gat_or.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "view-in-github"
   },
   "source": [
    "<a href=\"https://colab.research.google.com/github/njmarko/gat-or/blob/master/gat_or.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "1PQbehGvj_kX"
   },
   "source": [
    "# GATor\n",
    "\n",
    "GATor is a Graph Attention Network for object detection with relational reasoning.\n",
    "\n",
    "### Authors:\n",
    " - Marko Njegomir\n",
    " - Dušan Erdeljan"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## OUR LOWBUDGET GPU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'A100-SXM4-40GB'"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import torch\n",
    "torch.cuda.get_device_name(0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![My cpu is a nerual net processor. A learning computer](https://memegenerator.net/img/instances/62942383.jpg \"CPU\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "BtzUQf4vkMsG"
   },
   "source": [
    "### Installing dependencies and downloading dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "UKwSnuW4L-jj",
    "outputId": "5cf6fbad-364a-4b60-d106-827bc8d5d92c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1.10.1+cu113\n",
      "Collecting fiftyone\n",
      "  Downloading fiftyone-0.14.3-py3-none-any.whl (1.2 MB)\n",
      "     |████████████████████████████████| 1.2 MB 5.2 MB/s            \n",
      "\u001b[?25hCollecting aiofiles\n",
      "  Downloading aiofiles-0.8.0-py3-none-any.whl (13 kB)\n",
      "Requirement already satisfied: psutil in /opt/conda/lib/python3.7/site-packages (from fiftyone) (5.8.0)\n",
      "Requirement already satisfied: Pillow>=6.2 in /opt/conda/lib/python3.7/site-packages (from fiftyone) (8.4.0)\n",
      "Collecting voxel51-eta<0.7,>=0.6.2\n",
      "  Downloading voxel51_eta-0.6.2.2-py2.py3-none-any.whl (563 kB)\n",
      "     |████████████████████████████████| 563 kB 113.1 MB/s            \n",
      "\u001b[?25hCollecting pprintpp\n",
      "  Downloading pprintpp-0.4.0-py2.py3-none-any.whl (16 kB)\n",
      "Requirement already satisfied: pandas in /opt/conda/lib/python3.7/site-packages (from fiftyone) (1.3.5)\n",
      "Requirement already satisfied: future in /opt/conda/lib/python3.7/site-packages (from fiftyone) (0.18.2)\n",
      "Requirement already satisfied: pytz in /opt/conda/lib/python3.7/site-packages (from fiftyone) (2021.3)\n",
      "Collecting scikit-image\n",
      "  Downloading scikit_image-0.19.1-cp37-cp37m-manylinux_2_12_x86_64.manylinux2010_x86_64.whl (13.3 MB)\n",
      "     |████████████████████████████████| 13.3 MB 104.2 MB/s            \n",
      "\u001b[?25hCollecting fiftyone-db<0.4,>=0.3\n",
      "  Downloading fiftyone_db-0.3.0-py3-none-manylinux1_x86_64.whl (29.2 MB)\n",
      "     |████████████████████████████████| 29.2 MB 21.2 MB/s            \n",
      "\u001b[?25hRequirement already satisfied: numpy in /opt/conda/lib/python3.7/site-packages (from fiftyone) (1.19.5)\n",
      "Collecting eventlet\n",
      "  Downloading eventlet-0.33.0-py2.py3-none-any.whl (226 kB)\n",
      "     |████████████████████████████████| 226 kB 104.8 MB/s            \n",
      "\u001b[?25hCollecting tabulate\n",
      "  Downloading tabulate-0.8.9-py3-none-any.whl (25 kB)\n",
      "Collecting pymongo<4,>=3.11\n",
      "  Downloading pymongo-3.12.3-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (508 kB)\n",
      "     |████████████████████████████████| 508 kB 97.4 MB/s            \n",
      "\u001b[?25hCollecting fiftyone-brain<0.8,>=0.7.3\n",
      "  Downloading fiftyone_brain-0.7.3-cp37-cp37m-manylinux1_x86_64.whl (640 kB)\n",
      "     |████████████████████████████████| 640 kB 95.2 MB/s            \n",
      "\u001b[?25hRequirement already satisfied: scikit-learn in /opt/conda/lib/python3.7/site-packages (from fiftyone) (1.0.1)\n",
      "Requirement already satisfied: tornado<7,>=5.1.1 in /opt/conda/lib/python3.7/site-packages (from fiftyone) (6.1)\n",
      "Collecting xmltodict\n",
      "  Downloading xmltodict-0.12.0-py2.py3-none-any.whl (9.2 kB)\n",
      "Requirement already satisfied: packaging in /opt/conda/lib/python3.7/site-packages (from fiftyone) (21.3)\n",
      "Requirement already satisfied: matplotlib in /opt/conda/lib/python3.7/site-packages (from fiftyone) (3.5.1)\n",
      "Collecting motor<3,>=2.3\n",
      "  Downloading motor-2.5.1-py3-none-any.whl (55 kB)\n",
      "     |████████████████████████████████| 55 kB 7.3 MB/s             \n",
      "\u001b[?25hCollecting kaleido\n",
      "  Downloading kaleido-0.2.1-py2.py3-none-manylinux1_x86_64.whl (79.9 MB)\n",
      "     |████████████████████████████████| 79.9 MB 191 kB/s              \n",
      "\u001b[?25hRequirement already satisfied: PyYAML in /opt/conda/lib/python3.7/site-packages (from fiftyone) (6.0)\n",
      "Collecting universal-analytics-python3<2,>=1.0.1\n",
      "  Downloading universal_analytics_python3-1.1.1-py3-none-any.whl (10 kB)\n",
      "Requirement already satisfied: Jinja2 in /opt/conda/lib/python3.7/site-packages (from fiftyone) (3.0.3)\n",
      "Collecting boto3\n",
      "  Downloading boto3-1.20.42-py3-none-any.whl (131 kB)\n",
      "     |████████████████████████████████| 131 kB 99.6 MB/s            \n",
      "\u001b[?25hRequirement already satisfied: retrying in /opt/conda/lib/python3.7/site-packages (from fiftyone) (1.3.3)\n",
      "Requirement already satisfied: setuptools in /opt/conda/lib/python3.7/site-packages (from fiftyone) (59.6.0)\n",
      "Collecting opencv-python-headless\n",
      "  Downloading opencv_python_headless-4.5.5.62-cp36-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (47.7 MB)\n",
      "     |████████████████████████████████| 47.7 MB 96.3 MB/s            \n",
      "\u001b[?25hCollecting mongoengine==0.20.0\n",
      "  Downloading mongoengine-0.20.0-py3-none-any.whl (106 kB)\n",
      "     |████████████████████████████████| 106 kB 99.0 MB/s            \n",
      "\u001b[?25hRequirement already satisfied: argcomplete in /opt/conda/lib/python3.7/site-packages (from fiftyone) (1.12.3)\n",
      "Collecting Deprecated\n",
      "  Downloading Deprecated-1.2.13-py2.py3-none-any.whl (9.6 kB)\n",
      "Collecting plotly<5,>=4.14\n",
      "  Downloading plotly-4.14.3-py2.py3-none-any.whl (13.2 MB)\n",
      "     |████████████████████████████████| 13.2 MB 91.2 MB/s            \n",
      "\u001b[?25hRequirement already satisfied: scipy>=1.2.0 in /opt/conda/lib/python3.7/site-packages (from fiftyone-brain<0.8,>=0.7.3->fiftyone) (1.7.3)\n",
      "Requirement already satisfied: six in /opt/conda/lib/python3.7/site-packages (from plotly<5,>=4.14->fiftyone) (1.16.0)\n",
      "Collecting httpx>=0.10.0\n",
      "  Downloading httpx-0.21.3-py3-none-any.whl (83 kB)\n",
      "     |████████████████████████████████| 83 kB 4.3 MB/s             \n",
      "\u001b[?25hCollecting sortedcontainers\n",
      "  Downloading sortedcontainers-2.4.0-py2.py3-none-any.whl (29 kB)\n",
      "Collecting patool\n",
      "  Downloading patool-1.12-py2.py3-none-any.whl (77 kB)\n",
      "     |████████████████████████████████| 77 kB 12.3 MB/s            \n",
      "\u001b[?25hCollecting tzlocal\n",
      "  Downloading tzlocal-4.1-py3-none-any.whl (19 kB)\n",
      "Collecting dill\n",
      "  Downloading dill-0.3.4-py2.py3-none-any.whl (86 kB)\n",
      "     |████████████████████████████████| 86 kB 11.4 MB/s            \n",
      "\u001b[?25hRequirement already satisfied: python-dateutil in /opt/conda/lib/python3.7/site-packages (from voxel51-eta<0.7,>=0.6.2->fiftyone) (2.8.2)\n",
      "Collecting glob2\n",
      "  Downloading glob2-0.7.tar.gz (10 kB)\n",
      "  Preparing metadata (setup.py) ... \u001b[?25ldone\n",
      "\u001b[?25hCollecting ndjson\n",
      "  Downloading ndjson-0.3.1-py2.py3-none-any.whl (5.3 kB)\n",
      "Requirement already satisfied: importlib-metadata in /opt/conda/lib/python3.7/site-packages (from voxel51-eta<0.7,>=0.6.2->fiftyone) (4.9.0)\n",
      "Requirement already satisfied: requests in /opt/conda/lib/python3.7/site-packages (from voxel51-eta<0.7,>=0.6.2->fiftyone) (2.26.0)\n",
      "Collecting jmespath<1.0.0,>=0.7.1\n",
      "  Downloading jmespath-0.10.0-py2.py3-none-any.whl (24 kB)\n",
      "Collecting s3transfer<0.6.0,>=0.5.0\n",
      "  Downloading s3transfer-0.5.0-py3-none-any.whl (79 kB)\n",
      "     |████████████████████████████████| 79 kB 13.9 MB/s            \n",
      "\u001b[?25hCollecting botocore<1.24.0,>=1.23.42\n",
      "  Downloading botocore-1.23.42-py3-none-any.whl (8.5 MB)\n",
      "     |████████████████████████████████| 8.5 MB 94.9 MB/s            \n",
      "\u001b[?25hRequirement already satisfied: wrapt<2,>=1.10 in /opt/conda/lib/python3.7/site-packages (from Deprecated->fiftyone) (1.13.3)\n",
      "Collecting dnspython>=1.15.0\n",
      "  Downloading dnspython-2.2.0-py3-none-any.whl (266 kB)\n",
      "     |████████████████████████████████| 266 kB 104.9 MB/s            \n",
      "\u001b[?25hRequirement already satisfied: greenlet>=0.3 in /opt/conda/lib/python3.7/site-packages (from eventlet->fiftyone) (1.1.2)\n",
      "Requirement already satisfied: MarkupSafe>=2.0 in /opt/conda/lib/python3.7/site-packages (from Jinja2->fiftyone) (2.0.1)\n",
      "Requirement already satisfied: kiwisolver>=1.0.1 in /opt/conda/lib/python3.7/site-packages (from matplotlib->fiftyone) (1.3.2)\n",
      "Requirement already satisfied: pyparsing>=2.2.1 in /opt/conda/lib/python3.7/site-packages (from matplotlib->fiftyone) (3.0.6)\n",
      "Requirement already satisfied: cycler>=0.10 in /opt/conda/lib/python3.7/site-packages (from matplotlib->fiftyone) (0.11.0)\n",
      "Requirement already satisfied: fonttools>=4.22.0 in /opt/conda/lib/python3.7/site-packages (from matplotlib->fiftyone) (4.28.4)\n",
      "Requirement already satisfied: networkx>=2.2 in /opt/conda/lib/python3.7/site-packages (from scikit-image->fiftyone) (2.6.3)\n",
      "Collecting tifffile>=2019.7.26\n",
      "  Downloading tifffile-2021.11.2-py3-none-any.whl (178 kB)\n",
      "     |████████████████████████████████| 178 kB 103.8 MB/s            \n",
      "\u001b[?25hCollecting imageio>=2.4.1\n",
      "  Downloading imageio-2.14.1-py3-none-any.whl (3.3 MB)\n",
      "     |████████████████████████████████| 3.3 MB 100.2 MB/s            \n",
      "\u001b[?25hRequirement already satisfied: PyWavelets>=1.1.1 in /opt/conda/lib/python3.7/site-packages (from scikit-image->fiftyone) (1.2.0)\n",
      "Requirement already satisfied: threadpoolctl>=2.0.0 in /opt/conda/lib/python3.7/site-packages (from scikit-learn->fiftyone) (3.0.0)\n",
      "Requirement already satisfied: joblib>=0.11 in /opt/conda/lib/python3.7/site-packages (from scikit-learn->fiftyone) (1.0.1)\n",
      "Requirement already satisfied: urllib3<1.27,>=1.25.4 in /opt/conda/lib/python3.7/site-packages (from botocore<1.24.0,>=1.23.42->boto3->fiftyone) (1.26.7)\n",
      "Requirement already satisfied: sniffio in /opt/conda/lib/python3.7/site-packages (from httpx>=0.10.0->universal-analytics-python3<2,>=1.0.1->fiftyone) (1.2.0)\n",
      "Requirement already satisfied: certifi in /opt/conda/lib/python3.7/site-packages (from httpx>=0.10.0->universal-analytics-python3<2,>=1.0.1->fiftyone) (2021.10.8)\n",
      "Collecting httpcore<0.15.0,>=0.14.0\n",
      "  Downloading httpcore-0.14.5-py3-none-any.whl (67 kB)\n",
      "     |████████████████████████████████| 67 kB 11.6 MB/s            \n",
      "\u001b[?25hCollecting rfc3986[idna2008]<2,>=1.3\n",
      "  Downloading rfc3986-1.5.0-py2.py3-none-any.whl (31 kB)\n",
      "Requirement already satisfied: charset-normalizer in /opt/conda/lib/python3.7/site-packages (from httpx>=0.10.0->universal-analytics-python3<2,>=1.0.1->fiftyone) (2.0.9)\n",
      "Requirement already satisfied: typing-extensions>=3.6.4 in /opt/conda/lib/python3.7/site-packages (from importlib-metadata->voxel51-eta<0.7,>=0.6.2->fiftyone) (4.0.1)\n",
      "Requirement already satisfied: zipp>=0.5 in /opt/conda/lib/python3.7/site-packages (from importlib-metadata->voxel51-eta<0.7,>=0.6.2->fiftyone) (3.6.0)\n",
      "Requirement already satisfied: idna<4,>=2.5 in /opt/conda/lib/python3.7/site-packages (from requests->voxel51-eta<0.7,>=0.6.2->fiftyone) (3.1)\n",
      "Collecting backports.zoneinfo\n",
      "  Downloading backports.zoneinfo-0.2.1-cp37-cp37m-manylinux1_x86_64.whl (70 kB)\n",
      "     |████████████████████████████████| 70 kB 17.2 MB/s            \n",
      "\u001b[?25hCollecting pytz-deprecation-shim\n",
      "  Downloading pytz_deprecation_shim-0.1.0.post0-py2.py3-none-any.whl (15 kB)\n",
      "Requirement already satisfied: anyio==3.* in /opt/conda/lib/python3.7/site-packages (from httpcore<0.15.0,>=0.14.0->httpx>=0.10.0->universal-analytics-python3<2,>=1.0.1->fiftyone) (3.4.0)\n",
      "Collecting h11<0.13,>=0.11\n",
      "  Downloading h11-0.12.0-py3-none-any.whl (54 kB)\n",
      "     |████████████████████████████████| 54 kB 7.7 MB/s             \n",
      "\u001b[?25hCollecting tzdata\n",
      "  Downloading tzdata-2021.5-py2.py3-none-any.whl (339 kB)\n",
      "     |████████████████████████████████| 339 kB 103.8 MB/s            \n",
      "\u001b[?25hBuilding wheels for collected packages: glob2\n",
      "  Building wheel for glob2 (setup.py) ... \u001b[?25ldone\n",
      "\u001b[?25h  Created wheel for glob2: filename=glob2-0.7-py2.py3-none-any.whl size=9321 sha256=afc8291ba62a764f7d4a52ddeb36262fce0767687c5b99605b96e97ae2492e09\n",
      "  Stored in directory: /home/dusanerdeljan99/.cache/pip/wheels/d7/3c/72/5300602ba1269ffce8cff5dcf7b525fee756b57455903c37ba\n",
      "Successfully built glob2\n",
      "Installing collected packages: tzdata, rfc3986, jmespath, h11, backports.zoneinfo, tifffile, pytz-deprecation-shim, imageio, httpcore, botocore, tzlocal, tabulate, sortedcontainers, scikit-image, s3transfer, pymongo, patool, opencv-python-headless, ndjson, httpx, glob2, dnspython, dill, xmltodict, voxel51-eta, universal-analytics-python3, pprintpp, plotly, motor, mongoengine, kaleido, fiftyone-db, fiftyone-brain, eventlet, Deprecated, boto3, aiofiles, fiftyone\n",
      "Successfully installed Deprecated-1.2.13 aiofiles-0.8.0 backports.zoneinfo-0.2.1 boto3-1.20.42 botocore-1.23.42 dill-0.3.4 dnspython-2.2.0 eventlet-0.33.0 fiftyone-0.14.3 fiftyone-brain-0.7.3 fiftyone-db-0.3.0 glob2-0.7 h11-0.12.0 httpcore-0.14.5 httpx-0.21.3 imageio-2.14.1 jmespath-0.10.0 kaleido-0.2.1 mongoengine-0.20.0 motor-2.5.1 ndjson-0.3.1 opencv-python-headless-4.5.5.62 patool-1.12 plotly-4.14.3 pprintpp-0.4.0 pymongo-3.12.3 pytz-deprecation-shim-0.1.0.post0 rfc3986-1.5.0 s3transfer-0.5.0 scikit-image-0.19.1 sortedcontainers-2.4.0 tabulate-0.8.9 tifffile-2021.11.2 tzdata-2021.5 tzlocal-4.1 universal-analytics-python3-1.1.1 voxel51-eta-0.6.2.2 xmltodict-0.12.0\n",
      "Looking in links: https://data.pyg.org/whl/torch-1.10.0+cu113.html\n",
      "Collecting torch-scatter\n",
      "  Downloading https://data.pyg.org/whl/torch-1.10.0%2Bcu113/torch_scatter-2.0.9-cp37-cp37m-linux_x86_64.whl (7.9 MB)\n",
      "     |████████████████████████████████| 7.9 MB 1.5 MB/s            \n",
      "\u001b[?25hInstalling collected packages: torch-scatter\n",
      "Successfully installed torch-scatter-2.0.9\n",
      "Looking in links: https://data.pyg.org/whl/torch-1.10.0+cu113.html\n",
      "Collecting torch-sparse\n",
      "  Downloading https://data.pyg.org/whl/torch-1.10.0%2Bcu113/torch_sparse-0.6.12-cp37-cp37m-linux_x86_64.whl (3.5 MB)\n",
      "     |████████████████████████████████| 3.5 MB 4.9 MB/s            \n",
      "\u001b[?25hRequirement already satisfied: scipy in /opt/conda/lib/python3.7/site-packages (from torch-sparse) (1.7.3)\n",
      "Requirement already satisfied: numpy<1.23.0,>=1.16.5 in /opt/conda/lib/python3.7/site-packages (from scipy->torch-sparse) (1.19.5)\n",
      "Installing collected packages: torch-sparse\n",
      "Successfully installed torch-sparse-0.6.12\n",
      "Collecting torch-geometric\n",
      "  Downloading torch_geometric-2.0.3.tar.gz (370 kB)\n",
      "     |████████████████████████████████| 370 kB 5.1 MB/s            \n",
      "\u001b[?25h  Preparing metadata (setup.py) ... \u001b[?25ldone\n",
      "\u001b[?25hRequirement already satisfied: numpy in /opt/conda/lib/python3.7/site-packages (from torch-geometric) (1.19.5)\n",
      "Requirement already satisfied: tqdm in /opt/conda/lib/python3.7/site-packages (from torch-geometric) (4.62.3)\n",
      "Requirement already satisfied: scipy in /opt/conda/lib/python3.7/site-packages (from torch-geometric) (1.7.3)\n",
      "Requirement already satisfied: networkx in /opt/conda/lib/python3.7/site-packages (from torch-geometric) (2.6.3)\n",
      "Requirement already satisfied: scikit-learn in /opt/conda/lib/python3.7/site-packages (from torch-geometric) (1.0.1)\n",
      "Requirement already satisfied: requests in /opt/conda/lib/python3.7/site-packages (from torch-geometric) (2.26.0)\n",
      "Requirement already satisfied: pandas in /opt/conda/lib/python3.7/site-packages (from torch-geometric) (1.3.5)\n",
      "Collecting rdflib\n",
      "  Downloading rdflib-6.1.1-py3-none-any.whl (482 kB)\n",
      "     |████████████████████████████████| 482 kB 106.1 MB/s            \n",
      "\u001b[?25hCollecting googledrivedownloader\n",
      "  Downloading googledrivedownloader-0.4-py2.py3-none-any.whl (3.9 kB)\n",
      "Requirement already satisfied: jinja2 in /opt/conda/lib/python3.7/site-packages (from torch-geometric) (3.0.3)\n",
      "Requirement already satisfied: pyparsing in /opt/conda/lib/python3.7/site-packages (from torch-geometric) (3.0.6)\n",
      "Collecting yacs\n",
      "  Downloading yacs-0.1.8-py3-none-any.whl (14 kB)\n",
      "Requirement already satisfied: PyYAML in /opt/conda/lib/python3.7/site-packages (from torch-geometric) (6.0)\n",
      "Requirement already satisfied: MarkupSafe>=2.0 in /opt/conda/lib/python3.7/site-packages (from jinja2->torch-geometric) (2.0.1)\n",
      "Requirement already satisfied: python-dateutil>=2.7.3 in /opt/conda/lib/python3.7/site-packages (from pandas->torch-geometric) (2.8.2)\n",
      "Requirement already satisfied: pytz>=2017.3 in /opt/conda/lib/python3.7/site-packages (from pandas->torch-geometric) (2021.3)\n",
      "Requirement already satisfied: setuptools in /opt/conda/lib/python3.7/site-packages (from rdflib->torch-geometric) (59.6.0)\n",
      "Collecting isodate\n",
      "  Downloading isodate-0.6.1-py2.py3-none-any.whl (41 kB)\n",
      "     |████████████████████████████████| 41 kB 1.3 MB/s             \n",
      "\u001b[?25hRequirement already satisfied: importlib-metadata in /opt/conda/lib/python3.7/site-packages (from rdflib->torch-geometric) (4.9.0)\n",
      "Requirement already satisfied: charset-normalizer~=2.0.0 in /opt/conda/lib/python3.7/site-packages (from requests->torch-geometric) (2.0.9)\n",
      "Requirement already satisfied: idna<4,>=2.5 in /opt/conda/lib/python3.7/site-packages (from requests->torch-geometric) (3.1)\n",
      "Requirement already satisfied: urllib3<1.27,>=1.21.1 in /opt/conda/lib/python3.7/site-packages (from requests->torch-geometric) (1.26.7)\n",
      "Requirement already satisfied: certifi>=2017.4.17 in /opt/conda/lib/python3.7/site-packages (from requests->torch-geometric) (2021.10.8)\n",
      "Requirement already satisfied: threadpoolctl>=2.0.0 in /opt/conda/lib/python3.7/site-packages (from scikit-learn->torch-geometric) (3.0.0)\n",
      "Requirement already satisfied: joblib>=0.11 in /opt/conda/lib/python3.7/site-packages (from scikit-learn->torch-geometric) (1.0.1)\n",
      "Requirement already satisfied: six>=1.5 in /opt/conda/lib/python3.7/site-packages (from python-dateutil>=2.7.3->pandas->torch-geometric) (1.16.0)\n",
      "Requirement already satisfied: typing-extensions>=3.6.4 in /opt/conda/lib/python3.7/site-packages (from importlib-metadata->rdflib->torch-geometric) (4.0.1)\n",
      "Requirement already satisfied: zipp>=0.5 in /opt/conda/lib/python3.7/site-packages (from importlib-metadata->rdflib->torch-geometric) (3.6.0)\n",
      "Building wheels for collected packages: torch-geometric\n",
      "  Building wheel for torch-geometric (setup.py) ... \u001b[?25ldone\n",
      "\u001b[?25h  Created wheel for torch-geometric: filename=torch_geometric-2.0.3-py3-none-any.whl size=581969 sha256=0ce87cbb4cdd1634c42574c3d3250834efb686e773a91cc8007b97094940bdb8\n",
      "  Stored in directory: /home/dusanerdeljan99/.cache/pip/wheels/c3/2a/58/87ce0508964d4def1aafb92750c4f3ac77038efd1b9a89dcf5\n",
      "Successfully built torch-geometric\n",
      "Installing collected packages: isodate, yacs, rdflib, googledrivedownloader, torch-geometric\n",
      "Successfully installed googledrivedownloader-0.4 isodate-0.6.1 rdflib-6.1.1 torch-geometric-2.0.3 yacs-0.1.8\n",
      "Collecting pycocotools\n",
      "  Downloading pycocotools-2.0.4.tar.gz (106 kB)\n",
      "     |████████████████████████████████| 106 kB 5.3 MB/s            \n",
      "\u001b[?25h  Installing build dependencies ... \u001b[?25ldone\n",
      "\u001b[?25h  Getting requirements to build wheel ... \u001b[?25ldone\n",
      "\u001b[?25h  Preparing metadata (pyproject.toml) ... \u001b[?25ldone\n",
      "\u001b[?25hRequirement already satisfied: matplotlib>=2.1.0 in /opt/conda/lib/python3.7/site-packages (from pycocotools) (3.5.1)\n",
      "Requirement already satisfied: numpy in /opt/conda/lib/python3.7/site-packages (from pycocotools) (1.19.5)\n",
      "Requirement already satisfied: pyparsing>=2.2.1 in /opt/conda/lib/python3.7/site-packages (from matplotlib>=2.1.0->pycocotools) (3.0.6)\n",
      "Requirement already satisfied: kiwisolver>=1.0.1 in /opt/conda/lib/python3.7/site-packages (from matplotlib>=2.1.0->pycocotools) (1.3.2)\n",
      "Requirement already satisfied: fonttools>=4.22.0 in /opt/conda/lib/python3.7/site-packages (from matplotlib>=2.1.0->pycocotools) (4.28.4)\n",
      "Requirement already satisfied: cycler>=0.10 in /opt/conda/lib/python3.7/site-packages (from matplotlib>=2.1.0->pycocotools) (0.11.0)\n",
      "Requirement already satisfied: python-dateutil>=2.7 in /opt/conda/lib/python3.7/site-packages (from matplotlib>=2.1.0->pycocotools) (2.8.2)\n",
      "Requirement already satisfied: pillow>=6.2.0 in /opt/conda/lib/python3.7/site-packages (from matplotlib>=2.1.0->pycocotools) (8.4.0)\n",
      "Requirement already satisfied: packaging>=20.0 in /opt/conda/lib/python3.7/site-packages (from matplotlib>=2.1.0->pycocotools) (21.3)\n",
      "Requirement already satisfied: six>=1.5 in /opt/conda/lib/python3.7/site-packages (from python-dateutil>=2.7->matplotlib>=2.1.0->pycocotools) (1.16.0)\n",
      "Building wheels for collected packages: pycocotools\n",
      "  Building wheel for pycocotools (pyproject.toml) ... \u001b[?25ldone\n",
      "\u001b[?25h  Created wheel for pycocotools: filename=pycocotools-2.0.4-cp37-cp37m-linux_x86_64.whl size=326667 sha256=c9963704b9960e825b5432ff0c8fe7be3a80c1f78297c5af82a8c177dc3b3a27\n",
      "  Stored in directory: /home/dusanerdeljan99/.cache/pip/wheels/a3/5f/fa/f011e578cc76e1fc5be8dce30b3eb9fd00f337e744b3bba59b\n",
      "Successfully built pycocotools\n",
      "Installing collected packages: pycocotools\n",
      "Successfully installed pycocotools-2.0.4\n",
      "Collecting python-louvain\n",
      "  Downloading python-louvain-0.15.tar.gz (204 kB)\n",
      "     |████████████████████████████████| 204 kB 5.2 MB/s            \n",
      "\u001b[?25h  Preparing metadata (setup.py) ... \u001b[?25ldone\n",
      "\u001b[?25hRequirement already satisfied: networkx in /opt/conda/lib/python3.7/site-packages (from python-louvain) (2.6.3)\n",
      "Requirement already satisfied: numpy in /opt/conda/lib/python3.7/site-packages (from python-louvain) (1.19.5)\n",
      "Building wheels for collected packages: python-louvain\n",
      "  Building wheel for python-louvain (setup.py) ... \u001b[?25ldone\n",
      "\u001b[?25h  Created wheel for python-louvain: filename=python_louvain-0.15-py3-none-any.whl size=9413 sha256=1b07c496a4039c5c4a544efe9a3a4e2072366e20512804b57048092d0f98aab8\n",
      "  Stored in directory: /home/dusanerdeljan99/.cache/pip/wheels/c7/4c/ef/427db8777620f7fa12a6404f1527a39536a812f46a4d1044d3\n",
      "Successfully built python-louvain\n",
      "Installing collected packages: python-louvain\n",
      "Successfully installed python-louvain-0.15\n"
     ]
    }
   ],
   "source": [
    "!python -c \"import torch; print(torch.__version__)\"\n",
    "# These two command are used for googlecolab. They should not be run on cloud\n",
    "# !pip uninstall opencv-python-headless -y\n",
    "# !pip install opencv-python-headless==4.1.2.30\n",
    "\n",
    "!pip install fiftyone\n",
    "!pip install torch-scatter -f https://data.pyg.org/whl/torch-1.10.0+cu113.html\n",
    "!pip install torch-sparse -f https://data.pyg.org/whl/torch-1.10.0+cu113.html\n",
    "!pip install torch-geometric\n",
    "!pip install pycocotools\n",
    "!pip install python-louvain"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Cloning TorchVision utilities"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "id": "5FR4kjvz8heu"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cloning into 'vision'...\n",
      "remote: Enumerating objects: 82249, done.\u001b[K\n",
      "remote: Counting objects: 100% (17381/17381), done.\u001b[K\n",
      "remote: Compressing objects: 100% (1471/1471), done.\u001b[K\n",
      "remote: Total 82249 (delta 16007), reused 17040 (delta 15816), pack-reused 64868\n",
      "Receiving objects: 100% (82249/82249), 161.11 MiB | 55.57 MiB/s, done.\n",
      "Resolving deltas: 100% (69082/69082), done.\n",
      "Note: checking out '2f40a483d73018ae6e1488a484c5927f2b309969'.\n",
      "\n",
      "You are in 'detached HEAD' state. You can look around, make experimental\n",
      "changes and commit them, and you can discard any commits you make in this\n",
      "state without impacting any branches by performing another checkout.\n",
      "\n",
      "If you want to create a new branch to retain commits you create, you may\n",
      "do so (now or later) by using -b with the checkout command again. Example:\n",
      "\n",
      "  git checkout -b <new-branch-name>\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Download TorchVision repo to use some files from\n",
    "# references/detection\n",
    "!git clone https://github.com/pytorch/vision.git --branch  v0.8.2\n",
    "\n",
    "!cp vision/references/detection/utils.py ./\n",
    "!cp vision/references/detection/transforms.py ./\n",
    "!cp vision/references/detection/coco_eval.py ./\n",
    "!cp vision/references/detection/engine.py ./\n",
    "!cp vision/references/detection/coco_utils.py ./"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "id": "XN38W3_Wkccv"
   },
   "outputs": [],
   "source": [
    "import fiftyone as fo\n",
    "import fiftyone.zoo as foz\n",
    "import fiftyone.utils.coco as fouc\n",
    "from fiftyone import ViewField as VF\n",
    "import numpy as np\n",
    "from itertools import product\n",
    "from tqdm import tqdm\n",
    "import seaborn as sns\n",
    "import networkx as nx\n",
    "import community as community_louvain\n",
    "import torchvision\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "from torch.utils.data import DataLoader\n",
    "import torchvision.transforms as T\n",
    "from torchvision.transforms import functional as TVF\n",
    "from torchvision.ops import box_iou\n",
    "from torchvision.ops import boxes as box_ops\n",
    "from PIL import Image\n",
    "import matplotlib.pylab as plt\n",
    "import matplotlib.patches as patches\n",
    "from matplotlib.pyplot import figure\n",
    "import torch_geometric as pyg\n",
    "import torch_geometric.data as pyg_data\n",
    "from torch_geometric.nn import GATConv, GATv2Conv, MessagePassing\n",
    "\n",
    "from engine import train_one_epoch, evaluate\n",
    "from transforms import Compose, ToTensor\n",
    "import utils\n",
    "\n",
    "import math\n",
    "import os\n",
    "import os.path as osp"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Download COCO2017 train set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "PggQ7R73mOk9",
    "outputId": "58765394-a988-4e67-d20c-406e2885d847"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading split 'train' to '/home/dusanerdeljan99/fiftyone/coco-2017/train' if necessary\n",
      "Found annotations at '/home/dusanerdeljan99/fiftyone/coco-2017/raw/instances_train2017.json'\n",
      "Images already downloaded\n",
      "Existing download of split 'train' is sufficient\n",
      "Loading existing dataset 'coco-2017-train'. To reload from disk, either delete the existing dataset or provide a custom `dataset_name` to use\n"
     ]
    }
   ],
   "source": [
    "dataset = foz.load_zoo_dataset(\n",
    "    \"coco-2017\",\n",
    "    split=\"train\",\n",
    "#     max_samples=200,\n",
    ")\n",
    "dataset.persistent = True"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Download COCO2017 validation set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading split 'validation' to '/home/dusanerdeljan99/fiftyone/coco-2017/validation' if necessary\n",
      "Found annotations at '/home/dusanerdeljan99/fiftyone/coco-2017/raw/instances_val2017.json'\n",
      "Images already downloaded\n",
      "Existing download of split 'validation' is sufficient\n",
      "Loading existing dataset 'coco-2017-validation'. To reload from disk, either delete the existing dataset or provide a custom `dataset_name` to use\n"
     ]
    }
   ],
   "source": [
    "dataset_validation = foz.load_zoo_dataset(\n",
    "    \"coco-2017\",\n",
    "    split=\"validation\",\n",
    ")\n",
    "dataset_validation.persistent = True"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Inspect dataset in Fiftyone client "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# session = fo.launch_app(dataset_validation)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "8fNrbUQOm5Xc"
   },
   "source": [
    "### Visualizing co-occurrence matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "rpwK7r_HmWjJ",
    "outputId": "1e78822d-cd6e-4399-b761-90a0da8aa76d"
   },
   "outputs": [],
   "source": [
    "class_names = ['background','person','bicycle','car','motorcycle','airplane','bus','train','truck','boat','traffic light','fire hydrant','street sign','stop sign','parking meter','bench','bird','cat','dog','horse',\n",
    "'sheep','cow','elephant','bear','zebra','giraffe','hat','backpack','umbrella','shoe','eye glasses','handbag','tie','suitcase','frisbee','skis','snowboard','sports ball','kite','baseball bat','baseball glove','skateboard',\n",
    "'surfboard','tennis racket','bottle','plate','wine glass','cup','fork','knife','spoon','bowl','banana','apple','sandwich','orange','broccoli','carrot','hot dog','pizza','donut','cake','chair','couch','potted plant','bed',\n",
    "'mirror','dining table','window','desk','toilet','door','tv','laptop','mouse','remote','keyboard','cell phone','microwave','oven','toaster','sink','refrigerator','blender','book','clock','vase','scissors','teddy bear',\n",
    "'hair drier','toothbrush','hair brush'\n",
    "]\n",
    "\n",
    "class_name_to_idx = {elem:idx for idx, elem in enumerate(class_names)}\n",
    "n_classes = len(class_names)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 919
    },
    "id": "fA_VGr0roL5X",
    "outputId": "3c7548a9-a783-4fa1-86f4-7edf725b24b4"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 100% |███████████| 118287/118287 [6.1m elapsed, 0s remaining, 310.1 samples/s]       \n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1600x1600 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "co_occurrence_matrix = np.zeros((n_classes, n_classes))\n",
    "\n",
    "with fo.ProgressBar() as pb:\n",
    "  for sample in pb(dataset):\n",
    "    detections = sample.ground_truth.detections\n",
    "    for attention_pair in product(detections, detections):\n",
    "      i = class_name_to_idx[attention_pair[0].label]\n",
    "      j = class_name_to_idx[attention_pair[1].label]\n",
    "      co_occurrence_matrix[i][j] += 1\n",
    "      co_occurrence_matrix[j][i] += 1\n",
    "\n",
    "normalized_matrix = co_occurrence_matrix / np.amax(co_occurrence_matrix + 1.0, axis=1)\n",
    "\n",
    "figure(figsize=(20, 20), dpi=80)\n",
    "ax = sns.heatmap(normalized_matrix, linewidth=0.5, xticklabels=class_names, yticklabels=class_names, cmap=\"Blues\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "pmsMpYrnrDUA"
   },
   "source": [
    "### Defining PyTorch dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "id": "_OC7C5JevCgq"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "cuda\n"
     ]
    }
   ],
   "source": [
    "device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')\n",
    "print(device)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Resizing of input images and corresponding bbox targets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "id": "0f4N0hFYrmUq"
   },
   "outputs": [],
   "source": [
    "class Resize(object):\n",
    "  def __init__(self, size, interpolation=TVF.InterpolationMode.BILINEAR, max_size=None, antialias=None):\n",
    "    self.size = size\n",
    "    self.max_size = max_size\n",
    "    self.interpolation = interpolation\n",
    "    self.antialias = antialias\n",
    "\n",
    "  def __call__(self, image, target):\n",
    "    # Implemented rescaling in matrix form to speed up computation significantly \n",
    "    assert torch.is_tensor(image), \"Image is expected to be of type torch.tensor\"\n",
    "    original_shape = image.shape\n",
    "    image = TVF.resize(image, self.size, self.interpolation, self.max_size, self.antialias)\n",
    "    transformed_shape = image.shape\n",
    "    x_scale = transformed_shape[2] / original_shape[2]\n",
    "    y_scale = transformed_shape[1] / original_shape[1]\n",
    "    scale_tensor = torch.tensor([x_scale, y_scale]).repeat(target['boxes'].shape[0], 2)\n",
    "    target['boxes'] = torch.mul(target['boxes'], scale_tensor)\n",
    "    # Added ones to the target boxes width and high to ensure that bboxes would be at least 1x1 dim\n",
    "    target['boxes'] = target['boxes'] + torch.tensor([0.,0.,1.,1.])\n",
    "    return image, target"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Loading data in coco format"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "id": "OhBWtDFwqdEU"
   },
   "outputs": [],
   "source": [
    "class TorchCocoDataset(torch.utils.data.Dataset):\n",
    "  def __init__(self, fiftyone_dataset, transforms=None, gt_field=\"ground_truth\"):\n",
    "    self.samples = fiftyone_dataset\n",
    "    self.transforms = transforms\n",
    "    self.gt_field = gt_field\n",
    "    self.img_paths = self.samples.values(\"filepath\")\n",
    "    self.classes = class_names\n",
    "    self.labels_map_rev = class_name_to_idx\n",
    "\n",
    "  def __getitem__(self, idx):\n",
    "    img_path = self.img_paths[idx]\n",
    "    sample = self.samples[img_path]\n",
    "    metadata = sample.metadata\n",
    "    img = Image.open(img_path).convert(\"RGB\")\n",
    "\n",
    "    boxes = []\n",
    "    labels = []\n",
    "    area = []\n",
    "    iscrowd = []\n",
    "    detections = sample[self.gt_field].detections\n",
    "\n",
    "    for det in detections:\n",
    "      category_id = self.labels_map_rev[det.label]\n",
    "      coco_obj = fouc.COCOObject.from_label(\n",
    "          det, metadata, category_id=category_id,\n",
    "      )\n",
    "      x, y, w, h = coco_obj.bbox\n",
    "      boxes.append([x, y, x + w, y + h])\n",
    "      labels.append(coco_obj.category_id)\n",
    "      area.append(coco_obj.area)\n",
    "      iscrowd.append(coco_obj.iscrowd)\n",
    "\n",
    "    target = {}\n",
    "    target[\"boxes\"] = torch.as_tensor(boxes, dtype=torch.float32)\n",
    "    target[\"labels\"] = torch.as_tensor(labels, dtype=torch.int64)\n",
    "    target[\"image_id\"] = torch.as_tensor([idx])\n",
    "    target[\"area\"] = torch.as_tensor(area, dtype=torch.float32)\n",
    "    target[\"iscrowd\"] = torch.as_tensor(iscrowd, dtype=torch.int64)\n",
    "\n",
    "    if self.transforms:\n",
    "        img, target = self.transforms(img, target)\n",
    "\n",
    "    return img, target\n",
    "\n",
    "  def __len__(self):\n",
    "    return len(self.img_paths)\n",
    "\n",
    "  def get_classes(self):\n",
    "    return self.classes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Filtering out images with no bboxes and selecting train and validation split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "wKLyHQYlt-H_",
    "outputId": "4decc1b8-9a01-4da3-e3bd-aafd0ed9e7ac"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of training samples: 117266\n",
      "Number of test samples: 4952\n"
     ]
    }
   ],
   "source": [
    "has_bounding_box_view = dataset.filter_labels(\n",
    "        \"ground_truth\",\n",
    "        VF(\"label\").is_in(class_names)\n",
    ")\n",
    "has_bounding_box_view_validation = dataset_validation.filter_labels(\n",
    "        \"ground_truth\",\n",
    "        VF(\"label\").is_in(class_names)\n",
    ")\n",
    "\n",
    "train_transforms = Compose([ToTensor(), Resize((640, 480))])\n",
    "test_transforms = Compose([ToTensor()])\n",
    "\n",
    "# # For testing quickly in colab (peasants without v100 only)\n",
    "# train_view = has_bounding_box_view.take(10000, seed=51)\n",
    "# test_view = has_bounding_box_view.exclude([s.id for s in train_view]).take(1000)\n",
    "\n",
    "# all images\n",
    "train_view = has_bounding_box_view\n",
    "test_view = has_bounding_box_view_validation\n",
    "\n",
    "print(f\"Number of training samples: {len(train_view)}\")\n",
    "print(f\"Number of test samples: {len(test_view)}\")\n",
    "\n",
    "torch_dataset_train = TorchCocoDataset(train_view, train_transforms)\n",
    "torch_dataset_test = TorchCocoDataset(test_view, test_transforms)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Draw example image and bboxes to make sure that resize works"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "id": "-YHMZA5MuYh4"
   },
   "outputs": [],
   "source": [
    "def draw_image_with_bounding_boxes(image, target):\n",
    "  fig, ax = plt.subplots()\n",
    "  ax.imshow(image.permute(1, 2, 0))\n",
    "  for box in target['boxes']:\n",
    "    x = box[0]\n",
    "    y = box[1]\n",
    "    h = box[2] - x\n",
    "    w = box[3] - y\n",
    "    rect = patches.Rectangle((x, y), h, w, linewidth=1, edgecolor='r', facecolor='none')\n",
    "    ax.add_patch(rect)\n",
    "  plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 269
    },
    "id": "Bw-7xlh51xBd",
    "outputId": "76059553-0d40-46a9-e091-d6fb4a1c14e0"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "draw_image_with_bounding_boxes(*torch_dataset_train[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "RTJ22qC9xJXc"
   },
   "source": [
    "### Defining the model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Clustering algorithm for extracting global context data and adding it to the graph in form of virtual nodes and edges"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.cm as cm\n",
    "class GATorGPool(nn.Module):\n",
    "    \n",
    "    def __init__(self):\n",
    "        super(GATorGPool, self).__init__()\n",
    "        \n",
    "    def forward(self, graph):\n",
    "#         print(graph)\n",
    "#         print(graph.num_nodes)\n",
    "#         print(graph.num_edges)\n",
    "#         nx_graph = pyg.utils.to_networkx(graph, to_undirected=True)\n",
    "# #         nx.draw(nx_graph)\n",
    "# #         levels = community_louvain.generate_dendrogram(nx_graph)\n",
    "#         print(nx_graph)\n",
    "#         # compute the best partition\n",
    "#         partition = community_louvain.best_partition(nx_graph)\n",
    "#         print(max(partition.values()))\n",
    "#         nx.draw(nx_graph)\n",
    "#         print(partition)\n",
    "#         input(\"Press any key to continue\")\n",
    "#         # draw the graph\n",
    "#         pos = nx.spring_layout(nx_graph)\n",
    "#         # color the nodes according to their partition\n",
    "#         cmap = cm.get_cmap('viridis', max(partition.values()) + 1)\n",
    "#         nx.draw_networkx_nodes(nx_graph, pos, partition.keys(), node_size=40,\n",
    "#                                cmap=cmap, node_color=list(partition.values()))\n",
    "#         nx.draw_networkx_edges(nx_graph, pos, alpha=0.5)\n",
    "#         plt.show()\n",
    "#         print(levels)\n",
    "#         for level in levels:\n",
    "#             print(f\"==============================\\n{level}\\n=======================================\")\n",
    "        return graph\n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{0: 0, 1: 5, 2: 2, 3: 3, 4: 4, 5: 3, 6: 5, 7: 24, 8: 7, 9: 8, 10: 9, 11: 5, 12: 5, 13: 10, 14: 24, 15: 5, 16: 11, 17: 7, 18: 5, 19: 12, 20: 13, 21: 14, 22: 0, 23: 15, 24: 5, 25: 9, 26: 5, 27: 8, 28: 17, 29: 0, 30: 19, 31: 20, 32: 21, 33: 22, 34: 5, 35: 15, 36: 23, 37: 24, 38: 5, 39: 24, 40: 9, 41: 25, 42: 26, 43: 27, 44: 28, 45: 7, 46: 4, 47: 29, 48: 30, 49: 8, 50: 7, 51: 31, 52: 32, 53: 33, 54: 34, 55: 15, 56: 35, 57: 30, 58: 36, 59: 37, 60: 38, 61: 5, 62: 15, 63: 36, 64: 24, 65: 8, 66: 9, 67: 20, 68: 39, 69: 40, 70: 7, 71: 41, 72: 0, 73: 42, 74: 36, 75: 8, 76: 0, 77: 15, 78: 24, 79: 12, 80: 0, 81: 0, 82: 16, 83: 5, 84: 1, 85: 18, 86: 15, 87: 0, 88: 8, 89: 9, 90: 7, 91: 4, 92: 6, 93: 4, 94: 0, 95: 4, 96: 30, 97: 24, 98: 24, 99: 5}\n"
     ]
    }
   ],
   "source": [
    "G = nx.erdos_renyi_graph(100, 0.01)\n",
    "partion = community_louvain.best_partition(G)\n",
    "print(partion)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Building the graph from node embeddings and bboxes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "id": "kftiJY9CICT9"
   },
   "outputs": [],
   "source": [
    "class GATorGraph(object):\n",
    "  def __init__(self, on_data_created=None, method_name=\"simple_iou\"):\n",
    "    super(GATorGraph, self).__init__()\n",
    "#     self.graph_pool = GATorGPool()\n",
    "    self.on_data_created = on_data_created\n",
    "    self.edge_detection_method = self.select_edge_detetion_method(method_name)\n",
    "\n",
    "  def select_edge_detetion_method(self, method_name):\n",
    "    if method_name == \"simple_iou\":\n",
    "      return self.simple_box_iou_edge_detection\n",
    "    elif method_name == \"complex_distance_semantic\":\n",
    "      return self.complex_distance_semantic_edge_detection\n",
    "\n",
    "  \n",
    "  @torch.no_grad()\n",
    "  def __call__(self, x, batched_boxes, image_shapes):\n",
    "    # print(f\"GATorGraph: x shape before: {x.shape}\")\n",
    "    edges = []\n",
    "    for index, (boxes, image_shape) in enumerate(zip(batched_boxes, image_shapes)):\n",
    "      edge_list = self.edge_detection_method(boxes, x[index], image_shape)\n",
    "      # print(f\"GATorGraph: x shape in for loop: {x[index].shape}\") \n",
    "      # Deterministic edge creatinon (remove contiguous for stochastic sampling)\n",
    "      data = pyg_data.Data(x=x[index], edge_index=edge_list.t().contiguous(), boxes=boxes, image_shape=image_shape)\n",
    "#       data = self.graph_pool(data)\n",
    "#       print(f\"GATorGraph: data shape: {data}\")\n",
    "      if self.on_data_created:\n",
    "        self.on_data_created(data, index)\n",
    "      edges.append(data)\n",
    "\n",
    "    batched_edge_lists = pyg_data.Batch.from_data_list(edges)\n",
    "    # print(f\"GATorGraph: x shape after: {x.shape}\")\n",
    "    # print(f\"GATorGraph: data num nodes: {edges[0]}\")\n",
    "    return batched_edge_lists\n",
    "\n",
    "  def simple_box_iou_edge_detection(self, bboxes, _x, image_shape):\n",
    "      ious = box_iou(bboxes, bboxes)\n",
    "      ious = torch.tril(ious, diagonal=-1)\n",
    "      return (ious > 0.5).nonzero(as_tuple=False)\n",
    "\n",
    "\n",
    "  def complex_distance_semantic_edge_detection(self, bboxes, x, image_shape):\n",
    "\n",
    "    # central_boxes dim = (512,4)\n",
    "\n",
    "    # filter out overlapping boxes\n",
    "    ious = box_iou(bboxes, bboxes)\n",
    "    filter_mask = torch.where(ious > 0.8, 0, 1)\n",
    "    \n",
    "    semantic_similarity = self.calcualte_semantic_similarity(x, filter_mask)\n",
    "    # print(f\"GATorGraph: semantic_similarity shape: {semantic_similarity.shape}\")\n",
    "    # print(f\"GATorGraph: semantic_similarity count before filter: {len(semantic_similarity.nonzero(as_tuple=True)[0])}\")\n",
    "    spatial_relationships = self.calculate_complex_spatial_relationships(bboxes, image_shape, filter_mask)\n",
    "    # print(f\"GATorGraph: spatial_relationships shape: {semantic_similarity.shape}\")\n",
    "    # print(f\"GATorGraph: spatial_relationships count before filter: {len(spatial_relationships.nonzero(as_tuple=True)[0])}\")\n",
    "\n",
    "    edge_score = torch.add(semantic_similarity, spatial_relationships)\n",
    "    # print(f\"GATorGraph: edge_score shape: {semantic_similarity.shape}\")\n",
    "\n",
    "#     # filter out overlapping boxes\n",
    "#     ious = box_iou(bboxes, bboxes)\n",
    "#     filter_mask = torch.where(ious > 0.9, 0, 1)\n",
    "#     # print(f\"GATorGraph: edge_score count before filter: {len(edge_score.nonzero(as_tuple=True)[0])}\")\n",
    "#     filtered_scores = torch.mul(edge_score, filter_mask)\n",
    "    # print(f\"GATorGraph: edge_score count after filter: {len(filtered_scores.nonzero(as_tuple=True)[0])}\")\n",
    "\n",
    "    # chose only upper triange so there would be no duplicate edges because the graph is undirected\n",
    "#     edge_score = torch.tril(edge_score, diagonal=-1)\n",
    "\n",
    "    # TODO we should multiply the cosine_similarity and gauss_distance matrices\n",
    "    return edge_score.nonzero(as_tuple=False)\n",
    "\n",
    "  def calculate_complex_spatial_relationships(self, bboxes, image_shape, filter_mask, distance_threshold=0.5, topk=12):\n",
    "    # transform central_boxes coordinates into center coordinates\n",
    "    # TODO make a distance trheshold parameter\n",
    "    detection_distance = np.clip(np.sqrt(image_shape[0]**2 + image_shape[0]**2)/3, 200, 600)\n",
    "#     detection_distance = 250\n",
    "    centers = self.calculate_bbox_center(bboxes)\n",
    "    euclidean_distance = torch.cdist(bboxes, bboxes)\n",
    "    # constant to reduce the distance values\n",
    "    scaling_constant = np.log(distance_threshold)/(detection_distance)\n",
    "    gauss_distance = torch.exp(torch.mul(scaling_constant, euclidean_distance))\n",
    "\n",
    "    filtered_gauss_distances = torch.mul(gauss_distance, filter_mask)\n",
    "    \n",
    "    _, top_k_distances_indexes = torch.topk(filtered_gauss_distances, k=topk, dim= 1, largest=True)\n",
    "    dist_matrix = torch.scatter(torch.zeros(euclidean_distance.shape, device=device),1, top_k_distances_indexes, 1)\n",
    "    return dist_matrix\n",
    "\n",
    "  def calcualte_semantic_similarity(self, x, filter_mask):\n",
    "    # TODO we should generate a matrix of cosine similarities\n",
    "    # x shape = (512,1024)\n",
    "    # bboxes shape = (512, 4)\n",
    "    # out shape = (512, 512)\n",
    "    x_norm = x/x.norm(dim=1).unsqueeze(1)\n",
    "    cosine_similarity = torch.matmul(x_norm,x_norm.T)\n",
    "    # torch.set_printoptions(profile=\"default\")\n",
    "    # print(f\"GATorGraph: semantic_similarity matrix\\n: {cosine_similarity}\")\n",
    "    filtered_cosine_similarity = torch.mul(cosine_similarity, filter_mask)\n",
    "    cosine_similarity = torch.where(filtered_cosine_similarity > 0.9223145926, 1, 0)\n",
    "    return cosine_similarity\n",
    "\n",
    "  def calculate_bbox_center(self, bbox):\n",
    "    x = (bbox[:,0] + bbox[:,2])/2\n",
    "    # print(x)\n",
    "    y = (bbox[:,1] + bbox[:,3])/2\n",
    "    # print(y)\n",
    "    res = torch.concat((x.unsqueeze(1), y.unsqueeze(1)), dim=1)\n",
    "    # print(res)\n",
    "    return res\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Experimenting with various functions, operations and dimensions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "gDmUukN4v9vc",
    "outputId": "d3163785-97b9-4c35-8521-e701da740295"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([[0, 0, 0],\n",
      "        [4, 0, 0],\n",
      "        [7, 8, 0]])\n",
      "tensor([[  1.,  22.,  33.],\n",
      "        [ 44.,   5.,   6.],\n",
      "        [  7.,   8., 251.]])\n",
      "-0.0027725887222397813\n",
      "tensor([[0.9972, 0.9408, 0.9126],\n",
      "        [0.8852, 0.9862, 0.9835],\n",
      "        [0.9808, 0.9781, 0.4986]])\n",
      "tensor([[1., 1., 0.],\n",
      "        [0., 1., 1.],\n",
      "        [1., 1., 0.]])\n"
     ]
    }
   ],
   "source": [
    "t1 = torch.tensor([[1,2,3,4],[5,6,7,8]], dtype=torch.float)\n",
    "# print(t1)\n",
    "t2 = torch.tensor([[1,2,3,4],[5,6,7,8]], dtype=torch.float)\n",
    "# print(t2)\n",
    "# columns = (t1[:,2]/2 + t2[:,3]/2).unsqueeze(1)\n",
    "# print(columns)\n",
    "# x = (t1[:,0] + t1[:,2])/2\n",
    "# print(x)\n",
    "# y = (t1[:,1] + t1[:,3])/2\n",
    "# print(y)\n",
    "# res = torch.concat((x.unsqueeze(1), y.unsqueeze(1)), dim=1)\n",
    "# print(res)\n",
    "norm1 = t1/t1.norm(dim=1).unsqueeze(1)\n",
    "norm2 = t2/t2.norm(dim=1).unsqueeze(1)\n",
    "m = torch.matmul(norm1,norm1.T)\n",
    "# print(m)\n",
    "# sim = torch.cosine_similarity(torch.tensor([[1,2,3,4]],dtype=torch.float), torch.tensor([[5,6,7,8]],dtype=torch.float))\n",
    "# print(sim)\n",
    "# print(torch.mul(t1,t2))\n",
    "print(torch.tril(torch.tensor([[1,2,3],[4,5,6],[7,8,9]]), diagonal=-1))\n",
    "t3 = torch.tensor([[1,22,33],[44,5,6],[7,8,251]],dtype=torch.float)\n",
    "print(t3)\n",
    "k = np.log(0.5)/(250)\n",
    "print(k)\n",
    "gauss_distance = torch.exp(torch.mul(k, t3))\n",
    "print(gauss_distance)\n",
    "_, top_k_distances_indexes = torch.topk(gauss_distance,k=2, dim= 1, largest=True)\n",
    "dist_matrix = torch.scatter(torch.zeros(gauss_distance.shape),1, top_k_distances_indexes, 1.)\n",
    "print(dist_matrix)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Sanity check"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "CNwHuHzgynCp",
    "outputId": "4f86e66c-d1c0-412f-c34b-3c8617856f55"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AAAAAAAAAA\n"
     ]
    }
   ],
   "source": [
    "# Do not remove! Code doesn't work without it\n",
    "print(\"AAAAAAAAAA\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Main GAT stack module which processes image graph"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "id": "F5ViSKT680Xt"
   },
   "outputs": [],
   "source": [
    "class GATor(nn.Module):\n",
    "  def __init__(self, input_dim, hidden_dim, output_dim, layers, heads, dropout, edge_dim=None, layer_type=\"GATv1\"):\n",
    "    super(GATor, self).__init__()\n",
    "    self.convs = nn.ModuleList()\n",
    "    self.gnn_layer = self.choose_gnn_layer(layer_type)\n",
    "    self.convs.append(self.gnn_layer(in_channels=input_dim, out_channels=hidden_dim, heads=heads, edge_dim=edge_dim))\n",
    "    for l in range(layers-1):\n",
    "      self.convs.append(self.gnn_layer(in_channels=heads * hidden_dim, out_channels=hidden_dim, heads=heads, edge_dim=edge_dim))\n",
    "    # post-message-passing\n",
    "    self.post_mp = nn.Sequential(\n",
    "      nn.Linear(heads * hidden_dim, hidden_dim), nn.Dropout(dropout), \n",
    "      nn.Linear(hidden_dim, output_dim))\n",
    "    self.dropout = dropout\n",
    "    self.num_layers = layers\n",
    "\n",
    "  def choose_gnn_layer(self, layer_type):\n",
    "    if layer_type == \"GATv1\":\n",
    "      return GATConv\n",
    "    elif layer_type == \"GATv2\":\n",
    "      return GATv2Conv\n",
    "    \n",
    "\n",
    "  def forward(self, data):\n",
    "    x, edge_index, batch = data.x, data.edge_index, data.batch\n",
    "    # print(f\"GATor x shape: {x.shape}\")  \n",
    "    # print(f\"GATor edge_index shape: {edge_index.shape}\")\n",
    "    for i in range(self.num_layers):\n",
    "      # print(f\"GATor current layer index: {i}\")\n",
    "      # print(f\"GATor current x shape: {x.shape}\")\n",
    "      x = self.convs[i](x, edge_index)\n",
    "      x = F.relu(x)\n",
    "      x = F.dropout(x, p=self.dropout, training=self.training)\n",
    "    x = self.post_mp(x)\n",
    "    return x"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Wrapper module which caches bboxes and image shapes that are required for graph creation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "id": "PNU5fMCOBbbL"
   },
   "outputs": [],
   "source": [
    "class MultiScaleRoIAlignWrapper(nn.Module):\n",
    "  def __init__(self, roi_align):\n",
    "    super(MultiScaleRoIAlignWrapper, self).__init__()\n",
    "    self.roi_align = roi_align\n",
    "\n",
    "  def forward(self, x, boxes, image_shapes):\n",
    "    # print(f\"MULTI SCALE ROI: X dictionary: {x}\")\n",
    "    # print(f\"MULTI SCALE ROI: boxes len: {len(boxes)}\")\n",
    "    # print(f\"MULTI SCALE ROI: boxe[0] shape: {boxes[0].shape}\")\n",
    "    # print(f\"MULTI SCALE ROI: image_shapes len: {len(image_shapes)}\")\n",
    "    # for a in x:\n",
    "    #   print(f\"MULTI SCALE ROI: shapes o elements in X: {a}\")\n",
    "    ret_val = self.roi_align(x, boxes, image_shapes)\n",
    "    # print(f\"MULTI SCALE ROI: retval: {ret_val}\")\n",
    "    # print(f\"MULTI SCALE ROI: retval shape: {ret_val.shape}\")\n",
    "    cached['boxes'] = boxes\n",
    "    cached['image_shapes'] = image_shapes\n",
    "    return ret_val"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Wrapper module which initializes graph creating that will be placed after the model outputs final embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "id": "6sm1fgPfFTsX"
   },
   "outputs": [],
   "source": [
    "class TwoMLPHeadWrapper(nn.Module):\n",
    "  def __init__(self, box_head, gator_graph, gator):\n",
    "    super(TwoMLPHeadWrapper, self).__init__()\n",
    "    self.box_head = box_head\n",
    "    self.gator_graph = gator_graph\n",
    "    self.gator = gator\n",
    "\n",
    "  def forward(self, x):\n",
    "    # print(f\"TwoMLPHeadWrapper: x shape: {x.shape}\")\n",
    "    node_features = self.box_head(x)\n",
    "    # print(f\"TwoMLPHeadWrapper: node_features shape: {node_features.shape}\")\n",
    "    B = len(cached['boxes'])\n",
    "    graph = self.gator_graph(node_features.view(B,-1, node_features.shape[1]), cached['boxes'], cached['image_shapes'])\n",
    "    ret_val = self.gator(graph)\n",
    "    # print(f\"TwoMLPHeadWrapper: Evo me\")\n",
    "    return ret_val"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Graph message passing for custom layer building"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "id": "DZ2EgdT6-x3_"
   },
   "outputs": [],
   "source": [
    "class GATorLayer(MessagePassing):\n",
    "\n",
    "    def __init__(self, in_channels, out_channels, heads = 2,\n",
    "                 negative_slope = 0.2, dropout = 0., **kwargs):\n",
    "        super(GAT, self).__init__(node_dim=0, **kwargs)\n",
    "\n",
    "        self.in_channels = in_channels\n",
    "        self.out_channels = out_channels\n",
    "        self.heads = heads\n",
    "        self.negative_slope = negative_slope\n",
    "        self.dropout = dropout\n",
    "\n",
    "        self.lin_l = None\n",
    "        self.lin_r = None\n",
    "        self.att_l = None\n",
    "        self.att_r = None\n",
    "        \n",
    "\n",
    "\n",
    "        self.lin_l = Linear(in_channels, out_channels*self.heads)\n",
    "\n",
    "\n",
    "        self.lin_r = self.lin_l\n",
    "\n",
    "\n",
    "        self.att_l = Parameter(torch.Tensor(1, self.heads, out_channels))\n",
    "        self.att_r = Parameter(torch.Tensor(1, self.heads, out_channels))\n",
    "        self.attention = Parameter(torch.Tensor(1, self.heads, out_channels))\n",
    "\n",
    "\n",
    "        self.reset_parameters()\n",
    "\n",
    "    def reset_parameters(self):\n",
    "        nn.init.xavier_uniform_(self.lin_l.weight)\n",
    "        nn.init.xavier_uniform_(self.lin_r.weight)\n",
    "        nn.init.xavier_uniform_(self.att_l)\n",
    "        nn.init.xavier_uniform_(self.att_r)\n",
    "\n",
    "    def forward(self, x, edge_index, size = None):\n",
    "        \n",
    "        H, C = self.heads, self.out_channels\n",
    "\n",
    "        # print(f\"\\n========Forward========\")\n",
    "        # print(f\"x shape: {x.shape}\")\n",
    "        # print(f\"Edge index shape {edge_index.shape}\")\n",
    "        # print(f\"att_l shape: {self.att_l.shape}\")\n",
    "        # print(f\"att_r shape: {self.att_r.shape}\")\n",
    "\n",
    "\n",
    "        wh_l = self.lin_r(x).view(-1, H, C)\n",
    "        wh_r = self.lin_r(x).view(-1, H, C)\n",
    "        # print(f\"whl shape: {wh_l.shape}\")\n",
    "        # print(f\"whr shape: {wh_r.shape}\")\n",
    "\n",
    "        alpha_l = wh_l\n",
    "        alpha_r = wh_r\n",
    "        # print(f\"alpha_l shape: {alpha_l.shape}\")\n",
    "        # print(f\"alpha_r shape: {alpha_r.shape}\")\n",
    "\n",
    "        out = self.propagate(edge_index,x=(wh_l, wh_r), size=size, alpha=(alpha_l, alpha_r))\n",
    "        # print(f\"Shape of propagate out: {out.shape}\")\n",
    "\n",
    "        out = out.view(-1, H*C)\n",
    "        # print(f\"Shape of out after shape transformation to [N, H*C]: {out.shape}\")\n",
    "\n",
    "        return out\n",
    "\n",
    "\n",
    "    def message(self, x_j, alpha_j, alpha_i, index, ptr, size_i):\n",
    "\n",
    "        # Because TwoMLPHead layer from FasterRCNN gives embeddings for region proposals\n",
    "        # that are generated to be ready for classification and regression\n",
    "        # we can just use cosine similarity between the nodes to determine semantic similarity\n",
    "        # If semantic similarity is high, the central node i should pay more attention to that message\n",
    "\n",
    "        # print(f\"\\n========Message funnction========\")\n",
    "        # print(f\"x_j shape: {x_j.shape}\")\n",
    "        # print(f\"Alpha_j shape: {alpha_j.shape}\")\n",
    "        # print(f\"Alpha_i shape: {alpha_i.shape}\")\n",
    "        # print(f\"Index: {index.shape}\")\n",
    "        # print(f\"ptr shape: {ptr}\")\n",
    "        # print(f\"size_i: {size_i}\")\n",
    "\n",
    "        # TODO select just rows from x_j based on index\n",
    "        # alpha_i should be the central node embedding\n",
    "        cos_sim = torch.nn.CosineSimilarity(alpha_i, x_j)\n",
    "\n",
    "        attention = torch.mul(cos_sim, x_j)\n",
    "\n",
    "        attention = softmax(attention, ptr if ptr else index)\n",
    "        attention = F.dropout(att, self.dropout)\n",
    "\n",
    "        out = attention\n",
    "        # print(f\"out shape from message function should be [E, H, C]: {out.shape}\")\n",
    "\n",
    "        return out\n",
    "\n",
    "\n",
    "    def aggregate(self, inputs, index, dim_size = None):\n",
    "\n",
    "        out = torch_scatter.scatter(inputs, index = index, dim = self.node_dim, dim_size = dim_size, reduce = \"sum\")\n",
    "    \n",
    "        return out"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Building the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "id": "1ARkkb2hIpVg"
   },
   "outputs": [],
   "source": [
    "def build_model():\n",
    "  model = torchvision.models.detection.fasterrcnn_resnet50_fpn(pretrained=True)\n",
    "  gator_graph = GATorGraph(method_name=\"complex_distance_semantic\")\n",
    "  gator = GATor(input_dim=1024, hidden_dim=1024, output_dim=1024, layers=2, heads=1, dropout=0.5, edge_dim=None, layer_type=\"GATv2\")\n",
    "  # gat = GATModule(model.roi_heads.box_head, gator_graph)\n",
    "  # model.rpn = RPNWrapper(model.rpn)\n",
    "  model.roi_heads.box_roi_pool = MultiScaleRoIAlignWrapper(model.roi_heads.box_roi_pool)\n",
    "  model.roi_heads.box_head = TwoMLPHeadWrapper(model.roi_heads.box_head, gator_graph, gator)\n",
    "  # model.box_predictor.reset_parameters()\n",
    "  # model.roi_heads.box_head = gat\n",
    "  return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000,
     "referenced_widgets": [
      "10dc5c906a68455e8ab9966d2ecc1bbe",
      "38d76333d35b4823b6de75f6e30d159a",
      "314b12ed64c044208a8ff2a6e4fdc59c",
      "d5ae0d02c3b844bdafdec479bdb159e7",
      "452cc2936c62488aae19aab38eca67f5",
      "ab2c3648905e4c769249c935ceaf5ec5",
      "cf47b4248ecd459280bbe9f9c313c035",
      "e013dd58a6274902a2df85eb8c00e31d",
      "8ccf33413b0a41f5b6dd6765f7689fbc",
      "69fc04dc526f4d7188b2e0ea62402d8c",
      "e7cdbe0e4c024ca8a50aa6b5a8606280"
     ]
    },
    "id": "kNaBSo_AxikF",
    "outputId": "6b1b3a49-2391-4398-bc03-6c7b5097e454"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "FasterRCNN(\n",
       "  (transform): GeneralizedRCNNTransform(\n",
       "      Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])\n",
       "      Resize(min_size=(800,), max_size=1333, mode='bilinear')\n",
       "  )\n",
       "  (backbone): BackboneWithFPN(\n",
       "    (body): IntermediateLayerGetter(\n",
       "      (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)\n",
       "      (bn1): FrozenBatchNorm2d(64, eps=0.0)\n",
       "      (relu): ReLU(inplace=True)\n",
       "      (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
       "      (layer1): Sequential(\n",
       "        (0): Bottleneck(\n",
       "          (conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn1): FrozenBatchNorm2d(64, eps=0.0)\n",
       "          (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn2): FrozenBatchNorm2d(64, eps=0.0)\n",
       "          (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn3): FrozenBatchNorm2d(256, eps=0.0)\n",
       "          (relu): ReLU(inplace=True)\n",
       "          (downsample): Sequential(\n",
       "            (0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "            (1): FrozenBatchNorm2d(256, eps=0.0)\n",
       "          )\n",
       "        )\n",
       "        (1): Bottleneck(\n",
       "          (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn1): FrozenBatchNorm2d(64, eps=0.0)\n",
       "          (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn2): FrozenBatchNorm2d(64, eps=0.0)\n",
       "          (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn3): FrozenBatchNorm2d(256, eps=0.0)\n",
       "          (relu): ReLU(inplace=True)\n",
       "        )\n",
       "        (2): Bottleneck(\n",
       "          (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn1): FrozenBatchNorm2d(64, eps=0.0)\n",
       "          (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn2): FrozenBatchNorm2d(64, eps=0.0)\n",
       "          (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn3): FrozenBatchNorm2d(256, eps=0.0)\n",
       "          (relu): ReLU(inplace=True)\n",
       "        )\n",
       "      )\n",
       "      (layer2): Sequential(\n",
       "        (0): Bottleneck(\n",
       "          (conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn1): FrozenBatchNorm2d(128, eps=0.0)\n",
       "          (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "          (bn2): FrozenBatchNorm2d(128, eps=0.0)\n",
       "          (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn3): FrozenBatchNorm2d(512, eps=0.0)\n",
       "          (relu): ReLU(inplace=True)\n",
       "          (downsample): Sequential(\n",
       "            (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
       "            (1): FrozenBatchNorm2d(512, eps=0.0)\n",
       "          )\n",
       "        )\n",
       "        (1): Bottleneck(\n",
       "          (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn1): FrozenBatchNorm2d(128, eps=0.0)\n",
       "          (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn2): FrozenBatchNorm2d(128, eps=0.0)\n",
       "          (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn3): FrozenBatchNorm2d(512, eps=0.0)\n",
       "          (relu): ReLU(inplace=True)\n",
       "        )\n",
       "        (2): Bottleneck(\n",
       "          (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn1): FrozenBatchNorm2d(128, eps=0.0)\n",
       "          (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn2): FrozenBatchNorm2d(128, eps=0.0)\n",
       "          (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn3): FrozenBatchNorm2d(512, eps=0.0)\n",
       "          (relu): ReLU(inplace=True)\n",
       "        )\n",
       "        (3): Bottleneck(\n",
       "          (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn1): FrozenBatchNorm2d(128, eps=0.0)\n",
       "          (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn2): FrozenBatchNorm2d(128, eps=0.0)\n",
       "          (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn3): FrozenBatchNorm2d(512, eps=0.0)\n",
       "          (relu): ReLU(inplace=True)\n",
       "        )\n",
       "      )\n",
       "      (layer3): Sequential(\n",
       "        (0): Bottleneck(\n",
       "          (conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn1): FrozenBatchNorm2d(256, eps=0.0)\n",
       "          (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "          (bn2): FrozenBatchNorm2d(256, eps=0.0)\n",
       "          (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn3): FrozenBatchNorm2d(1024, eps=0.0)\n",
       "          (relu): ReLU(inplace=True)\n",
       "          (downsample): Sequential(\n",
       "            (0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
       "            (1): FrozenBatchNorm2d(1024, eps=0.0)\n",
       "          )\n",
       "        )\n",
       "        (1): Bottleneck(\n",
       "          (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn1): FrozenBatchNorm2d(256, eps=0.0)\n",
       "          (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn2): FrozenBatchNorm2d(256, eps=0.0)\n",
       "          (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn3): FrozenBatchNorm2d(1024, eps=0.0)\n",
       "          (relu): ReLU(inplace=True)\n",
       "        )\n",
       "        (2): Bottleneck(\n",
       "          (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn1): FrozenBatchNorm2d(256, eps=0.0)\n",
       "          (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn2): FrozenBatchNorm2d(256, eps=0.0)\n",
       "          (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn3): FrozenBatchNorm2d(1024, eps=0.0)\n",
       "          (relu): ReLU(inplace=True)\n",
       "        )\n",
       "        (3): Bottleneck(\n",
       "          (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn1): FrozenBatchNorm2d(256, eps=0.0)\n",
       "          (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn2): FrozenBatchNorm2d(256, eps=0.0)\n",
       "          (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn3): FrozenBatchNorm2d(1024, eps=0.0)\n",
       "          (relu): ReLU(inplace=True)\n",
       "        )\n",
       "        (4): Bottleneck(\n",
       "          (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn1): FrozenBatchNorm2d(256, eps=0.0)\n",
       "          (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn2): FrozenBatchNorm2d(256, eps=0.0)\n",
       "          (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn3): FrozenBatchNorm2d(1024, eps=0.0)\n",
       "          (relu): ReLU(inplace=True)\n",
       "        )\n",
       "        (5): Bottleneck(\n",
       "          (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn1): FrozenBatchNorm2d(256, eps=0.0)\n",
       "          (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn2): FrozenBatchNorm2d(256, eps=0.0)\n",
       "          (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn3): FrozenBatchNorm2d(1024, eps=0.0)\n",
       "          (relu): ReLU(inplace=True)\n",
       "        )\n",
       "      )\n",
       "      (layer4): Sequential(\n",
       "        (0): Bottleneck(\n",
       "          (conv1): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn1): FrozenBatchNorm2d(512, eps=0.0)\n",
       "          (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "          (bn2): FrozenBatchNorm2d(512, eps=0.0)\n",
       "          (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn3): FrozenBatchNorm2d(2048, eps=0.0)\n",
       "          (relu): ReLU(inplace=True)\n",
       "          (downsample): Sequential(\n",
       "            (0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
       "            (1): FrozenBatchNorm2d(2048, eps=0.0)\n",
       "          )\n",
       "        )\n",
       "        (1): Bottleneck(\n",
       "          (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn1): FrozenBatchNorm2d(512, eps=0.0)\n",
       "          (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn2): FrozenBatchNorm2d(512, eps=0.0)\n",
       "          (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn3): FrozenBatchNorm2d(2048, eps=0.0)\n",
       "          (relu): ReLU(inplace=True)\n",
       "        )\n",
       "        (2): Bottleneck(\n",
       "          (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn1): FrozenBatchNorm2d(512, eps=0.0)\n",
       "          (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "          (bn2): FrozenBatchNorm2d(512, eps=0.0)\n",
       "          (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (bn3): FrozenBatchNorm2d(2048, eps=0.0)\n",
       "          (relu): ReLU(inplace=True)\n",
       "        )\n",
       "      )\n",
       "    )\n",
       "    (fpn): FeaturePyramidNetwork(\n",
       "      (inner_blocks): ModuleList(\n",
       "        (0): Conv2d(256, 256, kernel_size=(1, 1), stride=(1, 1))\n",
       "        (1): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1))\n",
       "        (2): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
       "        (3): Conv2d(2048, 256, kernel_size=(1, 1), stride=(1, 1))\n",
       "      )\n",
       "      (layer_blocks): ModuleList(\n",
       "        (0): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
       "        (1): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
       "        (2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
       "        (3): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
       "      )\n",
       "      (extra_blocks): LastLevelMaxPool()\n",
       "    )\n",
       "  )\n",
       "  (rpn): RegionProposalNetwork(\n",
       "    (anchor_generator): AnchorGenerator()\n",
       "    (head): RPNHead(\n",
       "      (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
       "      (cls_logits): Conv2d(256, 3, kernel_size=(1, 1), stride=(1, 1))\n",
       "      (bbox_pred): Conv2d(256, 12, kernel_size=(1, 1), stride=(1, 1))\n",
       "    )\n",
       "  )\n",
       "  (roi_heads): RoIHeads(\n",
       "    (box_roi_pool): MultiScaleRoIAlignWrapper(\n",
       "      (roi_align): MultiScaleRoIAlign(featmap_names=['0', '1', '2', '3'], output_size=(7, 7), sampling_ratio=2)\n",
       "    )\n",
       "    (box_head): TwoMLPHeadWrapper(\n",
       "      (box_head): TwoMLPHead(\n",
       "        (fc6): Linear(in_features=12544, out_features=1024, bias=True)\n",
       "        (fc7): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "      )\n",
       "      (gator): GATor(\n",
       "        (convs): ModuleList(\n",
       "          (0): GATv2Conv(1024, 1024, heads=1)\n",
       "          (1): GATv2Conv(1024, 1024, heads=1)\n",
       "        )\n",
       "        (post_mp): Sequential(\n",
       "          (0): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "          (1): Dropout(p=0.5, inplace=False)\n",
       "          (2): Linear(in_features=1024, out_features=1024, bias=True)\n",
       "        )\n",
       "      )\n",
       "    )\n",
       "    (box_predictor): FastRCNNPredictor(\n",
       "      (cls_score): Linear(in_features=1024, out_features=91, bias=True)\n",
       "      (bbox_pred): Linear(in_features=1024, out_features=364, bias=True)\n",
       "    )\n",
       "  )\n",
       ")"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cached = {}\n",
    "model = build_model()\n",
    "model.to(device)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "-JLQATBxx1ZA"
   },
   "source": [
    "### Defining PyTorch dataloaders"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "id": "cspE-mQk1MS5"
   },
   "outputs": [],
   "source": [
    "def collate_fn(batch):\n",
    "  return tuple(zip(*batch))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "id": "z5ONelLMx9s2"
   },
   "outputs": [],
   "source": [
    "data_loader_train = DataLoader(torch_dataset_train, batch_size=48, shuffle=True, collate_fn=collate_fn)\n",
    "data_loader_test = DataLoader(torch_dataset_test, batch_size=1, shuffle=False, collate_fn=collate_fn)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "oSHErrODx4_c"
   },
   "source": [
    "### Training using standard COCO training pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "id": "bpFYMb_DIRXL"
   },
   "outputs": [],
   "source": [
    "def save_model(model, path):\n",
    "  torch.save(model.state_dict(), path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "id": "yY3YPOCYyOgB"
   },
   "outputs": [],
   "source": [
    "def train(model, train_loader, test_loader, n_epochs, save_dir='models'):\n",
    "  if not osp.exists(save_dir):\n",
    "    os.mkdir(save_dir)\n",
    "  device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')\n",
    "  model.train()\n",
    "  for param in model.parameters():\n",
    "    param.requires_grad = False\n",
    "  for param in model.roi_heads.box_head.gator.parameters():\n",
    "    param.requires_grad = True\n",
    "  for param in model.roi_heads.box_predictor.parameters():\n",
    "    param.requires_grad = True\n",
    "  model.to(device)\n",
    "  params = [p for p in model.parameters() if p.requires_grad]\n",
    "#   optimizer = torch.optim.RAdam(params, lr=0.02)\n",
    "  optimizer = torch.optim.SGD(params, lr=0.02, momentum=0.9, weight_decay=0.0005)\n",
    "\n",
    "  lr_scheduler = torch.optim.lr_scheduler.StepLR(optimizer,\n",
    "                                                step_size=3,\n",
    "                                                gamma=0.1)\n",
    "\n",
    "  for epoch in range(n_epochs):\n",
    "    # train for one epoch, printing every 10 iterations\n",
    "    train_one_epoch(model, optimizer, train_loader, device, epoch, print_freq=10)\n",
    "    # update the learning rate\n",
    "    lr_scheduler.step()\n",
    "    \n",
    "    # save the state dict\n",
    "    torch.save(model.state_dict(), osp.join(save_dir, f'faster-rcnn-gator-epochs-{epoch}.pt'))\n",
    "    \n",
    "    # evaluate on the test dataset\n",
    "    evaluate(model, test_loader, device=device)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# MAIN TRAINING CELL"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 542
    },
    "id": "lkjjcTSwyiBF",
    "outputId": "c27d9bd5-8c7f-4053-f2c0-f43f803bfbbc"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.7/site-packages/torch/functional.py:445: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at  ../aten/src/ATen/native/TensorShape.cpp:2157.)\n",
      "  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: [0]  [   0/2444]  eta: 1:15:52  lr: 0.000040  loss: 4.7253 (4.7253)  loss_classifier: 3.9863 (3.9863)  loss_box_reg: 0.5126 (0.5126)  loss_objectness: 0.1487 (0.1487)  loss_rpn_box_reg: 0.0778 (0.0778)  time: 1.8625  data: 0.6900  max mem: 19205\n",
      "Epoch: [0]  [  10/2444]  eta: 1:34:28  lr: 0.000240  loss: 4.4290 (4.2228)  loss_classifier: 3.7479 (3.5539)  loss_box_reg: 0.4989 (0.5125)  loss_objectness: 0.0719 (0.0891)  loss_rpn_box_reg: 0.0681 (0.0674)  time: 2.3291  data: 0.6698  max mem: 19259\n",
      "Epoch: [0]  [  20/2444]  eta: 1:32:20  lr: 0.000440  loss: 2.9115 (3.1486)  loss_classifier: 2.2240 (2.4915)  loss_box_reg: 0.5097 (0.5159)  loss_objectness: 0.0719 (0.0791)  loss_rpn_box_reg: 0.0558 (0.0621)  time: 2.3070  data: 0.6651  max mem: 19259\n",
      "Epoch: [0]  [  30/2444]  eta: 1:24:18  lr: 0.000639  loss: 1.8169 (2.7437)  loss_classifier: 1.1975 (2.1026)  loss_box_reg: 0.5097 (0.5130)  loss_objectness: 0.0550 (0.0706)  loss_rpn_box_reg: 0.0515 (0.0574)  time: 1.9667  data: 0.6637  max mem: 19259\n",
      "Epoch: [0]  [  40/2444]  eta: 1:20:16  lr: 0.000839  loss: 1.8814 (2.5348)  loss_classifier: 1.1993 (1.8826)  loss_box_reg: 0.5286 (0.5194)  loss_objectness: 0.0550 (0.0733)  loss_rpn_box_reg: 0.0551 (0.0595)  time: 1.7072  data: 0.6763  max mem: 19259\n",
      "Epoch: [0]  [  50/2444]  eta: 1:17:36  lr: 0.001039  loss: 1.6534 (2.3497)  loss_classifier: 1.0166 (1.7013)  loss_box_reg: 0.5248 (0.5178)  loss_objectness: 0.0687 (0.0717)  loss_rpn_box_reg: 0.0578 (0.0590)  time: 1.7119  data: 0.6813  max mem: 19259\n",
      "Epoch: [0]  [  60/2444]  eta: 1:15:54  lr: 0.001239  loss: 1.6358 (2.2386)  loss_classifier: 0.9769 (1.5832)  loss_box_reg: 0.5285 (0.5229)  loss_objectness: 0.0687 (0.0724)  loss_rpn_box_reg: 0.0601 (0.0602)  time: 1.7194  data: 0.6879  max mem: 19259\n",
      "Epoch: [0]  [  70/2444]  eta: 1:14:27  lr: 0.001439  loss: 1.6125 (2.1398)  loss_classifier: 0.9496 (1.4866)  loss_box_reg: 0.5285 (0.5212)  loss_objectness: 0.0734 (0.0722)  loss_rpn_box_reg: 0.0601 (0.0599)  time: 1.7209  data: 0.6888  max mem: 19259\n",
      "Epoch: [0]  [  80/2444]  eta: 1:13:20  lr: 0.001638  loss: 1.5329 (2.0697)  loss_classifier: 0.8658 (1.4125)  loss_box_reg: 0.5176 (0.5221)  loss_objectness: 0.0719 (0.0743)  loss_rpn_box_reg: 0.0574 (0.0607)  time: 1.7121  data: 0.6798  max mem: 19259\n",
      "Epoch: [0]  [  90/2444]  eta: 1:12:23  lr: 0.001838  loss: 1.5329 (2.0107)  loss_classifier: 0.8690 (1.3533)  loss_box_reg: 0.5224 (0.5227)  loss_objectness: 0.0705 (0.0740)  loss_rpn_box_reg: 0.0584 (0.0608)  time: 1.7145  data: 0.6805  max mem: 19259\n",
      "Epoch: [0]  [ 100/2444]  eta: 1:11:29  lr: 0.002038  loss: 1.4522 (1.9564)  loss_classifier: 0.8402 (1.3002)  loss_box_reg: 0.4942 (0.5211)  loss_objectness: 0.0602 (0.0743)  loss_rpn_box_reg: 0.0566 (0.0608)  time: 1.7023  data: 0.6691  max mem: 19259\n",
      "Epoch: [0]  [ 110/2444]  eta: 1:10:42  lr: 0.002238  loss: 1.3888 (1.9098)  loss_classifier: 0.7857 (1.2546)  loss_box_reg: 0.4942 (0.5196)  loss_objectness: 0.0647 (0.0748)  loss_rpn_box_reg: 0.0553 (0.0607)  time: 1.6943  data: 0.6619  max mem: 19259\n",
      "Epoch: [0]  [ 120/2444]  eta: 1:10:02  lr: 0.002438  loss: 1.3852 (1.8696)  loss_classifier: 0.7776 (1.2155)  loss_box_reg: 0.4947 (0.5186)  loss_objectness: 0.0647 (0.0745)  loss_rpn_box_reg: 0.0550 (0.0610)  time: 1.6979  data: 0.6659  max mem: 19259\n",
      "Epoch: [0]  [ 130/2444]  eta: 1:09:26  lr: 0.002637  loss: 1.4668 (1.8400)  loss_classifier: 0.7870 (1.1846)  loss_box_reg: 0.5190 (0.5201)  loss_objectness: 0.0658 (0.0740)  loss_rpn_box_reg: 0.0596 (0.0614)  time: 1.7035  data: 0.6716  max mem: 19259\n",
      "Epoch: [0]  [ 140/2444]  eta: 1:08:51  lr: 0.002837  loss: 1.4287 (1.8056)  loss_classifier: 0.7660 (1.1523)  loss_box_reg: 0.5090 (0.5183)  loss_objectness: 0.0692 (0.0740)  loss_rpn_box_reg: 0.0574 (0.0609)  time: 1.7037  data: 0.6705  max mem: 19259\n",
      "Epoch: [0]  [ 150/2444]  eta: 1:08:21  lr: 0.003037  loss: 1.3907 (1.7789)  loss_classifier: 0.7398 (1.1250)  loss_box_reg: 0.4982 (0.5180)  loss_objectness: 0.0731 (0.0747)  loss_rpn_box_reg: 0.0573 (0.0613)  time: 1.7052  data: 0.6731  max mem: 19259\n",
      "Epoch: [0]  [ 160/2444]  eta: 1:07:51  lr: 0.003237  loss: 1.4068 (1.7556)  loss_classifier: 0.7529 (1.1015)  loss_box_reg: 0.5197 (0.5183)  loss_objectness: 0.0679 (0.0745)  loss_rpn_box_reg: 0.0619 (0.0612)  time: 1.7088  data: 0.6783  max mem: 19259\n",
      "Epoch: [0]  [ 170/2444]  eta: 1:07:22  lr: 0.003437  loss: 1.4311 (1.7360)  loss_classifier: 0.7529 (1.0805)  loss_box_reg: 0.5268 (0.5187)  loss_objectness: 0.0643 (0.0755)  loss_rpn_box_reg: 0.0619 (0.0614)  time: 1.7021  data: 0.6710  max mem: 19259\n",
      "Epoch: [0]  [ 180/2444]  eta: 1:06:56  lr: 0.003636  loss: 1.4351 (1.7188)  loss_classifier: 0.7359 (1.0614)  loss_box_reg: 0.5327 (0.5194)  loss_objectness: 0.0835 (0.0763)  loss_rpn_box_reg: 0.0634 (0.0618)  time: 1.7048  data: 0.6730  max mem: 19259\n",
      "Epoch: [0]  [ 190/2444]  eta: 1:06:30  lr: 0.003836  loss: 1.4226 (1.7022)  loss_classifier: 0.7232 (1.0435)  loss_box_reg: 0.5395 (0.5202)  loss_objectness: 0.0835 (0.0766)  loss_rpn_box_reg: 0.0668 (0.0619)  time: 1.7088  data: 0.6772  max mem: 19259\n",
      "Epoch: [0]  [ 200/2444]  eta: 1:06:06  lr: 0.004036  loss: 1.3786 (1.6858)  loss_classifier: 0.7036 (1.0273)  loss_box_reg: 0.5349 (0.5203)  loss_objectness: 0.0740 (0.0764)  loss_rpn_box_reg: 0.0574 (0.0618)  time: 1.7064  data: 0.6752  max mem: 19259\n",
      "Epoch: [0]  [ 210/2444]  eta: 1:05:41  lr: 0.004236  loss: 1.3360 (1.6692)  loss_classifier: 0.6894 (1.0108)  loss_box_reg: 0.5109 (0.5203)  loss_objectness: 0.0723 (0.0765)  loss_rpn_box_reg: 0.0574 (0.0617)  time: 1.7044  data: 0.6726  max mem: 19259\n",
      "Epoch: [0]  [ 220/2444]  eta: 1:05:18  lr: 0.004436  loss: 1.3360 (1.6551)  loss_classifier: 0.6826 (0.9962)  loss_box_reg: 0.5162 (0.5203)  loss_objectness: 0.0760 (0.0766)  loss_rpn_box_reg: 0.0593 (0.0619)  time: 1.7090  data: 0.6754  max mem: 19259\n",
      "Epoch: [0]  [ 230/2444]  eta: 1:04:55  lr: 0.004635  loss: 1.3350 (1.6407)  loss_classifier: 0.6729 (0.9826)  loss_box_reg: 0.5237 (0.5202)  loss_objectness: 0.0683 (0.0761)  loss_rpn_box_reg: 0.0616 (0.0618)  time: 1.7099  data: 0.6767  max mem: 19259\n",
      "Epoch: [0]  [ 240/2444]  eta: 1:04:33  lr: 0.004835  loss: 1.2830 (1.6255)  loss_classifier: 0.6591 (0.9682)  loss_box_reg: 0.5197 (0.5198)  loss_objectness: 0.0533 (0.0756)  loss_rpn_box_reg: 0.0599 (0.0619)  time: 1.7078  data: 0.6749  max mem: 19259\n",
      "Epoch: [0]  [ 250/2444]  eta: 1:04:11  lr: 0.005035  loss: 1.3027 (1.6145)  loss_classifier: 0.6572 (0.9563)  loss_box_reg: 0.5240 (0.5206)  loss_objectness: 0.0616 (0.0756)  loss_rpn_box_reg: 0.0574 (0.0619)  time: 1.7098  data: 0.6767  max mem: 19259\n",
      "Epoch: [0]  [ 260/2444]  eta: 1:03:51  lr: 0.005235  loss: 1.3114 (1.6042)  loss_classifier: 0.6572 (0.9454)  loss_box_reg: 0.5394 (0.5212)  loss_objectness: 0.0682 (0.0755)  loss_rpn_box_reg: 0.0586 (0.0621)  time: 1.7142  data: 0.6818  max mem: 19259\n",
      "Epoch: [0]  [ 270/2444]  eta: 1:03:27  lr: 0.005435  loss: 1.2636 (1.5900)  loss_classifier: 0.6221 (0.9326)  loss_box_reg: 0.5099 (0.5203)  loss_objectness: 0.0666 (0.0751)  loss_rpn_box_reg: 0.0634 (0.0620)  time: 1.7010  data: 0.6687  max mem: 19259\n",
      "Epoch: [0]  [ 280/2444]  eta: 1:03:05  lr: 0.005634  loss: 1.2123 (1.5778)  loss_classifier: 0.5953 (0.9213)  loss_box_reg: 0.4863 (0.5200)  loss_objectness: 0.0615 (0.0746)  loss_rpn_box_reg: 0.0601 (0.0619)  time: 1.6831  data: 0.6532  max mem: 19259\n",
      "Epoch: [0]  [ 290/2444]  eta: 1:02:43  lr: 0.005834  loss: 1.2351 (1.5681)  loss_classifier: 0.6040 (0.9111)  loss_box_reg: 0.5054 (0.5205)  loss_objectness: 0.0632 (0.0745)  loss_rpn_box_reg: 0.0625 (0.0620)  time: 1.6902  data: 0.6631  max mem: 19259\n",
      "Epoch: [0]  [ 300/2444]  eta: 1:02:23  lr: 0.006034  loss: 1.3052 (1.5594)  loss_classifier: 0.6232 (0.9015)  loss_box_reg: 0.5307 (0.5212)  loss_objectness: 0.0648 (0.0745)  loss_rpn_box_reg: 0.0637 (0.0621)  time: 1.6995  data: 0.6744  max mem: 19259\n",
      "Epoch: [0]  [ 310/2444]  eta: 1:02:02  lr: 0.006234  loss: 1.3107 (1.5508)  loss_classifier: 0.6232 (0.8925)  loss_box_reg: 0.5422 (0.5217)  loss_objectness: 0.0644 (0.0744)  loss_rpn_box_reg: 0.0627 (0.0622)  time: 1.7006  data: 0.6773  max mem: 19259\n",
      "Epoch: [0]  [ 320/2444]  eta: 1:01:41  lr: 0.006434  loss: 1.2384 (1.5405)  loss_classifier: 0.6026 (0.8830)  loss_box_reg: 0.5266 (0.5214)  loss_objectness: 0.0574 (0.0740)  loss_rpn_box_reg: 0.0589 (0.0620)  time: 1.6905  data: 0.6673  max mem: 19259\n",
      "Epoch: [0]  [ 330/2444]  eta: 1:01:20  lr: 0.006633  loss: 1.2384 (1.5319)  loss_classifier: 0.5875 (0.8741)  loss_box_reg: 0.5265 (0.5215)  loss_objectness: 0.0678 (0.0741)  loss_rpn_box_reg: 0.0589 (0.0622)  time: 1.6920  data: 0.6689  max mem: 19259\n",
      "Epoch: [0]  [ 340/2444]  eta: 1:01:00  lr: 0.006833  loss: 1.2446 (1.5231)  loss_classifier: 0.5682 (0.8651)  loss_box_reg: 0.5143 (0.5212)  loss_objectness: 0.0754 (0.0746)  loss_rpn_box_reg: 0.0619 (0.0622)  time: 1.6940  data: 0.6732  max mem: 19259\n",
      "Epoch: [0]  [ 350/2444]  eta: 1:00:39  lr: 0.007033  loss: 1.2105 (1.5140)  loss_classifier: 0.5625 (0.8563)  loss_box_reg: 0.5049 (0.5208)  loss_objectness: 0.0760 (0.0747)  loss_rpn_box_reg: 0.0615 (0.0623)  time: 1.6823  data: 0.6630  max mem: 19259\n",
      "Epoch: [0]  [ 360/2444]  eta: 1:00:18  lr: 0.007233  loss: 1.1937 (1.5054)  loss_classifier: 0.5451 (0.8477)  loss_box_reg: 0.5049 (0.5206)  loss_objectness: 0.0660 (0.0746)  loss_rpn_box_reg: 0.0642 (0.0624)  time: 1.6796  data: 0.6594  max mem: 19259\n",
      "Epoch: [0]  [ 370/2444]  eta: 0:59:58  lr: 0.007433  loss: 1.1937 (1.4978)  loss_classifier: 0.5447 (0.8396)  loss_box_reg: 0.5095 (0.5207)  loss_objectness: 0.0739 (0.0750)  loss_rpn_box_reg: 0.0647 (0.0624)  time: 1.6864  data: 0.6652  max mem: 19259\n",
      "Epoch: [0]  [ 380/2444]  eta: 0:59:38  lr: 0.007632  loss: 1.1793 (1.4895)  loss_classifier: 0.5266 (0.8316)  loss_box_reg: 0.5143 (0.5205)  loss_objectness: 0.0753 (0.0750)  loss_rpn_box_reg: 0.0597 (0.0623)  time: 1.6887  data: 0.6682  max mem: 19259\n",
      "Epoch: [0]  [ 390/2444]  eta: 0:59:18  lr: 0.007832  loss: 1.1667 (1.4810)  loss_classifier: 0.5249 (0.8233)  loss_box_reg: 0.5143 (0.5203)  loss_objectness: 0.0711 (0.0750)  loss_rpn_box_reg: 0.0593 (0.0623)  time: 1.6839  data: 0.6644  max mem: 19259\n",
      "Epoch: [0]  [ 400/2444]  eta: 0:58:58  lr: 0.008032  loss: 1.1486 (1.4727)  loss_classifier: 0.4974 (0.8151)  loss_box_reg: 0.5138 (0.5203)  loss_objectness: 0.0698 (0.0750)  loss_rpn_box_reg: 0.0584 (0.0622)  time: 1.6803  data: 0.6615  max mem: 19259\n",
      "Epoch: [0]  [ 410/2444]  eta: 0:58:38  lr: 0.008232  loss: 1.1523 (1.4649)  loss_classifier: 0.4892 (0.8073)  loss_box_reg: 0.5207 (0.5203)  loss_objectness: 0.0714 (0.0751)  loss_rpn_box_reg: 0.0578 (0.0622)  time: 1.6806  data: 0.6593  max mem: 19259\n",
      "Epoch: [0]  [ 420/2444]  eta: 0:58:19  lr: 0.008432  loss: 1.1220 (1.4569)  loss_classifier: 0.4880 (0.7996)  loss_box_reg: 0.5191 (0.5201)  loss_objectness: 0.0702 (0.0750)  loss_rpn_box_reg: 0.0576 (0.0622)  time: 1.6823  data: 0.6608  max mem: 19259\n",
      "Epoch: [0]  [ 430/2444]  eta: 0:58:00  lr: 0.008631  loss: 1.0992 (1.4483)  loss_classifier: 0.4667 (0.7917)  loss_box_reg: 0.5191 (0.5200)  loss_objectness: 0.0618 (0.0746)  loss_rpn_box_reg: 0.0545 (0.0620)  time: 1.6919  data: 0.6720  max mem: 19259\n",
      "Epoch: [0]  [ 440/2444]  eta: 0:57:41  lr: 0.008831  loss: 1.0863 (1.4406)  loss_classifier: 0.4555 (0.7843)  loss_box_reg: 0.5141 (0.5200)  loss_objectness: 0.0602 (0.0744)  loss_rpn_box_reg: 0.0531 (0.0620)  time: 1.6920  data: 0.6707  max mem: 19259\n",
      "Epoch: [0]  [ 450/2444]  eta: 0:57:21  lr: 0.009031  loss: 1.0681 (1.4322)  loss_classifier: 0.4444 (0.7767)  loss_box_reg: 0.4995 (0.5194)  loss_objectness: 0.0646 (0.0742)  loss_rpn_box_reg: 0.0563 (0.0619)  time: 1.6797  data: 0.6592  max mem: 19259\n",
      "Epoch: [0]  [ 460/2444]  eta: 0:57:02  lr: 0.009231  loss: 1.0339 (1.4248)  loss_classifier: 0.4261 (0.7693)  loss_box_reg: 0.4897 (0.5190)  loss_objectness: 0.0683 (0.0745)  loss_rpn_box_reg: 0.0652 (0.0620)  time: 1.6807  data: 0.6612  max mem: 19259\n",
      "Epoch: [0]  [ 470/2444]  eta: 0:56:43  lr: 0.009431  loss: 1.0701 (1.4164)  loss_classifier: 0.4151 (0.7617)  loss_box_reg: 0.4897 (0.5186)  loss_objectness: 0.0605 (0.0743)  loss_rpn_box_reg: 0.0586 (0.0618)  time: 1.6838  data: 0.6646  max mem: 19259\n",
      "Epoch: [0]  [ 480/2444]  eta: 0:56:24  lr: 0.009630  loss: 1.0480 (1.4083)  loss_classifier: 0.4124 (0.7544)  loss_box_reg: 0.4916 (0.5181)  loss_objectness: 0.0584 (0.0740)  loss_rpn_box_reg: 0.0539 (0.0618)  time: 1.6819  data: 0.6632  max mem: 19259\n",
      "Epoch: [0]  [ 490/2444]  eta: 0:56:06  lr: 0.009830  loss: 1.0521 (1.4015)  loss_classifier: 0.4220 (0.7477)  loss_box_reg: 0.5009 (0.5181)  loss_objectness: 0.0664 (0.0740)  loss_rpn_box_reg: 0.0570 (0.0617)  time: 1.6870  data: 0.6660  max mem: 19259\n",
      "Epoch: [0]  [ 500/2444]  eta: 0:55:47  lr: 0.010030  loss: 1.0625 (1.3946)  loss_classifier: 0.4090 (0.7409)  loss_box_reg: 0.5041 (0.5179)  loss_objectness: 0.0664 (0.0740)  loss_rpn_box_reg: 0.0570 (0.0618)  time: 1.6866  data: 0.6655  max mem: 19259\n",
      "Epoch: [0]  [ 510/2444]  eta: 0:55:28  lr: 0.010230  loss: 1.0149 (1.3873)  loss_classifier: 0.3949 (0.7340)  loss_box_reg: 0.5041 (0.5177)  loss_objectness: 0.0648 (0.0739)  loss_rpn_box_reg: 0.0570 (0.0616)  time: 1.6796  data: 0.6607  max mem: 19259\n",
      "Epoch: [0]  [ 520/2444]  eta: 0:55:09  lr: 0.010430  loss: 1.0149 (1.3799)  loss_classifier: 0.3838 (0.7272)  loss_box_reg: 0.5013 (0.5172)  loss_objectness: 0.0681 (0.0738)  loss_rpn_box_reg: 0.0556 (0.0616)  time: 1.6763  data: 0.6567  max mem: 19259\n",
      "Epoch: [0]  [ 530/2444]  eta: 0:54:50  lr: 0.010629  loss: 1.0332 (1.3738)  loss_classifier: 0.3868 (0.7209)  loss_box_reg: 0.5054 (0.5175)  loss_objectness: 0.0723 (0.0738)  loss_rpn_box_reg: 0.0556 (0.0616)  time: 1.6721  data: 0.6526  max mem: 19259\n",
      "Epoch: [0]  [ 540/2444]  eta: 0:54:31  lr: 0.010829  loss: 1.0436 (1.3674)  loss_classifier: 0.3758 (0.7146)  loss_box_reg: 0.5196 (0.5173)  loss_objectness: 0.0666 (0.0739)  loss_rpn_box_reg: 0.0545 (0.0615)  time: 1.6713  data: 0.6513  max mem: 19259\n",
      "Epoch: [0]  [ 550/2444]  eta: 0:54:13  lr: 0.011029  loss: 0.9861 (1.3605)  loss_classifier: 0.3555 (0.7082)  loss_box_reg: 0.4967 (0.5170)  loss_objectness: 0.0666 (0.0738)  loss_rpn_box_reg: 0.0556 (0.0615)  time: 1.6801  data: 0.6591  max mem: 19259\n",
      "Epoch: [0]  [ 560/2444]  eta: 0:53:55  lr: 0.011229  loss: 1.0233 (1.3553)  loss_classifier: 0.3643 (0.7025)  loss_box_reg: 0.5014 (0.5171)  loss_objectness: 0.0747 (0.0743)  loss_rpn_box_reg: 0.0561 (0.0615)  time: 1.6867  data: 0.6660  max mem: 19259\n",
      "Epoch: [0]  [ 570/2444]  eta: 0:53:36  lr: 0.011429  loss: 1.0410 (1.3503)  loss_classifier: 0.3643 (0.6968)  loss_box_reg: 0.5215 (0.5174)  loss_objectness: 0.0787 (0.0744)  loss_rpn_box_reg: 0.0677 (0.0616)  time: 1.6840  data: 0.6639  max mem: 19259\n",
      "Epoch: [0]  [ 580/2444]  eta: 0:53:18  lr: 0.011628  loss: 1.0135 (1.3450)  loss_classifier: 0.3671 (0.6912)  loss_box_reg: 0.5305 (0.5175)  loss_objectness: 0.0752 (0.0747)  loss_rpn_box_reg: 0.0600 (0.0616)  time: 1.6809  data: 0.6619  max mem: 19259\n",
      "Epoch: [0]  [ 590/2444]  eta: 0:53:00  lr: 0.011828  loss: 1.0221 (1.3395)  loss_classifier: 0.3659 (0.6856)  loss_box_reg: 0.5177 (0.5175)  loss_objectness: 0.0782 (0.0747)  loss_rpn_box_reg: 0.0612 (0.0617)  time: 1.6867  data: 0.6677  max mem: 19259\n",
      "Epoch: [0]  [ 600/2444]  eta: 0:52:42  lr: 0.012028  loss: 1.0234 (1.3342)  loss_classifier: 0.3483 (0.6802)  loss_box_reg: 0.5110 (0.5175)  loss_objectness: 0.0806 (0.0748)  loss_rpn_box_reg: 0.0662 (0.0617)  time: 1.6924  data: 0.6715  max mem: 19259\n",
      "Epoch: [0]  [ 610/2444]  eta: 0:52:25  lr: 0.012228  loss: 0.9763 (1.3282)  loss_classifier: 0.3493 (0.6747)  loss_box_reg: 0.5127 (0.5173)  loss_objectness: 0.0668 (0.0745)  loss_rpn_box_reg: 0.0658 (0.0617)  time: 1.6934  data: 0.6717  max mem: 19259\n",
      "Epoch: [0]  [ 620/2444]  eta: 0:52:07  lr: 0.012428  loss: 0.9655 (1.3233)  loss_classifier: 0.3459 (0.6695)  loss_box_reg: 0.5020 (0.5173)  loss_objectness: 0.0578 (0.0747)  loss_rpn_box_reg: 0.0658 (0.0618)  time: 1.6904  data: 0.6688  max mem: 19259\n",
      "Epoch: [0]  [ 630/2444]  eta: 0:51:48  lr: 0.012627  loss: 0.9558 (1.3178)  loss_classifier: 0.3284 (0.6643)  loss_box_reg: 0.4986 (0.5170)  loss_objectness: 0.0693 (0.0747)  loss_rpn_box_reg: 0.0663 (0.0618)  time: 1.6839  data: 0.6624  max mem: 19259\n",
      "Epoch: [0]  [ 640/2444]  eta: 0:51:30  lr: 0.012827  loss: 0.9605 (1.3124)  loss_classifier: 0.3263 (0.6591)  loss_box_reg: 0.4966 (0.5169)  loss_objectness: 0.0599 (0.0745)  loss_rpn_box_reg: 0.0575 (0.0618)  time: 1.6826  data: 0.6629  max mem: 19259\n",
      "Epoch: [0]  [ 650/2444]  eta: 0:51:12  lr: 0.013027  loss: 0.9219 (1.3068)  loss_classifier: 0.3226 (0.6541)  loss_box_reg: 0.4961 (0.5166)  loss_objectness: 0.0576 (0.0745)  loss_rpn_box_reg: 0.0557 (0.0616)  time: 1.6811  data: 0.6616  max mem: 19259\n",
      "Epoch: [0]  [ 660/2444]  eta: 0:50:54  lr: 0.013227  loss: 0.9004 (1.3011)  loss_classifier: 0.3102 (0.6491)  loss_box_reg: 0.4860 (0.5162)  loss_objectness: 0.0571 (0.0743)  loss_rpn_box_reg: 0.0546 (0.0615)  time: 1.6814  data: 0.6605  max mem: 19259\n",
      "Epoch: [0]  [ 670/2444]  eta: 0:50:37  lr: 0.013427  loss: 0.9329 (1.2964)  loss_classifier: 0.3187 (0.6444)  loss_box_reg: 0.4901 (0.5161)  loss_objectness: 0.0607 (0.0744)  loss_rpn_box_reg: 0.0580 (0.0616)  time: 1.6912  data: 0.6693  max mem: 19259\n",
      "Epoch: [0]  [ 680/2444]  eta: 0:50:19  lr: 0.013626  loss: 0.9350 (1.2915)  loss_classifier: 0.3187 (0.6398)  loss_box_reg: 0.4934 (0.5159)  loss_objectness: 0.0753 (0.0743)  loss_rpn_box_reg: 0.0626 (0.0616)  time: 1.6864  data: 0.6655  max mem: 19259\n",
      "Epoch: [0]  [ 690/2444]  eta: 0:50:01  lr: 0.013826  loss: 0.9350 (1.2867)  loss_classifier: 0.3180 (0.6353)  loss_box_reg: 0.4959 (0.5157)  loss_objectness: 0.0570 (0.0741)  loss_rpn_box_reg: 0.0575 (0.0615)  time: 1.6858  data: 0.6671  max mem: 19259\n",
      "Epoch: [0]  [ 700/2444]  eta: 0:49:44  lr: 0.014026  loss: 0.9834 (1.2825)  loss_classifier: 0.3371 (0.6312)  loss_box_reg: 0.5003 (0.5158)  loss_objectness: 0.0570 (0.0740)  loss_rpn_box_reg: 0.0584 (0.0616)  time: 1.7001  data: 0.6806  max mem: 19259\n",
      "Epoch: [0]  [ 710/2444]  eta: 0:49:26  lr: 0.014226  loss: 0.9720 (1.2784)  loss_classifier: 0.3353 (0.6269)  loss_box_reg: 0.5105 (0.5157)  loss_objectness: 0.0659 (0.0742)  loss_rpn_box_reg: 0.0623 (0.0616)  time: 1.6925  data: 0.6719  max mem: 19259\n",
      "Epoch: [0]  [ 720/2444]  eta: 0:49:08  lr: 0.014426  loss: 0.9611 (1.2737)  loss_classifier: 0.3162 (0.6226)  loss_box_reg: 0.4954 (0.5153)  loss_objectness: 0.0716 (0.0741)  loss_rpn_box_reg: 0.0623 (0.0616)  time: 1.6807  data: 0.6609  max mem: 19259\n",
      "Epoch: [0]  [ 730/2444]  eta: 0:48:51  lr: 0.014625  loss: 0.9258 (1.2691)  loss_classifier: 0.3164 (0.6185)  loss_box_reg: 0.4918 (0.5150)  loss_objectness: 0.0719 (0.0740)  loss_rpn_box_reg: 0.0602 (0.0616)  time: 1.6874  data: 0.6663  max mem: 19259\n",
      "Epoch: [0]  [ 740/2444]  eta: 0:48:33  lr: 0.014825  loss: 0.9211 (1.2641)  loss_classifier: 0.3129 (0.6142)  loss_box_reg: 0.4900 (0.5146)  loss_objectness: 0.0568 (0.0738)  loss_rpn_box_reg: 0.0540 (0.0615)  time: 1.6870  data: 0.6654  max mem: 19259\n",
      "Epoch: [0]  [ 750/2444]  eta: 0:48:15  lr: 0.015025  loss: 0.9305 (1.2599)  loss_classifier: 0.3038 (0.6102)  loss_box_reg: 0.4837 (0.5143)  loss_objectness: 0.0645 (0.0739)  loss_rpn_box_reg: 0.0568 (0.0616)  time: 1.6766  data: 0.6567  max mem: 19259\n",
      "Epoch: [0]  [ 760/2444]  eta: 0:47:58  lr: 0.015225  loss: 0.9588 (1.2563)  loss_classifier: 0.3080 (0.6064)  loss_box_reg: 0.4975 (0.5141)  loss_objectness: 0.0751 (0.0741)  loss_rpn_box_reg: 0.0642 (0.0616)  time: 1.6827  data: 0.6630  max mem: 19259\n",
      "Epoch: [0]  [ 770/2444]  eta: 0:47:40  lr: 0.015425  loss: 0.9547 (1.2523)  loss_classifier: 0.3117 (0.6027)  loss_box_reg: 0.4908 (0.5139)  loss_objectness: 0.0745 (0.0741)  loss_rpn_box_reg: 0.0625 (0.0616)  time: 1.6912  data: 0.6703  max mem: 19259\n",
      "Epoch: [0]  [ 780/2444]  eta: 0:47:22  lr: 0.015624  loss: 0.9085 (1.2476)  loss_classifier: 0.3093 (0.5987)  loss_box_reg: 0.4879 (0.5134)  loss_objectness: 0.0637 (0.0739)  loss_rpn_box_reg: 0.0600 (0.0616)  time: 1.6799  data: 0.6589  max mem: 19259\n",
      "Epoch: [0]  [ 790/2444]  eta: 0:47:05  lr: 0.015824  loss: 0.9229 (1.2441)  loss_classifier: 0.3050 (0.5952)  loss_box_reg: 0.4900 (0.5132)  loss_objectness: 0.0688 (0.0740)  loss_rpn_box_reg: 0.0583 (0.0616)  time: 1.6888  data: 0.6685  max mem: 19259\n",
      "Epoch: [0]  [ 800/2444]  eta: 0:46:48  lr: 0.016024  loss: 0.9439 (1.2401)  loss_classifier: 0.3160 (0.5917)  loss_box_reg: 0.4947 (0.5128)  loss_objectness: 0.0758 (0.0740)  loss_rpn_box_reg: 0.0619 (0.0615)  time: 1.6973  data: 0.6781  max mem: 19259\n",
      "Epoch: [0]  [ 810/2444]  eta: 0:46:30  lr: 0.016224  loss: 0.9122 (1.2360)  loss_classifier: 0.3021 (0.5881)  loss_box_reg: 0.4692 (0.5123)  loss_objectness: 0.0676 (0.0740)  loss_rpn_box_reg: 0.0584 (0.0615)  time: 1.6829  data: 0.6642  max mem: 19259\n",
      "Epoch: [0]  [ 820/2444]  eta: 0:46:12  lr: 0.016424  loss: 0.8942 (1.2325)  loss_classifier: 0.3073 (0.5849)  loss_box_reg: 0.4692 (0.5121)  loss_objectness: 0.0646 (0.0740)  loss_rpn_box_reg: 0.0560 (0.0615)  time: 1.6831  data: 0.6624  max mem: 19259\n",
      "Epoch: [0]  [ 830/2444]  eta: 0:45:55  lr: 0.016623  loss: 0.9199 (1.2289)  loss_classifier: 0.3105 (0.5816)  loss_box_reg: 0.4794 (0.5118)  loss_objectness: 0.0759 (0.0740)  loss_rpn_box_reg: 0.0576 (0.0615)  time: 1.6878  data: 0.6675  max mem: 19259\n",
      "Epoch: [0]  [ 840/2444]  eta: 0:45:37  lr: 0.016823  loss: 0.8954 (1.2250)  loss_classifier: 0.3078 (0.5783)  loss_box_reg: 0.4688 (0.5114)  loss_objectness: 0.0677 (0.0739)  loss_rpn_box_reg: 0.0586 (0.0615)  time: 1.6837  data: 0.6648  max mem: 19259\n",
      "Epoch: [0]  [ 850/2444]  eta: 0:45:20  lr: 0.017023  loss: 0.8688 (1.2211)  loss_classifier: 0.2910 (0.5750)  loss_box_reg: 0.4667 (0.5109)  loss_objectness: 0.0556 (0.0738)  loss_rpn_box_reg: 0.0560 (0.0615)  time: 1.6779  data: 0.6582  max mem: 19259\n",
      "Epoch: [0]  [ 860/2444]  eta: 0:45:02  lr: 0.017223  loss: 0.8877 (1.2176)  loss_classifier: 0.3041 (0.5719)  loss_box_reg: 0.4714 (0.5105)  loss_objectness: 0.0650 (0.0738)  loss_rpn_box_reg: 0.0592 (0.0615)  time: 1.6824  data: 0.6621  max mem: 19259\n",
      "Epoch: [0]  [ 870/2444]  eta: 0:44:45  lr: 0.017423  loss: 0.8808 (1.2138)  loss_classifier: 0.2993 (0.5687)  loss_box_reg: 0.4614 (0.5100)  loss_objectness: 0.0740 (0.0738)  loss_rpn_box_reg: 0.0594 (0.0614)  time: 1.6929  data: 0.6736  max mem: 19259\n",
      "Epoch: [0]  [ 880/2444]  eta: 0:44:28  lr: 0.017622  loss: 0.8808 (1.2104)  loss_classifier: 0.2915 (0.5656)  loss_box_reg: 0.4586 (0.5095)  loss_objectness: 0.0739 (0.0738)  loss_rpn_box_reg: 0.0581 (0.0615)  time: 1.6946  data: 0.6748  max mem: 19259\n",
      "Epoch: [0]  [ 890/2444]  eta: 0:44:10  lr: 0.017822  loss: 0.8943 (1.2068)  loss_classifier: 0.2915 (0.5626)  loss_box_reg: 0.4746 (0.5092)  loss_objectness: 0.0616 (0.0736)  loss_rpn_box_reg: 0.0557 (0.0614)  time: 1.6835  data: 0.6627  max mem: 19259\n",
      "Epoch: [0]  [ 900/2444]  eta: 0:43:53  lr: 0.018022  loss: 0.8529 (1.2032)  loss_classifier: 0.2894 (0.5596)  loss_box_reg: 0.4640 (0.5087)  loss_objectness: 0.0560 (0.0735)  loss_rpn_box_reg: 0.0557 (0.0613)  time: 1.6819  data: 0.6618  max mem: 19259\n",
      "Epoch: [0]  [ 910/2444]  eta: 0:43:36  lr: 0.018222  loss: 0.8894 (1.1999)  loss_classifier: 0.2983 (0.5567)  loss_box_reg: 0.4653 (0.5083)  loss_objectness: 0.0560 (0.0735)  loss_rpn_box_reg: 0.0593 (0.0614)  time: 1.6911  data: 0.6700  max mem: 19259\n",
      "Epoch: [0]  [ 920/2444]  eta: 0:43:18  lr: 0.018422  loss: 0.9237 (1.1971)  loss_classifier: 0.2991 (0.5540)  loss_box_reg: 0.4672 (0.5080)  loss_objectness: 0.0686 (0.0737)  loss_rpn_box_reg: 0.0627 (0.0614)  time: 1.6951  data: 0.6726  max mem: 19259\n",
      "Epoch: [0]  [ 930/2444]  eta: 0:43:01  lr: 0.018621  loss: 0.8989 (1.1936)  loss_classifier: 0.2986 (0.5512)  loss_box_reg: 0.4672 (0.5075)  loss_objectness: 0.0674 (0.0736)  loss_rpn_box_reg: 0.0602 (0.0614)  time: 1.6863  data: 0.6647  max mem: 19259\n",
      "Epoch: [0]  [ 940/2444]  eta: 0:42:43  lr: 0.018821  loss: 0.8816 (1.1903)  loss_classifier: 0.2935 (0.5484)  loss_box_reg: 0.4694 (0.5071)  loss_objectness: 0.0579 (0.0735)  loss_rpn_box_reg: 0.0560 (0.0613)  time: 1.6767  data: 0.6570  max mem: 19259\n",
      "Epoch: [0]  [ 950/2444]  eta: 0:42:26  lr: 0.019021  loss: 0.8963 (1.1870)  loss_classifier: 0.2975 (0.5457)  loss_box_reg: 0.4639 (0.5066)  loss_objectness: 0.0650 (0.0734)  loss_rpn_box_reg: 0.0524 (0.0613)  time: 1.6752  data: 0.6548  max mem: 19259\n",
      "Epoch: [0]  [ 960/2444]  eta: 0:42:08  lr: 0.019221  loss: 0.8902 (1.1836)  loss_classifier: 0.2910 (0.5430)  loss_box_reg: 0.4570 (0.5061)  loss_objectness: 0.0669 (0.0733)  loss_rpn_box_reg: 0.0529 (0.0612)  time: 1.6741  data: 0.6530  max mem: 19259\n",
      "Epoch: [0]  [ 970/2444]  eta: 0:41:51  lr: 0.019421  loss: 0.8902 (1.1806)  loss_classifier: 0.2910 (0.5403)  loss_box_reg: 0.4689 (0.5058)  loss_objectness: 0.0659 (0.0733)  loss_rpn_box_reg: 0.0567 (0.0612)  time: 1.6759  data: 0.6563  max mem: 19259\n",
      "Epoch: [0]  [ 980/2444]  eta: 0:41:33  lr: 0.019620  loss: 0.8965 (1.1778)  loss_classifier: 0.2992 (0.5379)  loss_box_reg: 0.4727 (0.5054)  loss_objectness: 0.0655 (0.0733)  loss_rpn_box_reg: 0.0624 (0.0612)  time: 1.6769  data: 0.6569  max mem: 19259\n",
      "Epoch: [0]  [ 990/2444]  eta: 0:41:16  lr: 0.019820  loss: 0.9263 (1.1754)  loss_classifier: 0.3003 (0.5356)  loss_box_reg: 0.4781 (0.5051)  loss_objectness: 0.0810 (0.0735)  loss_rpn_box_reg: 0.0629 (0.0613)  time: 1.6850  data: 0.6636  max mem: 19259\n",
      "Epoch: [0]  [1000/2444]  eta: 0:40:59  lr: 0.020000  loss: 0.9189 (1.1722)  loss_classifier: 0.2992 (0.5330)  loss_box_reg: 0.4538 (0.5045)  loss_objectness: 0.0790 (0.0734)  loss_rpn_box_reg: 0.0622 (0.0613)  time: 1.6888  data: 0.6672  max mem: 19259\n",
      "Epoch: [0]  [1010/2444]  eta: 0:40:42  lr: 0.020000  loss: 0.8582 (1.1698)  loss_classifier: 0.2742 (0.5307)  loss_box_reg: 0.4537 (0.5042)  loss_objectness: 0.0684 (0.0735)  loss_rpn_box_reg: 0.0616 (0.0613)  time: 1.6808  data: 0.6598  max mem: 19259\n",
      "Epoch: [0]  [1020/2444]  eta: 0:40:24  lr: 0.020000  loss: 0.8896 (1.1672)  loss_classifier: 0.2960 (0.5285)  loss_box_reg: 0.4649 (0.5038)  loss_objectness: 0.0760 (0.0736)  loss_rpn_box_reg: 0.0628 (0.0613)  time: 1.6858  data: 0.6656  max mem: 19259\n",
      "Epoch: [0]  [1030/2444]  eta: 0:40:07  lr: 0.020000  loss: 0.8745 (1.1643)  loss_classifier: 0.2929 (0.5261)  loss_box_reg: 0.4624 (0.5033)  loss_objectness: 0.0722 (0.0736)  loss_rpn_box_reg: 0.0615 (0.0614)  time: 1.6892  data: 0.6686  max mem: 19259\n",
      "Epoch: [0]  [1040/2444]  eta: 0:39:50  lr: 0.020000  loss: 0.8366 (1.1616)  loss_classifier: 0.2700 (0.5238)  loss_box_reg: 0.4435 (0.5029)  loss_objectness: 0.0698 (0.0736)  loss_rpn_box_reg: 0.0590 (0.0613)  time: 1.6820  data: 0.6625  max mem: 19259\n",
      "Epoch: [0]  [1050/2444]  eta: 0:39:33  lr: 0.020000  loss: 0.8468 (1.1590)  loss_classifier: 0.2830 (0.5215)  loss_box_reg: 0.4458 (0.5025)  loss_objectness: 0.0718 (0.0736)  loss_rpn_box_reg: 0.0590 (0.0613)  time: 1.6868  data: 0.6681  max mem: 19259\n",
      "Epoch: [0]  [1060/2444]  eta: 0:39:15  lr: 0.020000  loss: 0.8468 (1.1560)  loss_classifier: 0.2830 (0.5192)  loss_box_reg: 0.4488 (0.5020)  loss_objectness: 0.0618 (0.0736)  loss_rpn_box_reg: 0.0540 (0.0613)  time: 1.6846  data: 0.6656  max mem: 19259\n",
      "Epoch: [0]  [1070/2444]  eta: 0:38:59  lr: 0.020000  loss: 0.8819 (1.1541)  loss_classifier: 0.2868 (0.5172)  loss_box_reg: 0.4759 (0.5019)  loss_objectness: 0.0636 (0.0737)  loss_rpn_box_reg: 0.0576 (0.0613)  time: 1.6980  data: 0.6791  max mem: 19259\n",
      "Epoch: [0]  [1080/2444]  eta: 0:38:41  lr: 0.020000  loss: 0.9037 (1.1517)  loss_classifier: 0.2983 (0.5151)  loss_box_reg: 0.4782 (0.5016)  loss_objectness: 0.0827 (0.0736)  loss_rpn_box_reg: 0.0617 (0.0613)  time: 1.7027  data: 0.6838  max mem: 19259\n",
      "Epoch: [0]  [1090/2444]  eta: 0:38:24  lr: 0.020000  loss: 0.8987 (1.1494)  loss_classifier: 0.2893 (0.5130)  loss_box_reg: 0.4671 (0.5013)  loss_objectness: 0.0789 (0.0737)  loss_rpn_box_reg: 0.0614 (0.0613)  time: 1.6885  data: 0.6697  max mem: 19259\n",
      "Epoch: [0]  [1100/2444]  eta: 0:38:07  lr: 0.020000  loss: 0.8728 (1.1472)  loss_classifier: 0.2880 (0.5110)  loss_box_reg: 0.4633 (0.5011)  loss_objectness: 0.0807 (0.0738)  loss_rpn_box_reg: 0.0622 (0.0613)  time: 1.6889  data: 0.6698  max mem: 19259\n",
      "Epoch: [0]  [1110/2444]  eta: 0:37:50  lr: 0.020000  loss: 0.8706 (1.1449)  loss_classifier: 0.2826 (0.5090)  loss_box_reg: 0.4555 (0.5007)  loss_objectness: 0.0807 (0.0739)  loss_rpn_box_reg: 0.0649 (0.0614)  time: 1.6853  data: 0.6680  max mem: 19259\n",
      "Epoch: [0]  [1120/2444]  eta: 0:37:33  lr: 0.020000  loss: 0.8607 (1.1424)  loss_classifier: 0.2750 (0.5069)  loss_box_reg: 0.4543 (0.5003)  loss_objectness: 0.0701 (0.0738)  loss_rpn_box_reg: 0.0651 (0.0614)  time: 1.6915  data: 0.6739  max mem: 19259\n",
      "Epoch: [0]  [1130/2444]  eta: 0:37:15  lr: 0.020000  loss: 0.8445 (1.1398)  loss_classifier: 0.2741 (0.5048)  loss_box_reg: 0.4591 (0.5000)  loss_objectness: 0.0586 (0.0737)  loss_rpn_box_reg: 0.0552 (0.0613)  time: 1.6908  data: 0.6721  max mem: 19259\n",
      "Epoch: [0]  [1140/2444]  eta: 0:36:58  lr: 0.020000  loss: 0.8445 (1.1374)  loss_classifier: 0.2792 (0.5029)  loss_box_reg: 0.4545 (0.4996)  loss_objectness: 0.0572 (0.0736)  loss_rpn_box_reg: 0.0537 (0.0613)  time: 1.6964  data: 0.6784  max mem: 19259\n",
      "Epoch: [0]  [1150/2444]  eta: 0:36:41  lr: 0.020000  loss: 0.8868 (1.1357)  loss_classifier: 0.2844 (0.5011)  loss_box_reg: 0.4548 (0.4995)  loss_objectness: 0.0680 (0.0738)  loss_rpn_box_reg: 0.0618 (0.0613)  time: 1.6992  data: 0.6810  max mem: 19259\n",
      "Epoch: [0]  [1160/2444]  eta: 0:36:24  lr: 0.020000  loss: 0.8827 (1.1334)  loss_classifier: 0.2795 (0.4991)  loss_box_reg: 0.4671 (0.4992)  loss_objectness: 0.0739 (0.0738)  loss_rpn_box_reg: 0.0625 (0.0613)  time: 1.6872  data: 0.6666  max mem: 19259\n",
      "Epoch: [0]  [1170/2444]  eta: 0:36:07  lr: 0.020000  loss: 0.8625 (1.1307)  loss_classifier: 0.2750 (0.4971)  loss_box_reg: 0.4590 (0.4986)  loss_objectness: 0.0641 (0.0737)  loss_rpn_box_reg: 0.0571 (0.0613)  time: 1.6821  data: 0.6615  max mem: 19259\n",
      "Epoch: [0]  [1180/2444]  eta: 0:35:50  lr: 0.020000  loss: 0.8649 (1.1287)  loss_classifier: 0.2675 (0.4952)  loss_box_reg: 0.4471 (0.4984)  loss_objectness: 0.0641 (0.0738)  loss_rpn_box_reg: 0.0593 (0.0613)  time: 1.6871  data: 0.6677  max mem: 19259\n",
      "Epoch: [0]  [1190/2444]  eta: 0:35:33  lr: 0.020000  loss: 0.8674 (1.1268)  loss_classifier: 0.2756 (0.4934)  loss_box_reg: 0.4621 (0.4982)  loss_objectness: 0.0704 (0.0738)  loss_rpn_box_reg: 0.0626 (0.0613)  time: 1.6928  data: 0.6717  max mem: 19259\n",
      "Epoch: [0]  [1200/2444]  eta: 0:35:15  lr: 0.020000  loss: 0.8553 (1.1246)  loss_classifier: 0.2753 (0.4917)  loss_box_reg: 0.4547 (0.4978)  loss_objectness: 0.0718 (0.0739)  loss_rpn_box_reg: 0.0575 (0.0613)  time: 1.6892  data: 0.6676  max mem: 19259\n",
      "Epoch: [0]  [1210/2444]  eta: 0:34:58  lr: 0.020000  loss: 0.8230 (1.1221)  loss_classifier: 0.2606 (0.4897)  loss_box_reg: 0.4283 (0.4972)  loss_objectness: 0.0729 (0.0739)  loss_rpn_box_reg: 0.0572 (0.0613)  time: 1.6783  data: 0.6583  max mem: 19259\n",
      "Epoch: [0]  [1220/2444]  eta: 0:34:41  lr: 0.020000  loss: 0.8162 (1.1197)  loss_classifier: 0.2576 (0.4879)  loss_box_reg: 0.4331 (0.4968)  loss_objectness: 0.0603 (0.0737)  loss_rpn_box_reg: 0.0587 (0.0613)  time: 1.6780  data: 0.6588  max mem: 19259\n",
      "Epoch: [0]  [1230/2444]  eta: 0:34:24  lr: 0.020000  loss: 0.8249 (1.1176)  loss_classifier: 0.2636 (0.4861)  loss_box_reg: 0.4476 (0.4965)  loss_objectness: 0.0582 (0.0738)  loss_rpn_box_reg: 0.0615 (0.0613)  time: 1.6855  data: 0.6640  max mem: 19259\n",
      "Epoch: [0]  [1240/2444]  eta: 0:34:07  lr: 0.020000  loss: 0.8310 (1.1156)  loss_classifier: 0.2665 (0.4844)  loss_box_reg: 0.4477 (0.4962)  loss_objectness: 0.0609 (0.0737)  loss_rpn_box_reg: 0.0619 (0.0613)  time: 1.6823  data: 0.6608  max mem: 19259\n",
      "Epoch: [0]  [1250/2444]  eta: 0:33:49  lr: 0.020000  loss: 0.8558 (1.1135)  loss_classifier: 0.2683 (0.4827)  loss_box_reg: 0.4481 (0.4958)  loss_objectness: 0.0648 (0.0737)  loss_rpn_box_reg: 0.0603 (0.0613)  time: 1.6811  data: 0.6616  max mem: 19259\n",
      "Epoch: [0]  [1260/2444]  eta: 0:33:32  lr: 0.020000  loss: 0.8611 (1.1113)  loss_classifier: 0.2683 (0.4809)  loss_box_reg: 0.4446 (0.4954)  loss_objectness: 0.0648 (0.0737)  loss_rpn_box_reg: 0.0576 (0.0613)  time: 1.6766  data: 0.6567  max mem: 19259\n",
      "Epoch: [0]  [1270/2444]  eta: 0:33:15  lr: 0.020000  loss: 0.8680 (1.1095)  loss_classifier: 0.2774 (0.4793)  loss_box_reg: 0.4386 (0.4950)  loss_objectness: 0.0643 (0.0738)  loss_rpn_box_reg: 0.0598 (0.0613)  time: 1.6847  data: 0.6645  max mem: 19259\n",
      "Epoch: [0]  [1280/2444]  eta: 0:32:58  lr: 0.020000  loss: 0.8443 (1.1074)  loss_classifier: 0.2700 (0.4777)  loss_box_reg: 0.4369 (0.4946)  loss_objectness: 0.0643 (0.0738)  loss_rpn_box_reg: 0.0648 (0.0613)  time: 1.7017  data: 0.6824  max mem: 19259\n",
      "Epoch: [0]  [1290/2444]  eta: 0:32:41  lr: 0.020000  loss: 0.8357 (1.1053)  loss_classifier: 0.2607 (0.4760)  loss_box_reg: 0.4366 (0.4942)  loss_objectness: 0.0661 (0.0738)  loss_rpn_box_reg: 0.0599 (0.0613)  time: 1.6904  data: 0.6712  max mem: 19259\n",
      "Epoch: [0]  [1300/2444]  eta: 0:32:24  lr: 0.020000  loss: 0.8357 (1.1032)  loss_classifier: 0.2687 (0.4744)  loss_box_reg: 0.4358 (0.4938)  loss_objectness: 0.0673 (0.0737)  loss_rpn_box_reg: 0.0597 (0.0613)  time: 1.6765  data: 0.6567  max mem: 19259\n",
      "Epoch: [0]  [1310/2444]  eta: 0:32:07  lr: 0.020000  loss: 0.8404 (1.1016)  loss_classifier: 0.2718 (0.4729)  loss_box_reg: 0.4454 (0.4935)  loss_objectness: 0.0673 (0.0738)  loss_rpn_box_reg: 0.0573 (0.0614)  time: 1.6832  data: 0.6633  max mem: 19259\n",
      "Epoch: [0]  [1320/2444]  eta: 0:31:50  lr: 0.020000  loss: 0.8499 (1.0995)  loss_classifier: 0.2702 (0.4713)  loss_box_reg: 0.4528 (0.4931)  loss_objectness: 0.0645 (0.0737)  loss_rpn_box_reg: 0.0573 (0.0614)  time: 1.6856  data: 0.6654  max mem: 19259\n",
      "Epoch: [0]  [1330/2444]  eta: 0:31:33  lr: 0.020000  loss: 0.8650 (1.0976)  loss_classifier: 0.2658 (0.4697)  loss_box_reg: 0.4535 (0.4928)  loss_objectness: 0.0691 (0.0737)  loss_rpn_box_reg: 0.0576 (0.0613)  time: 1.6878  data: 0.6665  max mem: 19259\n",
      "Epoch: [0]  [1340/2444]  eta: 0:31:15  lr: 0.020000  loss: 0.8249 (1.0955)  loss_classifier: 0.2621 (0.4682)  loss_box_reg: 0.4482 (0.4924)  loss_objectness: 0.0691 (0.0737)  loss_rpn_box_reg: 0.0576 (0.0613)  time: 1.6918  data: 0.6712  max mem: 19259\n",
      "Epoch: [0]  [1350/2444]  eta: 0:30:58  lr: 0.020000  loss: 0.8203 (1.0936)  loss_classifier: 0.2542 (0.4666)  loss_box_reg: 0.4328 (0.4920)  loss_objectness: 0.0651 (0.0737)  loss_rpn_box_reg: 0.0585 (0.0613)  time: 1.6824  data: 0.6645  max mem: 19259\n",
      "Epoch: [0]  [1360/2444]  eta: 0:30:41  lr: 0.020000  loss: 0.8321 (1.0919)  loss_classifier: 0.2727 (0.4652)  loss_box_reg: 0.4430 (0.4916)  loss_objectness: 0.0736 (0.0738)  loss_rpn_box_reg: 0.0622 (0.0613)  time: 1.6800  data: 0.6626  max mem: 19259\n",
      "Epoch: [0]  [1370/2444]  eta: 0:30:24  lr: 0.020000  loss: 0.8321 (1.0900)  loss_classifier: 0.2689 (0.4637)  loss_box_reg: 0.4495 (0.4912)  loss_objectness: 0.0736 (0.0738)  loss_rpn_box_reg: 0.0577 (0.0613)  time: 1.6866  data: 0.6681  max mem: 19259\n",
      "Epoch: [0]  [1380/2444]  eta: 0:30:07  lr: 0.020000  loss: 0.8015 (1.0882)  loss_classifier: 0.2620 (0.4623)  loss_box_reg: 0.4443 (0.4909)  loss_objectness: 0.0637 (0.0737)  loss_rpn_box_reg: 0.0591 (0.0613)  time: 1.6794  data: 0.6608  max mem: 19259\n",
      "Epoch: [0]  [1390/2444]  eta: 0:29:50  lr: 0.020000  loss: 0.8143 (1.0865)  loss_classifier: 0.2575 (0.4610)  loss_box_reg: 0.4352 (0.4905)  loss_objectness: 0.0632 (0.0737)  loss_rpn_box_reg: 0.0610 (0.0614)  time: 1.6795  data: 0.6612  max mem: 19259\n",
      "Epoch: [0]  [1400/2444]  eta: 0:29:33  lr: 0.020000  loss: 0.8242 (1.0849)  loss_classifier: 0.2681 (0.4596)  loss_box_reg: 0.4387 (0.4902)  loss_objectness: 0.0750 (0.0737)  loss_rpn_box_reg: 0.0612 (0.0614)  time: 1.6869  data: 0.6684  max mem: 19259\n",
      "Epoch: [0]  [1410/2444]  eta: 0:29:16  lr: 0.020000  loss: 0.8592 (1.0832)  loss_classifier: 0.2700 (0.4583)  loss_box_reg: 0.4383 (0.4899)  loss_objectness: 0.0693 (0.0737)  loss_rpn_box_reg: 0.0644 (0.0614)  time: 1.6886  data: 0.6698  max mem: 19259\n",
      "Epoch: [0]  [1420/2444]  eta: 0:28:59  lr: 0.020000  loss: 0.8460 (1.0816)  loss_classifier: 0.2649 (0.4569)  loss_box_reg: 0.4369 (0.4896)  loss_objectness: 0.0693 (0.0737)  loss_rpn_box_reg: 0.0614 (0.0614)  time: 1.6854  data: 0.6652  max mem: 19259\n",
      "Epoch: [0]  [1430/2444]  eta: 0:28:42  lr: 0.020000  loss: 0.8237 (1.0797)  loss_classifier: 0.2579 (0.4555)  loss_box_reg: 0.4369 (0.4892)  loss_objectness: 0.0619 (0.0736)  loss_rpn_box_reg: 0.0586 (0.0614)  time: 1.6845  data: 0.6646  max mem: 19259\n",
      "Epoch: [0]  [1440/2444]  eta: 0:28:24  lr: 0.020000  loss: 0.8203 (1.0781)  loss_classifier: 0.2513 (0.4541)  loss_box_reg: 0.4284 (0.4888)  loss_objectness: 0.0711 (0.0737)  loss_rpn_box_reg: 0.0635 (0.0614)  time: 1.6864  data: 0.6676  max mem: 19259\n",
      "Epoch: [0]  [1450/2444]  eta: 0:28:07  lr: 0.020000  loss: 0.8534 (1.0765)  loss_classifier: 0.2600 (0.4529)  loss_box_reg: 0.4307 (0.4884)  loss_objectness: 0.0718 (0.0737)  loss_rpn_box_reg: 0.0641 (0.0614)  time: 1.6874  data: 0.6679  max mem: 19259\n",
      "Epoch: [0]  [1460/2444]  eta: 0:27:50  lr: 0.020000  loss: 0.8141 (1.0745)  loss_classifier: 0.2549 (0.4515)  loss_box_reg: 0.4252 (0.4879)  loss_objectness: 0.0696 (0.0737)  loss_rpn_box_reg: 0.0526 (0.0614)  time: 1.6836  data: 0.6658  max mem: 19259\n",
      "Epoch: [0]  [1470/2444]  eta: 0:27:33  lr: 0.020000  loss: 0.8091 (1.0727)  loss_classifier: 0.2475 (0.4502)  loss_box_reg: 0.4161 (0.4875)  loss_objectness: 0.0643 (0.0736)  loss_rpn_box_reg: 0.0547 (0.0614)  time: 1.6834  data: 0.6659  max mem: 19259\n",
      "Epoch: [0]  [1480/2444]  eta: 0:27:16  lr: 0.020000  loss: 0.7976 (1.0709)  loss_classifier: 0.2475 (0.4489)  loss_box_reg: 0.4215 (0.4871)  loss_objectness: 0.0621 (0.0736)  loss_rpn_box_reg: 0.0571 (0.0614)  time: 1.6879  data: 0.6695  max mem: 19259\n",
      "Epoch: [0]  [1490/2444]  eta: 0:26:59  lr: 0.020000  loss: 0.7650 (1.0687)  loss_classifier: 0.2348 (0.4474)  loss_box_reg: 0.3988 (0.4864)  loss_objectness: 0.0574 (0.0736)  loss_rpn_box_reg: 0.0543 (0.0613)  time: 1.6750  data: 0.6574  max mem: 19259\n",
      "Epoch: [0]  [1500/2444]  eta: 0:26:42  lr: 0.020000  loss: 0.7745 (1.0671)  loss_classifier: 0.2489 (0.4462)  loss_box_reg: 0.4053 (0.4860)  loss_objectness: 0.0681 (0.0736)  loss_rpn_box_reg: 0.0546 (0.0613)  time: 1.6783  data: 0.6603  max mem: 19259\n",
      "Epoch: [0]  [1510/2444]  eta: 0:26:25  lr: 0.020000  loss: 0.8151 (1.0656)  loss_classifier: 0.2636 (0.4450)  loss_box_reg: 0.4236 (0.4857)  loss_objectness: 0.0739 (0.0736)  loss_rpn_box_reg: 0.0582 (0.0613)  time: 1.6877  data: 0.6682  max mem: 19259\n",
      "Epoch: [0]  [1520/2444]  eta: 0:26:08  lr: 0.020000  loss: 0.8057 (1.0639)  loss_classifier: 0.2620 (0.4437)  loss_box_reg: 0.4287 (0.4853)  loss_objectness: 0.0759 (0.0736)  loss_rpn_box_reg: 0.0582 (0.0613)  time: 1.6847  data: 0.6631  max mem: 19259\n",
      "Epoch: [0]  [1530/2444]  eta: 0:25:51  lr: 0.020000  loss: 0.7816 (1.0621)  loss_classifier: 0.2412 (0.4424)  loss_box_reg: 0.4125 (0.4848)  loss_objectness: 0.0651 (0.0736)  loss_rpn_box_reg: 0.0570 (0.0613)  time: 1.6864  data: 0.6649  max mem: 19259\n",
      "Epoch: [0]  [1540/2444]  eta: 0:25:34  lr: 0.020000  loss: 0.8182 (1.0608)  loss_classifier: 0.2602 (0.4414)  loss_box_reg: 0.4159 (0.4845)  loss_objectness: 0.0722 (0.0736)  loss_rpn_box_reg: 0.0611 (0.0613)  time: 1.6869  data: 0.6675  max mem: 19259\n",
      "Epoch: [0]  [1550/2444]  eta: 0:25:17  lr: 0.020000  loss: 0.7813 (1.0589)  loss_classifier: 0.2481 (0.4400)  loss_box_reg: 0.4159 (0.4840)  loss_objectness: 0.0724 (0.0737)  loss_rpn_box_reg: 0.0551 (0.0613)  time: 1.6761  data: 0.6568  max mem: 19259\n",
      "Epoch: [0]  [1560/2444]  eta: 0:25:00  lr: 0.020000  loss: 0.7558 (1.0571)  loss_classifier: 0.2303 (0.4388)  loss_box_reg: 0.3981 (0.4834)  loss_objectness: 0.0654 (0.0736)  loss_rpn_box_reg: 0.0549 (0.0612)  time: 1.6753  data: 0.6548  max mem: 19259\n",
      "Epoch: [0]  [1570/2444]  eta: 0:24:43  lr: 0.020000  loss: 0.7889 (1.0558)  loss_classifier: 0.2551 (0.4377)  loss_box_reg: 0.4123 (0.4831)  loss_objectness: 0.0667 (0.0737)  loss_rpn_box_reg: 0.0580 (0.0613)  time: 1.6943  data: 0.6742  max mem: 19259\n",
      "Epoch: [0]  [1580/2444]  eta: 0:24:26  lr: 0.020000  loss: 0.8228 (1.0542)  loss_classifier: 0.2638 (0.4366)  loss_box_reg: 0.4226 (0.4827)  loss_objectness: 0.0720 (0.0737)  loss_rpn_box_reg: 0.0621 (0.0613)  time: 1.6940  data: 0.6737  max mem: 19259\n",
      "Epoch: [0]  [1590/2444]  eta: 0:24:09  lr: 0.020000  loss: 0.7787 (1.0525)  loss_classifier: 0.2435 (0.4354)  loss_box_reg: 0.4010 (0.4822)  loss_objectness: 0.0695 (0.0737)  loss_rpn_box_reg: 0.0619 (0.0613)  time: 1.6829  data: 0.6618  max mem: 19259\n",
      "Epoch: [0]  [1600/2444]  eta: 0:23:52  lr: 0.020000  loss: 0.7719 (1.0508)  loss_classifier: 0.2443 (0.4342)  loss_box_reg: 0.3902 (0.4816)  loss_objectness: 0.0645 (0.0737)  loss_rpn_box_reg: 0.0584 (0.0613)  time: 1.6797  data: 0.6597  max mem: 19259\n",
      "Epoch: [0]  [1610/2444]  eta: 0:23:35  lr: 0.020000  loss: 0.8003 (1.0494)  loss_classifier: 0.2538 (0.4332)  loss_box_reg: 0.3967 (0.4813)  loss_objectness: 0.0661 (0.0737)  loss_rpn_box_reg: 0.0580 (0.0613)  time: 1.6839  data: 0.6646  max mem: 19259\n",
      "Epoch: [0]  [1620/2444]  eta: 0:23:17  lr: 0.020000  loss: 0.8018 (1.0479)  loss_classifier: 0.2582 (0.4321)  loss_box_reg: 0.4091 (0.4809)  loss_objectness: 0.0627 (0.0736)  loss_rpn_box_reg: 0.0606 (0.0613)  time: 1.6864  data: 0.6666  max mem: 19259\n",
      "Epoch: [0]  [1630/2444]  eta: 0:23:00  lr: 0.020000  loss: 0.8030 (1.0465)  loss_classifier: 0.2663 (0.4311)  loss_box_reg: 0.4185 (0.4805)  loss_objectness: 0.0625 (0.0736)  loss_rpn_box_reg: 0.0633 (0.0613)  time: 1.6851  data: 0.6654  max mem: 19259\n",
      "Epoch: [0]  [1640/2444]  eta: 0:22:43  lr: 0.020000  loss: 0.8067 (1.0447)  loss_classifier: 0.2663 (0.4299)  loss_box_reg: 0.4185 (0.4799)  loss_objectness: 0.0651 (0.0736)  loss_rpn_box_reg: 0.0547 (0.0612)  time: 1.6850  data: 0.6650  max mem: 19870\n",
      "Epoch: [0]  [1650/2444]  eta: 0:22:26  lr: 0.020000  loss: 0.7724 (1.0432)  loss_classifier: 0.2516 (0.4289)  loss_box_reg: 0.4174 (0.4795)  loss_objectness: 0.0594 (0.0735)  loss_rpn_box_reg: 0.0509 (0.0612)  time: 1.6802  data: 0.6612  max mem: 19870\n",
      "Epoch: [0]  [1660/2444]  eta: 0:22:09  lr: 0.020000  loss: 0.8063 (1.0417)  loss_classifier: 0.2534 (0.4279)  loss_box_reg: 0.4031 (0.4791)  loss_objectness: 0.0594 (0.0735)  loss_rpn_box_reg: 0.0591 (0.0612)  time: 1.6748  data: 0.6572  max mem: 19870\n",
      "Epoch: [0]  [1670/2444]  eta: 0:21:52  lr: 0.020000  loss: 0.8076 (1.0405)  loss_classifier: 0.2580 (0.4269)  loss_box_reg: 0.4031 (0.4787)  loss_objectness: 0.0757 (0.0737)  loss_rpn_box_reg: 0.0639 (0.0613)  time: 1.6792  data: 0.6601  max mem: 19870\n",
      "Epoch: [0]  [1680/2444]  eta: 0:21:35  lr: 0.020000  loss: 0.8294 (1.0391)  loss_classifier: 0.2644 (0.4259)  loss_box_reg: 0.4076 (0.4783)  loss_objectness: 0.0636 (0.0736)  loss_rpn_box_reg: 0.0633 (0.0613)  time: 1.6857  data: 0.6662  max mem: 19870\n",
      "Epoch: [0]  [1690/2444]  eta: 0:21:18  lr: 0.020000  loss: 0.7628 (1.0375)  loss_classifier: 0.2504 (0.4248)  loss_box_reg: 0.3873 (0.4777)  loss_objectness: 0.0615 (0.0736)  loss_rpn_box_reg: 0.0611 (0.0613)  time: 1.6812  data: 0.6632  max mem: 19870\n",
      "Epoch: [0]  [1700/2444]  eta: 0:21:01  lr: 0.020000  loss: 0.7483 (1.0358)  loss_classifier: 0.2422 (0.4238)  loss_box_reg: 0.3843 (0.4772)  loss_objectness: 0.0639 (0.0736)  loss_rpn_box_reg: 0.0568 (0.0613)  time: 1.6857  data: 0.6667  max mem: 19870\n",
      "Epoch: [0]  [1710/2444]  eta: 0:20:44  lr: 0.020000  loss: 0.7766 (1.0345)  loss_classifier: 0.2531 (0.4229)  loss_box_reg: 0.3976 (0.4768)  loss_objectness: 0.0654 (0.0735)  loss_rpn_box_reg: 0.0576 (0.0613)  time: 1.6922  data: 0.6721  max mem: 19870\n",
      "Epoch: [0]  [1720/2444]  eta: 0:20:27  lr: 0.020000  loss: 0.7635 (1.0328)  loss_classifier: 0.2510 (0.4218)  loss_box_reg: 0.3928 (0.4763)  loss_objectness: 0.0630 (0.0735)  loss_rpn_box_reg: 0.0537 (0.0612)  time: 1.6855  data: 0.6660  max mem: 19870\n",
      "Epoch: [0]  [1730/2444]  eta: 0:20:10  lr: 0.020000  loss: 0.7633 (1.0314)  loss_classifier: 0.2479 (0.4209)  loss_box_reg: 0.3815 (0.4758)  loss_objectness: 0.0590 (0.0735)  loss_rpn_box_reg: 0.0591 (0.0613)  time: 1.6845  data: 0.6653  max mem: 19870\n",
      "Epoch: [0]  [1740/2444]  eta: 0:19:53  lr: 0.020000  loss: 0.7989 (1.0302)  loss_classifier: 0.2526 (0.4199)  loss_box_reg: 0.4071 (0.4754)  loss_objectness: 0.0694 (0.0736)  loss_rpn_box_reg: 0.0640 (0.0613)  time: 1.6958  data: 0.6770  max mem: 19870\n",
      "Epoch: [0]  [1750/2444]  eta: 0:19:36  lr: 0.020000  loss: 0.7956 (1.0287)  loss_classifier: 0.2553 (0.4190)  loss_box_reg: 0.3914 (0.4748)  loss_objectness: 0.0704 (0.0736)  loss_rpn_box_reg: 0.0573 (0.0613)  time: 1.6955  data: 0.6777  max mem: 19870\n",
      "Epoch: [0]  [1760/2444]  eta: 0:19:19  lr: 0.020000  loss: 0.7675 (1.0273)  loss_classifier: 0.2553 (0.4181)  loss_box_reg: 0.3866 (0.4744)  loss_objectness: 0.0628 (0.0736)  loss_rpn_box_reg: 0.0604 (0.0613)  time: 1.6830  data: 0.6669  max mem: 19870\n",
      "Epoch: [0]  [1770/2444]  eta: 0:19:02  lr: 0.020000  loss: 0.7767 (1.0259)  loss_classifier: 0.2520 (0.4171)  loss_box_reg: 0.3880 (0.4739)  loss_objectness: 0.0611 (0.0736)  loss_rpn_box_reg: 0.0638 (0.0613)  time: 1.6735  data: 0.6560  max mem: 19870\n",
      "Epoch: [0]  [1780/2444]  eta: 0:18:45  lr: 0.020000  loss: 0.7709 (1.0244)  loss_classifier: 0.2520 (0.4162)  loss_box_reg: 0.3845 (0.4733)  loss_objectness: 0.0664 (0.0737)  loss_rpn_box_reg: 0.0647 (0.0613)  time: 1.6688  data: 0.6498  max mem: 19870\n",
      "Epoch: [0]  [1790/2444]  eta: 0:18:28  lr: 0.020000  loss: 0.7387 (1.0228)  loss_classifier: 0.2450 (0.4152)  loss_box_reg: 0.3673 (0.4727)  loss_objectness: 0.0625 (0.0736)  loss_rpn_box_reg: 0.0546 (0.0613)  time: 1.6669  data: 0.6492  max mem: 19870\n",
      "Epoch: [0]  [1800/2444]  eta: 0:18:11  lr: 0.020000  loss: 0.7671 (1.0214)  loss_classifier: 0.2478 (0.4143)  loss_box_reg: 0.3753 (0.4721)  loss_objectness: 0.0625 (0.0737)  loss_rpn_box_reg: 0.0562 (0.0613)  time: 1.6717  data: 0.6536  max mem: 19870\n",
      "Epoch: [0]  [1810/2444]  eta: 0:17:54  lr: 0.020000  loss: 0.7671 (1.0200)  loss_classifier: 0.2527 (0.4135)  loss_box_reg: 0.3753 (0.4716)  loss_objectness: 0.0679 (0.0736)  loss_rpn_box_reg: 0.0574 (0.0613)  time: 1.6853  data: 0.6661  max mem: 19870\n",
      "Epoch: [0]  [1820/2444]  eta: 0:17:37  lr: 0.020000  loss: 0.7580 (1.0188)  loss_classifier: 0.2556 (0.4127)  loss_box_reg: 0.3815 (0.4712)  loss_objectness: 0.0634 (0.0736)  loss_rpn_box_reg: 0.0582 (0.0613)  time: 1.6926  data: 0.6750  max mem: 19870\n",
      "Epoch: [0]  [1830/2444]  eta: 0:17:20  lr: 0.020000  loss: 0.7751 (1.0175)  loss_classifier: 0.2613 (0.4118)  loss_box_reg: 0.3746 (0.4706)  loss_objectness: 0.0634 (0.0737)  loss_rpn_box_reg: 0.0656 (0.0613)  time: 1.6866  data: 0.6702  max mem: 19870\n",
      "Epoch: [0]  [1840/2444]  eta: 0:17:03  lr: 0.020000  loss: 0.7580 (1.0160)  loss_classifier: 0.2508 (0.4110)  loss_box_reg: 0.3782 (0.4701)  loss_objectness: 0.0612 (0.0736)  loss_rpn_box_reg: 0.0598 (0.0613)  time: 1.6770  data: 0.6598  max mem: 19870\n",
      "Epoch: [0]  [1850/2444]  eta: 0:16:46  lr: 0.020000  loss: 0.7487 (1.0147)  loss_classifier: 0.2508 (0.4101)  loss_box_reg: 0.3650 (0.4696)  loss_objectness: 0.0628 (0.0736)  loss_rpn_box_reg: 0.0588 (0.0614)  time: 1.6842  data: 0.6664  max mem: 19870\n",
      "Epoch: [0]  [1860/2444]  eta: 0:16:29  lr: 0.020000  loss: 0.7267 (1.0131)  loss_classifier: 0.2442 (0.4092)  loss_box_reg: 0.3519 (0.4689)  loss_objectness: 0.0606 (0.0736)  loss_rpn_box_reg: 0.0574 (0.0613)  time: 1.6896  data: 0.6709  max mem: 19870\n",
      "Epoch: [0]  [1870/2444]  eta: 0:16:12  lr: 0.020000  loss: 0.7193 (1.0120)  loss_classifier: 0.2496 (0.4085)  loss_box_reg: 0.3473 (0.4684)  loss_objectness: 0.0872 (0.0737)  loss_rpn_box_reg: 0.0574 (0.0614)  time: 1.6802  data: 0.6619  max mem: 19870\n",
      "Epoch: [0]  [1880/2444]  eta: 0:15:55  lr: 0.020000  loss: 0.7567 (1.0105)  loss_classifier: 0.2496 (0.4076)  loss_box_reg: 0.3497 (0.4678)  loss_objectness: 0.0883 (0.0738)  loss_rpn_box_reg: 0.0562 (0.0613)  time: 1.6732  data: 0.6560  max mem: 19870\n",
      "Epoch: [0]  [1890/2444]  eta: 0:15:38  lr: 0.020000  loss: 0.7535 (1.0091)  loss_classifier: 0.2442 (0.4068)  loss_box_reg: 0.3655 (0.4673)  loss_objectness: 0.0722 (0.0737)  loss_rpn_box_reg: 0.0522 (0.0613)  time: 1.6775  data: 0.6519  max mem: 19870\n",
      "Epoch: [0]  [1900/2444]  eta: 0:15:21  lr: 0.020000  loss: 0.7805 (1.0081)  loss_classifier: 0.2590 (0.4061)  loss_box_reg: 0.3850 (0.4669)  loss_objectness: 0.0720 (0.0737)  loss_rpn_box_reg: 0.0629 (0.0614)  time: 1.6918  data: 0.6657  max mem: 19870\n",
      "Epoch: [0]  [1910/2444]  eta: 0:15:04  lr: 0.020000  loss: 0.7716 (1.0067)  loss_classifier: 0.2667 (0.4053)  loss_box_reg: 0.3771 (0.4663)  loss_objectness: 0.0695 (0.0737)  loss_rpn_box_reg: 0.0629 (0.0614)  time: 1.6908  data: 0.6738  max mem: 19870\n",
      "Epoch: [0]  [1920/2444]  eta: 0:14:47  lr: 0.020000  loss: 0.7716 (1.0055)  loss_classifier: 0.2713 (0.4046)  loss_box_reg: 0.3669 (0.4658)  loss_objectness: 0.0641 (0.0737)  loss_rpn_box_reg: 0.0620 (0.0614)  time: 1.6785  data: 0.6619  max mem: 19870\n",
      "Epoch: [0]  [1930/2444]  eta: 0:14:30  lr: 0.020000  loss: 0.7684 (1.0040)  loss_classifier: 0.2674 (0.4038)  loss_box_reg: 0.3700 (0.4652)  loss_objectness: 0.0640 (0.0737)  loss_rpn_box_reg: 0.0596 (0.0613)  time: 1.6753  data: 0.6576  max mem: 19870\n",
      "Epoch: [0]  [1940/2444]  eta: 0:14:13  lr: 0.020000  loss: 0.7572 (1.0028)  loss_classifier: 0.2605 (0.4031)  loss_box_reg: 0.3639 (0.4647)  loss_objectness: 0.0675 (0.0737)  loss_rpn_box_reg: 0.0575 (0.0614)  time: 1.6884  data: 0.6690  max mem: 19870\n",
      "Epoch: [0]  [1950/2444]  eta: 0:13:56  lr: 0.020000  loss: 0.7202 (1.0015)  loss_classifier: 0.2561 (0.4023)  loss_box_reg: 0.3602 (0.4641)  loss_objectness: 0.0684 (0.0737)  loss_rpn_box_reg: 0.0593 (0.0614)  time: 1.6859  data: 0.6661  max mem: 19870\n",
      "Epoch: [0]  [1960/2444]  eta: 0:13:39  lr: 0.020000  loss: 0.7202 (1.0000)  loss_classifier: 0.2439 (0.4015)  loss_box_reg: 0.3459 (0.4635)  loss_objectness: 0.0658 (0.0737)  loss_rpn_box_reg: 0.0597 (0.0614)  time: 1.6722  data: 0.6540  max mem: 19870\n",
      "Epoch: [0]  [1970/2444]  eta: 0:13:22  lr: 0.020000  loss: 0.7355 (0.9988)  loss_classifier: 0.2465 (0.4008)  loss_box_reg: 0.3465 (0.4630)  loss_objectness: 0.0638 (0.0737)  loss_rpn_box_reg: 0.0599 (0.0614)  time: 1.6797  data: 0.6616  max mem: 19870\n",
      "Epoch: [0]  [1980/2444]  eta: 0:13:06  lr: 0.020000  loss: 0.7108 (0.9973)  loss_classifier: 0.2436 (0.4000)  loss_box_reg: 0.3353 (0.4623)  loss_objectness: 0.0646 (0.0737)  loss_rpn_box_reg: 0.0599 (0.0614)  time: 1.6792  data: 0.6616  max mem: 19870\n",
      "Epoch: [0]  [1990/2444]  eta: 0:12:49  lr: 0.020000  loss: 0.7108 (0.9962)  loss_classifier: 0.2425 (0.3993)  loss_box_reg: 0.3338 (0.4618)  loss_objectness: 0.0713 (0.0737)  loss_rpn_box_reg: 0.0586 (0.0614)  time: 1.6768  data: 0.6578  max mem: 19870\n",
      "Epoch: [0]  [2000/2444]  eta: 0:12:32  lr: 0.020000  loss: 0.7349 (0.9949)  loss_classifier: 0.2535 (0.3986)  loss_box_reg: 0.3523 (0.4612)  loss_objectness: 0.0817 (0.0737)  loss_rpn_box_reg: 0.0586 (0.0614)  time: 1.6849  data: 0.6646  max mem: 19870\n",
      "Epoch: [0]  [2010/2444]  eta: 0:12:15  lr: 0.020000  loss: 0.7481 (0.9937)  loss_classifier: 0.2539 (0.3979)  loss_box_reg: 0.3486 (0.4606)  loss_objectness: 0.0853 (0.0738)  loss_rpn_box_reg: 0.0599 (0.0614)  time: 1.6848  data: 0.6663  max mem: 19870\n",
      "Epoch: [0]  [2020/2444]  eta: 0:11:58  lr: 0.020000  loss: 0.7559 (0.9926)  loss_classifier: 0.2580 (0.3972)  loss_box_reg: 0.3494 (0.4602)  loss_objectness: 0.0791 (0.0738)  loss_rpn_box_reg: 0.0649 (0.0614)  time: 1.6887  data: 0.6693  max mem: 19870\n",
      "Epoch: [0]  [2030/2444]  eta: 0:11:41  lr: 0.020000  loss: 0.7559 (0.9913)  loss_classifier: 0.2622 (0.3965)  loss_box_reg: 0.3462 (0.4596)  loss_objectness: 0.0710 (0.0738)  loss_rpn_box_reg: 0.0587 (0.0614)  time: 1.6855  data: 0.6650  max mem: 19870\n",
      "Epoch: [0]  [2040/2444]  eta: 0:11:24  lr: 0.020000  loss: 0.6812 (0.9899)  loss_classifier: 0.2459 (0.3958)  loss_box_reg: 0.3306 (0.4590)  loss_objectness: 0.0643 (0.0737)  loss_rpn_box_reg: 0.0540 (0.0614)  time: 1.6790  data: 0.6596  max mem: 19870\n",
      "Epoch: [0]  [2050/2444]  eta: 0:11:07  lr: 0.020000  loss: 0.7354 (0.9887)  loss_classifier: 0.2562 (0.3951)  loss_box_reg: 0.3498 (0.4585)  loss_objectness: 0.0593 (0.0737)  loss_rpn_box_reg: 0.0596 (0.0614)  time: 1.6844  data: 0.6652  max mem: 19870\n",
      "Epoch: [0]  [2060/2444]  eta: 0:10:50  lr: 0.020000  loss: 0.7480 (0.9876)  loss_classifier: 0.2572 (0.3945)  loss_box_reg: 0.3513 (0.4579)  loss_objectness: 0.0797 (0.0738)  loss_rpn_box_reg: 0.0637 (0.0614)  time: 1.6880  data: 0.6700  max mem: 19870\n",
      "Epoch: [0]  [2070/2444]  eta: 0:10:33  lr: 0.020000  loss: 0.7055 (0.9863)  loss_classifier: 0.2447 (0.3938)  loss_box_reg: 0.3340 (0.4573)  loss_objectness: 0.0770 (0.0738)  loss_rpn_box_reg: 0.0616 (0.0614)  time: 1.6764  data: 0.6607  max mem: 19870\n",
      "Epoch: [0]  [2080/2444]  eta: 0:10:16  lr: 0.020000  loss: 0.7055 (0.9852)  loss_classifier: 0.2393 (0.3931)  loss_box_reg: 0.3364 (0.4568)  loss_objectness: 0.0723 (0.0739)  loss_rpn_box_reg: 0.0616 (0.0614)  time: 1.6677  data: 0.6516  max mem: 19870\n",
      "Epoch: [0]  [2090/2444]  eta: 0:09:59  lr: 0.020000  loss: 0.6934 (0.9837)  loss_classifier: 0.2373 (0.3923)  loss_box_reg: 0.3264 (0.4561)  loss_objectness: 0.0723 (0.0739)  loss_rpn_box_reg: 0.0634 (0.0614)  time: 1.6665  data: 0.6491  max mem: 19870\n",
      "Epoch: [0]  [2100/2444]  eta: 0:09:42  lr: 0.020000  loss: 0.6908 (0.9824)  loss_classifier: 0.2373 (0.3916)  loss_box_reg: 0.3245 (0.4555)  loss_objectness: 0.0575 (0.0738)  loss_rpn_box_reg: 0.0601 (0.0614)  time: 1.6773  data: 0.6535  max mem: 19870\n",
      "Epoch: [0]  [2110/2444]  eta: 0:09:25  lr: 0.020000  loss: 0.7203 (0.9813)  loss_classifier: 0.2511 (0.3910)  loss_box_reg: 0.3385 (0.4550)  loss_objectness: 0.0613 (0.0738)  loss_rpn_box_reg: 0.0609 (0.0614)  time: 1.6893  data: 0.6670  max mem: 19870\n",
      "Epoch: [0]  [2120/2444]  eta: 0:09:08  lr: 0.020000  loss: 0.7273 (0.9800)  loss_classifier: 0.2562 (0.3903)  loss_box_reg: 0.3393 (0.4545)  loss_objectness: 0.0659 (0.0738)  loss_rpn_box_reg: 0.0590 (0.0614)  time: 1.6755  data: 0.6597  max mem: 19870\n",
      "Epoch: [0]  [2130/2444]  eta: 0:08:51  lr: 0.020000  loss: 0.6825 (0.9787)  loss_classifier: 0.2475 (0.3896)  loss_box_reg: 0.3289 (0.4539)  loss_objectness: 0.0555 (0.0737)  loss_rpn_box_reg: 0.0544 (0.0614)  time: 1.6682  data: 0.6500  max mem: 19870\n",
      "Epoch: [0]  [2140/2444]  eta: 0:08:34  lr: 0.020000  loss: 0.6825 (0.9775)  loss_classifier: 0.2450 (0.3890)  loss_box_reg: 0.3289 (0.4534)  loss_objectness: 0.0532 (0.0737)  loss_rpn_box_reg: 0.0523 (0.0614)  time: 1.6705  data: 0.6529  max mem: 19870\n",
      "Epoch: [0]  [2150/2444]  eta: 0:08:17  lr: 0.020000  loss: 0.7350 (0.9766)  loss_classifier: 0.2575 (0.3885)  loss_box_reg: 0.3491 (0.4530)  loss_objectness: 0.0777 (0.0738)  loss_rpn_box_reg: 0.0604 (0.0614)  time: 1.6769  data: 0.6611  max mem: 19870\n",
      "Epoch: [0]  [2160/2444]  eta: 0:08:00  lr: 0.020000  loss: 0.7144 (0.9752)  loss_classifier: 0.2522 (0.3878)  loss_box_reg: 0.3407 (0.4524)  loss_objectness: 0.0665 (0.0737)  loss_rpn_box_reg: 0.0572 (0.0614)  time: 1.6803  data: 0.6628  max mem: 19870\n",
      "Epoch: [0]  [2170/2444]  eta: 0:07:43  lr: 0.020000  loss: 0.6922 (0.9738)  loss_classifier: 0.2399 (0.3871)  loss_box_reg: 0.3241 (0.4517)  loss_objectness: 0.0540 (0.0736)  loss_rpn_box_reg: 0.0581 (0.0614)  time: 1.6641  data: 0.6459  max mem: 19870\n",
      "Epoch: [0]  [2180/2444]  eta: 0:07:26  lr: 0.020000  loss: 0.7280 (0.9727)  loss_classifier: 0.2497 (0.3865)  loss_box_reg: 0.3332 (0.4512)  loss_objectness: 0.0702 (0.0737)  loss_rpn_box_reg: 0.0642 (0.0614)  time: 1.6649  data: 0.6469  max mem: 19870\n",
      "Epoch: [0]  [2190/2444]  eta: 0:07:09  lr: 0.020000  loss: 0.7349 (0.9716)  loss_classifier: 0.2584 (0.3858)  loss_box_reg: 0.3342 (0.4507)  loss_objectness: 0.0751 (0.0737)  loss_rpn_box_reg: 0.0636 (0.0614)  time: 1.6760  data: 0.6587  max mem: 19870\n",
      "Epoch: [0]  [2200/2444]  eta: 0:06:52  lr: 0.020000  loss: 0.6889 (0.9704)  loss_classifier: 0.2454 (0.3852)  loss_box_reg: 0.3319 (0.4502)  loss_objectness: 0.0587 (0.0736)  loss_rpn_box_reg: 0.0555 (0.0614)  time: 1.6688  data: 0.6532  max mem: 19870\n",
      "Epoch: [0]  [2210/2444]  eta: 0:06:35  lr: 0.020000  loss: 0.6999 (0.9691)  loss_classifier: 0.2570 (0.3846)  loss_box_reg: 0.3317 (0.4496)  loss_objectness: 0.0587 (0.0736)  loss_rpn_box_reg: 0.0552 (0.0614)  time: 1.6682  data: 0.6515  max mem: 19870\n",
      "Epoch: [0]  [2220/2444]  eta: 0:06:19  lr: 0.020000  loss: 0.7116 (0.9680)  loss_classifier: 0.2565 (0.3840)  loss_box_reg: 0.3212 (0.4490)  loss_objectness: 0.0700 (0.0736)  loss_rpn_box_reg: 0.0610 (0.0614)  time: 1.6732  data: 0.6549  max mem: 19870\n",
      "Epoch: [0]  [2230/2444]  eta: 0:06:02  lr: 0.020000  loss: 0.7041 (0.9668)  loss_classifier: 0.2511 (0.3834)  loss_box_reg: 0.3254 (0.4485)  loss_objectness: 0.0586 (0.0735)  loss_rpn_box_reg: 0.0619 (0.0614)  time: 1.6713  data: 0.6529  max mem: 19870\n",
      "Epoch: [0]  [2240/2444]  eta: 0:05:45  lr: 0.020000  loss: 0.6913 (0.9654)  loss_classifier: 0.2408 (0.3827)  loss_box_reg: 0.3180 (0.4479)  loss_objectness: 0.0604 (0.0735)  loss_rpn_box_reg: 0.0613 (0.0614)  time: 1.6664  data: 0.6470  max mem: 19870\n",
      "Epoch: [0]  [2250/2444]  eta: 0:05:28  lr: 0.020000  loss: 0.6957 (0.9643)  loss_classifier: 0.2441 (0.3821)  loss_box_reg: 0.3267 (0.4473)  loss_objectness: 0.0622 (0.0735)  loss_rpn_box_reg: 0.0613 (0.0614)  time: 1.6681  data: 0.6483  max mem: 19870\n",
      "Epoch: [0]  [2260/2444]  eta: 0:05:11  lr: 0.020000  loss: 0.7152 (0.9631)  loss_classifier: 0.2484 (0.3815)  loss_box_reg: 0.3305 (0.4468)  loss_objectness: 0.0650 (0.0735)  loss_rpn_box_reg: 0.0600 (0.0614)  time: 1.6715  data: 0.6531  max mem: 19870\n",
      "Epoch: [0]  [2270/2444]  eta: 0:04:54  lr: 0.020000  loss: 0.7404 (0.9621)  loss_classifier: 0.2521 (0.3809)  loss_box_reg: 0.3340 (0.4462)  loss_objectness: 0.0693 (0.0735)  loss_rpn_box_reg: 0.0606 (0.0614)  time: 1.6722  data: 0.6553  max mem: 19870\n",
      "Epoch: [0]  [2280/2444]  eta: 0:04:37  lr: 0.020000  loss: 0.7337 (0.9609)  loss_classifier: 0.2494 (0.3803)  loss_box_reg: 0.3213 (0.4457)  loss_objectness: 0.0660 (0.0735)  loss_rpn_box_reg: 0.0620 (0.0614)  time: 1.6811  data: 0.6635  max mem: 19870\n",
      "Epoch: [0]  [2290/2444]  eta: 0:04:20  lr: 0.020000  loss: 0.6793 (0.9598)  loss_classifier: 0.2481 (0.3798)  loss_box_reg: 0.3208 (0.4452)  loss_objectness: 0.0596 (0.0735)  loss_rpn_box_reg: 0.0566 (0.0614)  time: 1.6822  data: 0.6653  max mem: 19870\n",
      "Epoch: [0]  [2300/2444]  eta: 0:04:03  lr: 0.020000  loss: 0.6699 (0.9586)  loss_classifier: 0.2398 (0.3792)  loss_box_reg: 0.3148 (0.4447)  loss_objectness: 0.0596 (0.0734)  loss_rpn_box_reg: 0.0560 (0.0614)  time: 1.6806  data: 0.6587  max mem: 19870\n",
      "Epoch: [0]  [2310/2444]  eta: 0:03:46  lr: 0.020000  loss: 0.6865 (0.9576)  loss_classifier: 0.2416 (0.3786)  loss_box_reg: 0.3221 (0.4442)  loss_objectness: 0.0596 (0.0734)  loss_rpn_box_reg: 0.0543 (0.0614)  time: 1.6793  data: 0.6562  max mem: 19870\n",
      "Epoch: [0]  [2320/2444]  eta: 0:03:29  lr: 0.020000  loss: 0.7135 (0.9566)  loss_classifier: 0.2426 (0.3781)  loss_box_reg: 0.3221 (0.4436)  loss_objectness: 0.0718 (0.0735)  loss_rpn_box_reg: 0.0644 (0.0614)  time: 1.6720  data: 0.6542  max mem: 19870\n",
      "Epoch: [0]  [2330/2444]  eta: 0:03:12  lr: 0.020000  loss: 0.6980 (0.9555)  loss_classifier: 0.2463 (0.3775)  loss_box_reg: 0.3148 (0.4431)  loss_objectness: 0.0718 (0.0735)  loss_rpn_box_reg: 0.0633 (0.0614)  time: 1.6720  data: 0.6549  max mem: 19870\n",
      "Epoch: [0]  [2340/2444]  eta: 0:02:55  lr: 0.020000  loss: 0.6980 (0.9544)  loss_classifier: 0.2463 (0.3769)  loss_box_reg: 0.3148 (0.4426)  loss_objectness: 0.0666 (0.0734)  loss_rpn_box_reg: 0.0575 (0.0614)  time: 1.6740  data: 0.6578  max mem: 19870\n",
      "Epoch: [0]  [2350/2444]  eta: 0:02:38  lr: 0.020000  loss: 0.7043 (0.9534)  loss_classifier: 0.2485 (0.3764)  loss_box_reg: 0.3266 (0.4421)  loss_objectness: 0.0652 (0.0734)  loss_rpn_box_reg: 0.0569 (0.0614)  time: 1.6731  data: 0.6568  max mem: 19870\n",
      "Epoch: [0]  [2360/2444]  eta: 0:02:22  lr: 0.020000  loss: 0.7043 (0.9523)  loss_classifier: 0.2437 (0.3759)  loss_box_reg: 0.3232 (0.4416)  loss_objectness: 0.0684 (0.0734)  loss_rpn_box_reg: 0.0603 (0.0614)  time: 1.6751  data: 0.6592  max mem: 19870\n",
      "Epoch: [0]  [2370/2444]  eta: 0:02:05  lr: 0.020000  loss: 0.6860 (0.9514)  loss_classifier: 0.2437 (0.3754)  loss_box_reg: 0.3219 (0.4412)  loss_objectness: 0.0684 (0.0734)  loss_rpn_box_reg: 0.0604 (0.0614)  time: 1.6739  data: 0.6580  max mem: 19870\n",
      "Epoch: [0]  [2380/2444]  eta: 0:01:48  lr: 0.020000  loss: 0.7192 (0.9505)  loss_classifier: 0.2587 (0.3749)  loss_box_reg: 0.3272 (0.4407)  loss_objectness: 0.0772 (0.0735)  loss_rpn_box_reg: 0.0597 (0.0614)  time: 1.6645  data: 0.6489  max mem: 19870\n",
      "Epoch: [0]  [2390/2444]  eta: 0:01:31  lr: 0.020000  loss: 0.7135 (0.9495)  loss_classifier: 0.2587 (0.3744)  loss_box_reg: 0.3272 (0.4402)  loss_objectness: 0.0734 (0.0735)  loss_rpn_box_reg: 0.0598 (0.0614)  time: 1.6704  data: 0.6555  max mem: 19870\n",
      "Epoch: [0]  [2400/2444]  eta: 0:01:14  lr: 0.020000  loss: 0.7014 (0.9486)  loss_classifier: 0.2506 (0.3739)  loss_box_reg: 0.3226 (0.4398)  loss_objectness: 0.0697 (0.0736)  loss_rpn_box_reg: 0.0615 (0.0614)  time: 1.6764  data: 0.6617  max mem: 19870\n",
      "Epoch: [0]  [2410/2444]  eta: 0:00:57  lr: 0.020000  loss: 0.7399 (0.9477)  loss_classifier: 0.2538 (0.3734)  loss_box_reg: 0.3305 (0.4393)  loss_objectness: 0.0741 (0.0736)  loss_rpn_box_reg: 0.0638 (0.0615)  time: 1.6771  data: 0.6617  max mem: 19870\n",
      "Epoch: [0]  [2420/2444]  eta: 0:00:40  lr: 0.020000  loss: 0.6952 (0.9468)  loss_classifier: 0.2448 (0.3729)  loss_box_reg: 0.3261 (0.4388)  loss_objectness: 0.0741 (0.0736)  loss_rpn_box_reg: 0.0612 (0.0615)  time: 1.6794  data: 0.6629  max mem: 19870\n",
      "Epoch: [0]  [2430/2444]  eta: 0:00:23  lr: 0.020000  loss: 0.6952 (0.9458)  loss_classifier: 0.2489 (0.3724)  loss_box_reg: 0.3284 (0.4384)  loss_objectness: 0.0670 (0.0736)  loss_rpn_box_reg: 0.0595 (0.0615)  time: 1.6777  data: 0.6611  max mem: 19870\n",
      "Epoch: [0]  [2440/2444]  eta: 0:00:06  lr: 0.020000  loss: 0.6944 (0.9448)  loss_classifier: 0.2567 (0.3719)  loss_box_reg: 0.3284 (0.4379)  loss_objectness: 0.0617 (0.0736)  loss_rpn_box_reg: 0.0605 (0.0615)  time: 1.6763  data: 0.6590  max mem: 19870\n",
      "Epoch: [0]  [2443/2444]  eta: 0:00:01  lr: 0.020000  loss: 0.6827 (0.9444)  loss_classifier: 0.2441 (0.3717)  loss_box_reg: 0.3205 (0.4378)  loss_objectness: 0.0603 (0.0735)  loss_rpn_box_reg: 0.0559 (0.0614)  time: 1.5958  data: 0.6266  max mem: 19870\n",
      "Epoch: [0] Total time: 1:08:50 (1.6899 s / it)\n",
      "creating index...\n",
      "index created!\n",
      "Test:  [   0/4952]  eta: 0:34:29  model_time: 0.3787 (0.3787)  evaluator_time: 0.0177 (0.0177)  time: 0.4179  data: 0.0202  max mem: 19870\n",
      "Test:  [ 100/4952]  eta: 0:09:52  model_time: 0.0332 (0.0966)  evaluator_time: 0.0088 (0.0099)  time: 0.0620  data: 0.0168  max mem: 19870\n",
      "Test:  [ 200/4952]  eta: 0:07:11  model_time: 0.0327 (0.0652)  evaluator_time: 0.0052 (0.0099)  time: 0.0596  data: 0.0147  max mem: 19870\n",
      "Test:  [ 300/4952]  eta: 0:06:16  model_time: 0.0329 (0.0546)  evaluator_time: 0.0078 (0.0105)  time: 0.0631  data: 0.0161  max mem: 19870\n",
      "Test:  [ 400/4952]  eta: 0:05:42  model_time: 0.0328 (0.0493)  evaluator_time: 0.0064 (0.0103)  time: 0.0581  data: 0.0144  max mem: 19870\n",
      "Test:  [ 500/4952]  eta: 0:05:19  model_time: 0.0327 (0.0461)  evaluator_time: 0.0066 (0.0101)  time: 0.0587  data: 0.0148  max mem: 19870\n",
      "Test:  [ 600/4952]  eta: 0:05:04  model_time: 0.0328 (0.0440)  evaluator_time: 0.0055 (0.0103)  time: 0.0563  data: 0.0141  max mem: 19870\n",
      "Test:  [ 700/4952]  eta: 0:04:51  model_time: 0.0327 (0.0425)  evaluator_time: 0.0059 (0.0104)  time: 0.0573  data: 0.0149  max mem: 19870\n",
      "Test:  [ 800/4952]  eta: 0:04:40  model_time: 0.0337 (0.0414)  evaluator_time: 0.0088 (0.0104)  time: 0.0625  data: 0.0159  max mem: 19870\n",
      "Test:  [ 900/4952]  eta: 0:04:30  model_time: 0.0329 (0.0405)  evaluator_time: 0.0055 (0.0107)  time: 0.0680  data: 0.0141  max mem: 19870\n",
      "Test:  [1000/4952]  eta: 0:04:20  model_time: 0.0326 (0.0397)  evaluator_time: 0.0062 (0.0105)  time: 0.0546  data: 0.0135  max mem: 19870\n",
      "Test:  [1100/4952]  eta: 0:04:11  model_time: 0.0325 (0.0391)  evaluator_time: 0.0053 (0.0105)  time: 0.0558  data: 0.0143  max mem: 19870\n",
      "Test:  [1200/4952]  eta: 0:04:03  model_time: 0.0335 (0.0387)  evaluator_time: 0.0115 (0.0106)  time: 0.0665  data: 0.0168  max mem: 19870\n",
      "Test:  [1300/4952]  eta: 0:03:55  model_time: 0.0325 (0.0382)  evaluator_time: 0.0058 (0.0106)  time: 0.0599  data: 0.0151  max mem: 19870\n",
      "Test:  [1400/4952]  eta: 0:03:47  model_time: 0.0328 (0.0379)  evaluator_time: 0.0060 (0.0106)  time: 0.0594  data: 0.0148  max mem: 19870\n",
      "Test:  [1500/4952]  eta: 0:03:40  model_time: 0.0324 (0.0376)  evaluator_time: 0.0067 (0.0105)  time: 0.0600  data: 0.0159  max mem: 19870\n",
      "Test:  [1600/4952]  eta: 0:03:33  model_time: 0.0326 (0.0373)  evaluator_time: 0.0072 (0.0106)  time: 0.0617  data: 0.0157  max mem: 19870\n",
      "Test:  [1700/4952]  eta: 0:03:26  model_time: 0.0326 (0.0371)  evaluator_time: 0.0052 (0.0107)  time: 0.0738  data: 0.0151  max mem: 19870\n",
      "Test:  [1800/4952]  eta: 0:03:19  model_time: 0.0334 (0.0369)  evaluator_time: 0.0054 (0.0107)  time: 0.0604  data: 0.0147  max mem: 19870\n",
      "Test:  [1900/4952]  eta: 0:03:12  model_time: 0.0327 (0.0367)  evaluator_time: 0.0064 (0.0106)  time: 0.0573  data: 0.0145  max mem: 19870\n",
      "Test:  [2000/4952]  eta: 0:03:05  model_time: 0.0330 (0.0365)  evaluator_time: 0.0082 (0.0106)  time: 0.0631  data: 0.0161  max mem: 19870\n",
      "Test:  [2100/4952]  eta: 0:02:58  model_time: 0.0328 (0.0363)  evaluator_time: 0.0071 (0.0106)  time: 0.0607  data: 0.0156  max mem: 19870\n",
      "Test:  [2200/4952]  eta: 0:02:51  model_time: 0.0336 (0.0362)  evaluator_time: 0.0065 (0.0106)  time: 0.0633  data: 0.0157  max mem: 19870\n",
      "Test:  [2300/4952]  eta: 0:02:45  model_time: 0.0330 (0.0361)  evaluator_time: 0.0058 (0.0105)  time: 0.0624  data: 0.0157  max mem: 19870\n",
      "Test:  [2400/4952]  eta: 0:02:38  model_time: 0.0327 (0.0359)  evaluator_time: 0.0100 (0.0105)  time: 0.0639  data: 0.0164  max mem: 19870\n",
      "Test:  [2500/4952]  eta: 0:02:32  model_time: 0.0329 (0.0358)  evaluator_time: 0.0071 (0.0105)  time: 0.0621  data: 0.0167  max mem: 19870\n",
      "Test:  [2600/4952]  eta: 0:02:25  model_time: 0.0328 (0.0357)  evaluator_time: 0.0068 (0.0105)  time: 0.0596  data: 0.0141  max mem: 19870\n",
      "Test:  [2700/4952]  eta: 0:02:19  model_time: 0.0324 (0.0356)  evaluator_time: 0.0070 (0.0105)  time: 0.0582  data: 0.0143  max mem: 19870\n",
      "Test:  [2800/4952]  eta: 0:02:13  model_time: 0.0328 (0.0355)  evaluator_time: 0.0075 (0.0105)  time: 0.0596  data: 0.0149  max mem: 19870\n",
      "Test:  [2900/4952]  eta: 0:02:06  model_time: 0.0328 (0.0355)  evaluator_time: 0.0058 (0.0105)  time: 0.0621  data: 0.0162  max mem: 19870\n",
      "Test:  [3000/4952]  eta: 0:02:00  model_time: 0.0329 (0.0354)  evaluator_time: 0.0080 (0.0104)  time: 0.0600  data: 0.0149  max mem: 19870\n",
      "Test:  [3100/4952]  eta: 0:01:54  model_time: 0.0327 (0.0353)  evaluator_time: 0.0069 (0.0104)  time: 0.0611  data: 0.0154  max mem: 19870\n",
      "Test:  [3200/4952]  eta: 0:01:47  model_time: 0.0326 (0.0353)  evaluator_time: 0.0075 (0.0104)  time: 0.0609  data: 0.0157  max mem: 19870\n",
      "Test:  [3300/4952]  eta: 0:01:41  model_time: 0.0326 (0.0352)  evaluator_time: 0.0067 (0.0104)  time: 0.0594  data: 0.0154  max mem: 19870\n",
      "Test:  [3400/4952]  eta: 0:01:35  model_time: 0.0329 (0.0351)  evaluator_time: 0.0049 (0.0105)  time: 0.0610  data: 0.0155  max mem: 19870\n",
      "Test:  [3500/4952]  eta: 0:01:29  model_time: 0.0327 (0.0351)  evaluator_time: 0.0074 (0.0105)  time: 0.0552  data: 0.0136  max mem: 19870\n",
      "Test:  [3600/4952]  eta: 0:01:22  model_time: 0.0331 (0.0350)  evaluator_time: 0.0073 (0.0105)  time: 0.0578  data: 0.0154  max mem: 19870\n",
      "Test:  [3700/4952]  eta: 0:01:16  model_time: 0.0325 (0.0350)  evaluator_time: 0.0057 (0.0104)  time: 0.0561  data: 0.0142  max mem: 19870\n",
      "Test:  [3800/4952]  eta: 0:01:10  model_time: 0.0327 (0.0349)  evaluator_time: 0.0060 (0.0104)  time: 0.0549  data: 0.0135  max mem: 19870\n",
      "Test:  [3900/4952]  eta: 0:01:04  model_time: 0.0328 (0.0349)  evaluator_time: 0.0071 (0.0104)  time: 0.0574  data: 0.0147  max mem: 19870\n",
      "Test:  [4000/4952]  eta: 0:00:58  model_time: 0.0332 (0.0348)  evaluator_time: 0.0088 (0.0104)  time: 0.0600  data: 0.0148  max mem: 19870\n",
      "Test:  [4100/4952]  eta: 0:00:52  model_time: 0.0330 (0.0348)  evaluator_time: 0.0063 (0.0104)  time: 0.0584  data: 0.0148  max mem: 19870\n",
      "Test:  [4200/4952]  eta: 0:00:45  model_time: 0.0323 (0.0348)  evaluator_time: 0.0050 (0.0104)  time: 0.0561  data: 0.0135  max mem: 19870\n",
      "Test:  [4300/4952]  eta: 0:00:39  model_time: 0.0327 (0.0347)  evaluator_time: 0.0059 (0.0104)  time: 0.0566  data: 0.0139  max mem: 19870\n",
      "Test:  [4400/4952]  eta: 0:00:33  model_time: 0.0326 (0.0347)  evaluator_time: 0.0080 (0.0104)  time: 0.0578  data: 0.0148  max mem: 19870\n",
      "Test:  [4500/4952]  eta: 0:00:27  model_time: 0.0327 (0.0347)  evaluator_time: 0.0074 (0.0104)  time: 0.0608  data: 0.0156  max mem: 19870\n",
      "Test:  [4600/4952]  eta: 0:00:21  model_time: 0.0325 (0.0346)  evaluator_time: 0.0059 (0.0104)  time: 0.0564  data: 0.0150  max mem: 19870\n",
      "Test:  [4700/4952]  eta: 0:00:15  model_time: 0.0328 (0.0346)  evaluator_time: 0.0053 (0.0103)  time: 0.0559  data: 0.0144  max mem: 19870\n",
      "Test:  [4800/4952]  eta: 0:00:09  model_time: 0.0325 (0.0346)  evaluator_time: 0.0074 (0.0103)  time: 0.0588  data: 0.0150  max mem: 19870\n",
      "Test:  [4900/4952]  eta: 0:00:03  model_time: 0.0330 (0.0345)  evaluator_time: 0.0048 (0.0103)  time: 0.0559  data: 0.0136  max mem: 19870\n",
      "Test:  [4951/4952]  eta: 0:00:00  model_time: 0.0329 (0.0345)  evaluator_time: 0.0063 (0.0103)  time: 0.0581  data: 0.0139  max mem: 19870\n",
      "Test: Total time: 0:05:00 (0.0608 s / it)\n",
      "Averaged stats: model_time: 0.0329 (0.0345)  evaluator_time: 0.0063 (0.0103)\n",
      "Accumulating evaluation results...\n",
      "DONE (t=7.84s).\n",
      "IoU metric: bbox\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.308\n",
      " Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.528\n",
      " Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.324\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.126\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.306\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.418\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.268\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.446\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.475\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.260\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.478\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.582\n",
      "Epoch: [1]  [   0/2444]  eta: 1:12:19  lr: 0.020000  loss: 0.6757 (0.6757)  loss_classifier: 0.2464 (0.2464)  loss_box_reg: 0.3180 (0.3180)  loss_objectness: 0.0504 (0.0504)  loss_rpn_box_reg: 0.0608 (0.0608)  time: 1.7755  data: 0.7321  max mem: 19870\n",
      "Epoch: [1]  [  10/2444]  eta: 1:08:48  lr: 0.020000  loss: 0.6757 (0.6913)  loss_classifier: 0.2419 (0.2410)  loss_box_reg: 0.3271 (0.3181)  loss_objectness: 0.0623 (0.0720)  loss_rpn_box_reg: 0.0599 (0.0602)  time: 1.6962  data: 0.6754  max mem: 19870\n",
      "Epoch: [1]  [  20/2444]  eta: 1:08:30  lr: 0.020000  loss: 0.6907 (0.6970)  loss_classifier: 0.2455 (0.2455)  loss_box_reg: 0.3201 (0.3182)  loss_objectness: 0.0722 (0.0742)  loss_rpn_box_reg: 0.0591 (0.0591)  time: 1.6917  data: 0.6733  max mem: 19870\n",
      "Epoch: [1]  [  30/2444]  eta: 1:08:23  lr: 0.020000  loss: 0.7041 (0.6938)  loss_classifier: 0.2463 (0.2449)  loss_box_reg: 0.3179 (0.3179)  loss_objectness: 0.0722 (0.0718)  loss_rpn_box_reg: 0.0591 (0.0592)  time: 1.7021  data: 0.6826  max mem: 19870\n",
      "Epoch: [1]  [  40/2444]  eta: 1:08:00  lr: 0.020000  loss: 0.7140 (0.6978)  loss_classifier: 0.2460 (0.2456)  loss_box_reg: 0.3196 (0.3188)  loss_objectness: 0.0687 (0.0734)  loss_rpn_box_reg: 0.0591 (0.0600)  time: 1.6989  data: 0.6775  max mem: 19870\n",
      "Epoch: [1]  [  50/2444]  eta: 1:07:45  lr: 0.020000  loss: 0.7303 (0.7043)  loss_classifier: 0.2586 (0.2484)  loss_box_reg: 0.3311 (0.3221)  loss_objectness: 0.0706 (0.0737)  loss_rpn_box_reg: 0.0588 (0.0602)  time: 1.6950  data: 0.6744  max mem: 19870\n",
      "Epoch: [1]  [  60/2444]  eta: 1:07:29  lr: 0.020000  loss: 0.7185 (0.7069)  loss_classifier: 0.2564 (0.2490)  loss_box_reg: 0.3366 (0.3230)  loss_objectness: 0.0707 (0.0748)  loss_rpn_box_reg: 0.0578 (0.0600)  time: 1.7018  data: 0.6812  max mem: 19870\n",
      "Epoch: [1]  [  70/2444]  eta: 1:07:11  lr: 0.020000  loss: 0.7014 (0.7047)  loss_classifier: 0.2466 (0.2487)  loss_box_reg: 0.3154 (0.3225)  loss_objectness: 0.0699 (0.0737)  loss_rpn_box_reg: 0.0571 (0.0597)  time: 1.6984  data: 0.6775  max mem: 19870\n",
      "Epoch: [1]  [  80/2444]  eta: 1:06:53  lr: 0.020000  loss: 0.6852 (0.7013)  loss_classifier: 0.2507 (0.2477)  loss_box_reg: 0.3154 (0.3217)  loss_objectness: 0.0602 (0.0725)  loss_rpn_box_reg: 0.0571 (0.0594)  time: 1.6940  data: 0.6744  max mem: 19870\n",
      "Epoch: [1]  [  90/2444]  eta: 1:06:35  lr: 0.020000  loss: 0.6855 (0.7020)  loss_classifier: 0.2475 (0.2479)  loss_box_reg: 0.3206 (0.3226)  loss_objectness: 0.0647 (0.0719)  loss_rpn_box_reg: 0.0623 (0.0596)  time: 1.6949  data: 0.6758  max mem: 19870\n",
      "Epoch: [1]  [ 100/2444]  eta: 1:06:16  lr: 0.020000  loss: 0.6896 (0.7054)  loss_classifier: 0.2475 (0.2485)  loss_box_reg: 0.3277 (0.3232)  loss_objectness: 0.0720 (0.0736)  loss_rpn_box_reg: 0.0626 (0.0600)  time: 1.6912  data: 0.6729  max mem: 19870\n",
      "Epoch: [1]  [ 110/2444]  eta: 1:05:55  lr: 0.020000  loss: 0.7558 (0.7063)  loss_classifier: 0.2543 (0.2494)  loss_box_reg: 0.3324 (0.3238)  loss_objectness: 0.0745 (0.0731)  loss_rpn_box_reg: 0.0571 (0.0599)  time: 1.6819  data: 0.6646  max mem: 19870\n",
      "Epoch: [1]  [ 120/2444]  eta: 1:05:38  lr: 0.020000  loss: 0.7356 (0.7055)  loss_classifier: 0.2518 (0.2488)  loss_box_reg: 0.3206 (0.3230)  loss_objectness: 0.0697 (0.0732)  loss_rpn_box_reg: 0.0602 (0.0606)  time: 1.6876  data: 0.6699  max mem: 19870\n",
      "Epoch: [1]  [ 130/2444]  eta: 1:05:21  lr: 0.020000  loss: 0.6729 (0.7044)  loss_classifier: 0.2400 (0.2484)  loss_box_reg: 0.3107 (0.3231)  loss_objectness: 0.0635 (0.0721)  loss_rpn_box_reg: 0.0609 (0.0609)  time: 1.6956  data: 0.6761  max mem: 19870\n",
      "Epoch: [1]  [ 140/2444]  eta: 1:05:05  lr: 0.020000  loss: 0.7021 (0.7054)  loss_classifier: 0.2505 (0.2490)  loss_box_reg: 0.3294 (0.3234)  loss_objectness: 0.0635 (0.0719)  loss_rpn_box_reg: 0.0609 (0.0612)  time: 1.6976  data: 0.6784  max mem: 19870\n",
      "Epoch: [1]  [ 150/2444]  eta: 1:04:51  lr: 0.020000  loss: 0.7044 (0.7058)  loss_classifier: 0.2584 (0.2492)  loss_box_reg: 0.3336 (0.3240)  loss_objectness: 0.0666 (0.0716)  loss_rpn_box_reg: 0.0594 (0.0610)  time: 1.7066  data: 0.6878  max mem: 19870\n",
      "Epoch: [1]  [ 160/2444]  eta: 1:04:34  lr: 0.020000  loss: 0.6970 (0.7055)  loss_classifier: 0.2465 (0.2492)  loss_box_reg: 0.3262 (0.3239)  loss_objectness: 0.0665 (0.0715)  loss_rpn_box_reg: 0.0582 (0.0610)  time: 1.7044  data: 0.6853  max mem: 19870\n",
      "Epoch: [1]  [ 170/2444]  eta: 1:04:17  lr: 0.020000  loss: 0.7042 (0.7060)  loss_classifier: 0.2463 (0.2490)  loss_box_reg: 0.3207 (0.3241)  loss_objectness: 0.0710 (0.0718)  loss_rpn_box_reg: 0.0588 (0.0611)  time: 1.6960  data: 0.6773  max mem: 19870\n",
      "Epoch: [1]  [ 180/2444]  eta: 1:03:59  lr: 0.020000  loss: 0.7044 (0.7055)  loss_classifier: 0.2497 (0.2488)  loss_box_reg: 0.3105 (0.3238)  loss_objectness: 0.0748 (0.0720)  loss_rpn_box_reg: 0.0566 (0.0608)  time: 1.6930  data: 0.6741  max mem: 19870\n",
      "Epoch: [1]  [ 190/2444]  eta: 1:03:42  lr: 0.020000  loss: 0.7115 (0.7065)  loss_classifier: 0.2517 (0.2491)  loss_box_reg: 0.3183 (0.3242)  loss_objectness: 0.0687 (0.0721)  loss_rpn_box_reg: 0.0569 (0.0611)  time: 1.6927  data: 0.6732  max mem: 19870\n",
      "Epoch: [1]  [ 200/2444]  eta: 1:03:26  lr: 0.020000  loss: 0.7185 (0.7064)  loss_classifier: 0.2541 (0.2489)  loss_box_reg: 0.3284 (0.3241)  loss_objectness: 0.0687 (0.0722)  loss_rpn_box_reg: 0.0630 (0.0611)  time: 1.6967  data: 0.6758  max mem: 19870\n",
      "Epoch: [1]  [ 210/2444]  eta: 1:03:12  lr: 0.020000  loss: 0.7093 (0.7066)  loss_classifier: 0.2518 (0.2491)  loss_box_reg: 0.3285 (0.3241)  loss_objectness: 0.0759 (0.0723)  loss_rpn_box_reg: 0.0593 (0.0611)  time: 1.7138  data: 0.6921  max mem: 19870\n",
      "Epoch: [1]  [ 220/2444]  eta: 1:02:56  lr: 0.020000  loss: 0.7145 (0.7076)  loss_classifier: 0.2492 (0.2494)  loss_box_reg: 0.3207 (0.3244)  loss_objectness: 0.0733 (0.0727)  loss_rpn_box_reg: 0.0596 (0.0611)  time: 1.7171  data: 0.6968  max mem: 19870\n",
      "Epoch: [1]  [ 230/2444]  eta: 1:02:42  lr: 0.020000  loss: 0.7210 (0.7082)  loss_classifier: 0.2492 (0.2496)  loss_box_reg: 0.3245 (0.3246)  loss_objectness: 0.0733 (0.0728)  loss_rpn_box_reg: 0.0603 (0.0612)  time: 1.7209  data: 0.7008  max mem: 19870\n",
      "Epoch: [1]  [ 240/2444]  eta: 1:02:24  lr: 0.020000  loss: 0.6803 (0.7068)  loss_classifier: 0.2434 (0.2495)  loss_box_reg: 0.3230 (0.3241)  loss_objectness: 0.0687 (0.0724)  loss_rpn_box_reg: 0.0555 (0.0608)  time: 1.7118  data: 0.6906  max mem: 19870\n",
      "Epoch: [1]  [ 250/2444]  eta: 1:02:06  lr: 0.020000  loss: 0.6605 (0.7063)  loss_classifier: 0.2379 (0.2494)  loss_box_reg: 0.3146 (0.3235)  loss_objectness: 0.0669 (0.0725)  loss_rpn_box_reg: 0.0536 (0.0609)  time: 1.6873  data: 0.6659  max mem: 19870\n",
      "Epoch: [1]  [ 260/2444]  eta: 1:01:50  lr: 0.020000  loss: 0.6917 (0.7068)  loss_classifier: 0.2547 (0.2497)  loss_box_reg: 0.3149 (0.3237)  loss_objectness: 0.0741 (0.0726)  loss_rpn_box_reg: 0.0585 (0.0608)  time: 1.6939  data: 0.6733  max mem: 19870\n",
      "Epoch: [1]  [ 270/2444]  eta: 1:01:31  lr: 0.020000  loss: 0.6896 (0.7056)  loss_classifier: 0.2474 (0.2493)  loss_box_reg: 0.3160 (0.3229)  loss_objectness: 0.0672 (0.0725)  loss_rpn_box_reg: 0.0585 (0.0608)  time: 1.6915  data: 0.6722  max mem: 19870\n",
      "Epoch: [1]  [ 280/2444]  eta: 1:01:15  lr: 0.020000  loss: 0.6658 (0.7054)  loss_classifier: 0.2334 (0.2492)  loss_box_reg: 0.3136 (0.3229)  loss_objectness: 0.0614 (0.0724)  loss_rpn_box_reg: 0.0593 (0.0609)  time: 1.6918  data: 0.6717  max mem: 19870\n",
      "Epoch: [1]  [ 290/2444]  eta: 1:00:56  lr: 0.020000  loss: 0.6884 (0.7048)  loss_classifier: 0.2432 (0.2490)  loss_box_reg: 0.3187 (0.3227)  loss_objectness: 0.0624 (0.0723)  loss_rpn_box_reg: 0.0583 (0.0607)  time: 1.6900  data: 0.6702  max mem: 19870\n",
      "Epoch: [1]  [ 300/2444]  eta: 1:00:40  lr: 0.020000  loss: 0.7043 (0.7056)  loss_classifier: 0.2467 (0.2491)  loss_box_reg: 0.3250 (0.3228)  loss_objectness: 0.0722 (0.0728)  loss_rpn_box_reg: 0.0600 (0.0609)  time: 1.6933  data: 0.6733  max mem: 19870\n",
      "Epoch: [1]  [ 310/2444]  eta: 1:00:22  lr: 0.020000  loss: 0.7306 (0.7057)  loss_classifier: 0.2437 (0.2488)  loss_box_reg: 0.3250 (0.3229)  loss_objectness: 0.0811 (0.0731)  loss_rpn_box_reg: 0.0620 (0.0610)  time: 1.6992  data: 0.6787  max mem: 19870\n",
      "Epoch: [1]  [ 320/2444]  eta: 1:00:06  lr: 0.020000  loss: 0.7224 (0.7066)  loss_classifier: 0.2385 (0.2489)  loss_box_reg: 0.3264 (0.3233)  loss_objectness: 0.0755 (0.0733)  loss_rpn_box_reg: 0.0599 (0.0610)  time: 1.6959  data: 0.6761  max mem: 19870\n",
      "Epoch: [1]  [ 330/2444]  eta: 0:59:48  lr: 0.020000  loss: 0.7139 (0.7059)  loss_classifier: 0.2443 (0.2487)  loss_box_reg: 0.3252 (0.3230)  loss_objectness: 0.0686 (0.0731)  loss_rpn_box_reg: 0.0605 (0.0611)  time: 1.6973  data: 0.6757  max mem: 19870\n",
      "Epoch: [1]  [ 340/2444]  eta: 0:59:31  lr: 0.020000  loss: 0.6892 (0.7048)  loss_classifier: 0.2436 (0.2486)  loss_box_reg: 0.3147 (0.3228)  loss_objectness: 0.0529 (0.0726)  loss_rpn_box_reg: 0.0530 (0.0608)  time: 1.6902  data: 0.6698  max mem: 19870\n",
      "Epoch: [1]  [ 350/2444]  eta: 0:59:14  lr: 0.020000  loss: 0.6920 (0.7052)  loss_classifier: 0.2456 (0.2488)  loss_box_reg: 0.3257 (0.3231)  loss_objectness: 0.0599 (0.0725)  loss_rpn_box_reg: 0.0537 (0.0608)  time: 1.6917  data: 0.6727  max mem: 19870\n",
      "Epoch: [1]  [ 360/2444]  eta: 0:58:57  lr: 0.020000  loss: 0.7369 (0.7058)  loss_classifier: 0.2512 (0.2489)  loss_box_reg: 0.3316 (0.3234)  loss_objectness: 0.0729 (0.0726)  loss_rpn_box_reg: 0.0606 (0.0608)  time: 1.7011  data: 0.6815  max mem: 19870\n",
      "Epoch: [1]  [ 370/2444]  eta: 0:58:40  lr: 0.020000  loss: 0.7243 (0.7064)  loss_classifier: 0.2512 (0.2490)  loss_box_reg: 0.3310 (0.3234)  loss_objectness: 0.0779 (0.0731)  loss_rpn_box_reg: 0.0629 (0.0609)  time: 1.7012  data: 0.6822  max mem: 19870\n",
      "Epoch: [1]  [ 380/2444]  eta: 0:58:22  lr: 0.020000  loss: 0.6810 (0.7055)  loss_classifier: 0.2417 (0.2488)  loss_box_reg: 0.3092 (0.3230)  loss_objectness: 0.0607 (0.0729)  loss_rpn_box_reg: 0.0566 (0.0609)  time: 1.6893  data: 0.6694  max mem: 19870\n",
      "Epoch: [1]  [ 390/2444]  eta: 0:58:06  lr: 0.020000  loss: 0.6815 (0.7066)  loss_classifier: 0.2456 (0.2489)  loss_box_reg: 0.3287 (0.3234)  loss_objectness: 0.0605 (0.0731)  loss_rpn_box_reg: 0.0653 (0.0612)  time: 1.6975  data: 0.6756  max mem: 19870\n",
      "Epoch: [1]  [ 400/2444]  eta: 0:57:49  lr: 0.020000  loss: 0.7254 (0.7065)  loss_classifier: 0.2521 (0.2489)  loss_box_reg: 0.3318 (0.3233)  loss_objectness: 0.0730 (0.0731)  loss_rpn_box_reg: 0.0663 (0.0612)  time: 1.7021  data: 0.6809  max mem: 19870\n",
      "Epoch: [1]  [ 410/2444]  eta: 0:57:32  lr: 0.020000  loss: 0.6874 (0.7057)  loss_classifier: 0.2412 (0.2485)  loss_box_reg: 0.3200 (0.3231)  loss_objectness: 0.0713 (0.0730)  loss_rpn_box_reg: 0.0609 (0.0611)  time: 1.6923  data: 0.6720  max mem: 19870\n",
      "Epoch: [1]  [ 420/2444]  eta: 0:57:14  lr: 0.020000  loss: 0.6719 (0.7050)  loss_classifier: 0.2319 (0.2484)  loss_box_reg: 0.3200 (0.3228)  loss_objectness: 0.0704 (0.0728)  loss_rpn_box_reg: 0.0562 (0.0610)  time: 1.6912  data: 0.6701  max mem: 19870\n",
      "Epoch: [1]  [ 430/2444]  eta: 0:56:58  lr: 0.020000  loss: 0.6799 (0.7048)  loss_classifier: 0.2370 (0.2482)  loss_box_reg: 0.3115 (0.3226)  loss_objectness: 0.0722 (0.0729)  loss_rpn_box_reg: 0.0562 (0.0611)  time: 1.6981  data: 0.6771  max mem: 19870\n",
      "Epoch: [1]  [ 440/2444]  eta: 0:56:41  lr: 0.020000  loss: 0.6933 (0.7045)  loss_classifier: 0.2379 (0.2481)  loss_box_reg: 0.3158 (0.3224)  loss_objectness: 0.0724 (0.0728)  loss_rpn_box_reg: 0.0638 (0.0612)  time: 1.7018  data: 0.6811  max mem: 19870\n",
      "Epoch: [1]  [ 450/2444]  eta: 0:56:23  lr: 0.020000  loss: 0.6853 (0.7042)  loss_classifier: 0.2414 (0.2481)  loss_box_reg: 0.3085 (0.3222)  loss_objectness: 0.0662 (0.0728)  loss_rpn_box_reg: 0.0627 (0.0611)  time: 1.6923  data: 0.6721  max mem: 19870\n",
      "Epoch: [1]  [ 460/2444]  eta: 0:56:06  lr: 0.020000  loss: 0.7035 (0.7042)  loss_classifier: 0.2458 (0.2483)  loss_box_reg: 0.3132 (0.3222)  loss_objectness: 0.0655 (0.0726)  loss_rpn_box_reg: 0.0586 (0.0611)  time: 1.6916  data: 0.6726  max mem: 19870\n",
      "Epoch: [1]  [ 470/2444]  eta: 0:55:49  lr: 0.020000  loss: 0.7225 (0.7042)  loss_classifier: 0.2494 (0.2482)  loss_box_reg: 0.3215 (0.3222)  loss_objectness: 0.0655 (0.0727)  loss_rpn_box_reg: 0.0586 (0.0611)  time: 1.6949  data: 0.6754  max mem: 19870\n",
      "Epoch: [1]  [ 480/2444]  eta: 0:55:32  lr: 0.020000  loss: 0.7007 (0.7036)  loss_classifier: 0.2371 (0.2480)  loss_box_reg: 0.3100 (0.3219)  loss_objectness: 0.0667 (0.0726)  loss_rpn_box_reg: 0.0593 (0.0611)  time: 1.6943  data: 0.6723  max mem: 19870\n",
      "Epoch: [1]  [ 490/2444]  eta: 0:55:15  lr: 0.020000  loss: 0.6628 (0.7033)  loss_classifier: 0.2371 (0.2479)  loss_box_reg: 0.3020 (0.3216)  loss_objectness: 0.0663 (0.0726)  loss_rpn_box_reg: 0.0593 (0.0612)  time: 1.6877  data: 0.6654  max mem: 19870\n",
      "Epoch: [1]  [ 500/2444]  eta: 0:54:58  lr: 0.020000  loss: 0.6838 (0.7040)  loss_classifier: 0.2433 (0.2481)  loss_box_reg: 0.3149 (0.3218)  loss_objectness: 0.0746 (0.0728)  loss_rpn_box_reg: 0.0670 (0.0614)  time: 1.6955  data: 0.6751  max mem: 19870\n",
      "Epoch: [1]  [ 510/2444]  eta: 0:54:41  lr: 0.020000  loss: 0.7108 (0.7037)  loss_classifier: 0.2433 (0.2479)  loss_box_reg: 0.3216 (0.3216)  loss_objectness: 0.0746 (0.0729)  loss_rpn_box_reg: 0.0603 (0.0613)  time: 1.7027  data: 0.6835  max mem: 19870\n",
      "Epoch: [1]  [ 520/2444]  eta: 0:54:24  lr: 0.020000  loss: 0.6311 (0.7021)  loss_classifier: 0.2236 (0.2473)  loss_box_reg: 0.2909 (0.3209)  loss_objectness: 0.0584 (0.0727)  loss_rpn_box_reg: 0.0549 (0.0612)  time: 1.6893  data: 0.6697  max mem: 19870\n",
      "Epoch: [1]  [ 530/2444]  eta: 0:54:06  lr: 0.020000  loss: 0.6311 (0.7013)  loss_classifier: 0.2236 (0.2471)  loss_box_reg: 0.2826 (0.3205)  loss_objectness: 0.0557 (0.0725)  loss_rpn_box_reg: 0.0544 (0.0612)  time: 1.6836  data: 0.6635  max mem: 19870\n",
      "Epoch: [1]  [ 540/2444]  eta: 0:53:50  lr: 0.020000  loss: 0.6596 (0.7009)  loss_classifier: 0.2414 (0.2471)  loss_box_reg: 0.3066 (0.3204)  loss_objectness: 0.0611 (0.0724)  loss_rpn_box_reg: 0.0544 (0.0611)  time: 1.6953  data: 0.6741  max mem: 19870\n",
      "Epoch: [1]  [ 550/2444]  eta: 0:53:33  lr: 0.020000  loss: 0.6837 (0.7013)  loss_classifier: 0.2469 (0.2471)  loss_box_reg: 0.3163 (0.3205)  loss_objectness: 0.0682 (0.0726)  loss_rpn_box_reg: 0.0601 (0.0611)  time: 1.7041  data: 0.6818  max mem: 19870\n",
      "Epoch: [1]  [ 560/2444]  eta: 0:53:16  lr: 0.020000  loss: 0.6894 (0.7010)  loss_classifier: 0.2469 (0.2470)  loss_box_reg: 0.3163 (0.3203)  loss_objectness: 0.0663 (0.0725)  loss_rpn_box_reg: 0.0624 (0.0612)  time: 1.7029  data: 0.6811  max mem: 19870\n",
      "Epoch: [1]  [ 570/2444]  eta: 0:53:00  lr: 0.020000  loss: 0.6732 (0.7008)  loss_classifier: 0.2414 (0.2469)  loss_box_reg: 0.3098 (0.3201)  loss_objectness: 0.0607 (0.0726)  loss_rpn_box_reg: 0.0573 (0.0612)  time: 1.7053  data: 0.6836  max mem: 19870\n",
      "Epoch: [1]  [ 580/2444]  eta: 0:52:43  lr: 0.020000  loss: 0.6855 (0.7013)  loss_classifier: 0.2458 (0.2471)  loss_box_reg: 0.3164 (0.3203)  loss_objectness: 0.0649 (0.0727)  loss_rpn_box_reg: 0.0593 (0.0613)  time: 1.7032  data: 0.6829  max mem: 19870\n",
      "Epoch: [1]  [ 590/2444]  eta: 0:52:26  lr: 0.020000  loss: 0.6838 (0.7009)  loss_classifier: 0.2545 (0.2470)  loss_box_reg: 0.3164 (0.3200)  loss_objectness: 0.0631 (0.0727)  loss_rpn_box_reg: 0.0581 (0.0612)  time: 1.6957  data: 0.6766  max mem: 19870\n",
      "Epoch: [1]  [ 600/2444]  eta: 0:52:08  lr: 0.020000  loss: 0.6838 (0.7010)  loss_classifier: 0.2465 (0.2469)  loss_box_reg: 0.3099 (0.3199)  loss_objectness: 0.0727 (0.0730)  loss_rpn_box_reg: 0.0566 (0.0612)  time: 1.6922  data: 0.6721  max mem: 19870\n",
      "Epoch: [1]  [ 610/2444]  eta: 0:51:52  lr: 0.020000  loss: 0.7035 (0.7011)  loss_classifier: 0.2469 (0.2469)  loss_box_reg: 0.3233 (0.3199)  loss_objectness: 0.0757 (0.0730)  loss_rpn_box_reg: 0.0629 (0.0613)  time: 1.6974  data: 0.6783  max mem: 19870\n",
      "Epoch: [1]  [ 620/2444]  eta: 0:51:34  lr: 0.020000  loss: 0.6958 (0.7008)  loss_classifier: 0.2442 (0.2468)  loss_box_reg: 0.3115 (0.3198)  loss_objectness: 0.0649 (0.0730)  loss_rpn_box_reg: 0.0616 (0.0612)  time: 1.6923  data: 0.6736  max mem: 19870\n",
      "Epoch: [1]  [ 630/2444]  eta: 0:51:18  lr: 0.020000  loss: 0.6958 (0.7009)  loss_classifier: 0.2439 (0.2468)  loss_box_reg: 0.3082 (0.3197)  loss_objectness: 0.0650 (0.0731)  loss_rpn_box_reg: 0.0592 (0.0612)  time: 1.7005  data: 0.6745  max mem: 19870\n",
      "Epoch: [1]  [ 640/2444]  eta: 0:51:01  lr: 0.020000  loss: 0.7102 (0.7019)  loss_classifier: 0.2594 (0.2472)  loss_box_reg: 0.3305 (0.3201)  loss_objectness: 0.0711 (0.0733)  loss_rpn_box_reg: 0.0630 (0.0613)  time: 1.7118  data: 0.6854  max mem: 19870\n",
      "Epoch: [1]  [ 650/2444]  eta: 0:50:44  lr: 0.020000  loss: 0.7254 (0.7023)  loss_classifier: 0.2605 (0.2473)  loss_box_reg: 0.3257 (0.3202)  loss_objectness: 0.0711 (0.0733)  loss_rpn_box_reg: 0.0646 (0.0614)  time: 1.7029  data: 0.6829  max mem: 19870\n",
      "Epoch: [1]  [ 660/2444]  eta: 0:50:27  lr: 0.020000  loss: 0.7219 (0.7023)  loss_classifier: 0.2500 (0.2473)  loss_box_reg: 0.3155 (0.3201)  loss_objectness: 0.0712 (0.0734)  loss_rpn_box_reg: 0.0646 (0.0615)  time: 1.6952  data: 0.6759  max mem: 19870\n",
      "Epoch: [1]  [ 670/2444]  eta: 0:50:10  lr: 0.020000  loss: 0.6769 (0.7023)  loss_classifier: 0.2428 (0.2473)  loss_box_reg: 0.3063 (0.3200)  loss_objectness: 0.0708 (0.0734)  loss_rpn_box_reg: 0.0598 (0.0616)  time: 1.6911  data: 0.6719  max mem: 19870\n",
      "Epoch: [1]  [ 680/2444]  eta: 0:49:53  lr: 0.020000  loss: 0.6744 (0.7025)  loss_classifier: 0.2417 (0.2474)  loss_box_reg: 0.3030 (0.3200)  loss_objectness: 0.0708 (0.0735)  loss_rpn_box_reg: 0.0587 (0.0616)  time: 1.6917  data: 0.6727  max mem: 19870\n",
      "Epoch: [1]  [ 690/2444]  eta: 0:49:36  lr: 0.020000  loss: 0.7025 (0.7028)  loss_classifier: 0.2533 (0.2476)  loss_box_reg: 0.3143 (0.3200)  loss_objectness: 0.0713 (0.0736)  loss_rpn_box_reg: 0.0587 (0.0616)  time: 1.6943  data: 0.6758  max mem: 19870\n",
      "Epoch: [1]  [ 700/2444]  eta: 0:49:19  lr: 0.020000  loss: 0.6905 (0.7022)  loss_classifier: 0.2464 (0.2474)  loss_box_reg: 0.3139 (0.3198)  loss_objectness: 0.0660 (0.0735)  loss_rpn_box_reg: 0.0564 (0.0615)  time: 1.6936  data: 0.6744  max mem: 19870\n",
      "Epoch: [1]  [ 710/2444]  eta: 0:49:02  lr: 0.020000  loss: 0.6839 (0.7023)  loss_classifier: 0.2334 (0.2473)  loss_box_reg: 0.3071 (0.3198)  loss_objectness: 0.0660 (0.0735)  loss_rpn_box_reg: 0.0581 (0.0616)  time: 1.6959  data: 0.6759  max mem: 19870\n",
      "Epoch: [1]  [ 720/2444]  eta: 0:48:45  lr: 0.020000  loss: 0.7051 (0.7021)  loss_classifier: 0.2485 (0.2473)  loss_box_reg: 0.3188 (0.3197)  loss_objectness: 0.0709 (0.0736)  loss_rpn_box_reg: 0.0611 (0.0615)  time: 1.6993  data: 0.6793  max mem: 19870\n",
      "Epoch: [1]  [ 730/2444]  eta: 0:48:28  lr: 0.020000  loss: 0.7029 (0.7017)  loss_classifier: 0.2494 (0.2471)  loss_box_reg: 0.3105 (0.3194)  loss_objectness: 0.0816 (0.0736)  loss_rpn_box_reg: 0.0587 (0.0615)  time: 1.6898  data: 0.6703  max mem: 19870\n",
      "Epoch: [1]  [ 740/2444]  eta: 0:48:11  lr: 0.020000  loss: 0.7029 (0.7019)  loss_classifier: 0.2407 (0.2471)  loss_box_reg: 0.3117 (0.3194)  loss_objectness: 0.0793 (0.0739)  loss_rpn_box_reg: 0.0617 (0.0615)  time: 1.6868  data: 0.6683  max mem: 19870\n",
      "Epoch: [1]  [ 750/2444]  eta: 0:47:53  lr: 0.020000  loss: 0.7017 (0.7013)  loss_classifier: 0.2380 (0.2470)  loss_box_reg: 0.3015 (0.3191)  loss_objectness: 0.0720 (0.0738)  loss_rpn_box_reg: 0.0544 (0.0614)  time: 1.6908  data: 0.6720  max mem: 19870\n",
      "Epoch: [1]  [ 760/2444]  eta: 0:47:37  lr: 0.020000  loss: 0.6678 (0.7015)  loss_classifier: 0.2371 (0.2470)  loss_box_reg: 0.3046 (0.3192)  loss_objectness: 0.0713 (0.0739)  loss_rpn_box_reg: 0.0534 (0.0614)  time: 1.6966  data: 0.6765  max mem: 19870\n",
      "Epoch: [1]  [ 770/2444]  eta: 0:47:20  lr: 0.020000  loss: 0.6678 (0.7010)  loss_classifier: 0.2391 (0.2469)  loss_box_reg: 0.3139 (0.3191)  loss_objectness: 0.0674 (0.0737)  loss_rpn_box_reg: 0.0598 (0.0614)  time: 1.6999  data: 0.6780  max mem: 19870\n",
      "Epoch: [1]  [ 780/2444]  eta: 0:47:03  lr: 0.020000  loss: 0.6495 (0.7006)  loss_classifier: 0.2323 (0.2468)  loss_box_reg: 0.3073 (0.3189)  loss_objectness: 0.0553 (0.0735)  loss_rpn_box_reg: 0.0602 (0.0613)  time: 1.6960  data: 0.6737  max mem: 19870\n",
      "Epoch: [1]  [ 790/2444]  eta: 0:46:46  lr: 0.020000  loss: 0.6782 (0.7003)  loss_classifier: 0.2323 (0.2467)  loss_box_reg: 0.3145 (0.3188)  loss_objectness: 0.0626 (0.0734)  loss_rpn_box_reg: 0.0587 (0.0613)  time: 1.6912  data: 0.6703  max mem: 19870\n",
      "Epoch: [1]  [ 800/2444]  eta: 0:46:29  lr: 0.020000  loss: 0.6889 (0.7001)  loss_classifier: 0.2319 (0.2466)  loss_box_reg: 0.3145 (0.3186)  loss_objectness: 0.0741 (0.0736)  loss_rpn_box_reg: 0.0587 (0.0613)  time: 1.6915  data: 0.6717  max mem: 19870\n",
      "Epoch: [1]  [ 810/2444]  eta: 0:46:12  lr: 0.020000  loss: 0.6751 (0.7000)  loss_classifier: 0.2319 (0.2465)  loss_box_reg: 0.3035 (0.3185)  loss_objectness: 0.0770 (0.0737)  loss_rpn_box_reg: 0.0601 (0.0614)  time: 1.6949  data: 0.6751  max mem: 19870\n",
      "Epoch: [1]  [ 820/2444]  eta: 0:45:55  lr: 0.020000  loss: 0.7048 (0.7004)  loss_classifier: 0.2428 (0.2466)  loss_box_reg: 0.3195 (0.3187)  loss_objectness: 0.0661 (0.0736)  loss_rpn_box_reg: 0.0689 (0.0615)  time: 1.6998  data: 0.6796  max mem: 19870\n",
      "Epoch: [1]  [ 830/2444]  eta: 0:45:38  lr: 0.020000  loss: 0.7063 (0.7001)  loss_classifier: 0.2406 (0.2465)  loss_box_reg: 0.3156 (0.3185)  loss_objectness: 0.0684 (0.0736)  loss_rpn_box_reg: 0.0651 (0.0615)  time: 1.7061  data: 0.6812  max mem: 19870\n",
      "Epoch: [1]  [ 840/2444]  eta: 0:45:21  lr: 0.020000  loss: 0.6916 (0.6997)  loss_classifier: 0.2280 (0.2464)  loss_box_reg: 0.3037 (0.3184)  loss_objectness: 0.0642 (0.0736)  loss_rpn_box_reg: 0.0603 (0.0615)  time: 1.7055  data: 0.6804  max mem: 19870\n",
      "Epoch: [1]  [ 850/2444]  eta: 0:45:04  lr: 0.020000  loss: 0.6926 (0.7001)  loss_classifier: 0.2456 (0.2464)  loss_box_reg: 0.3125 (0.3184)  loss_objectness: 0.0739 (0.0738)  loss_rpn_box_reg: 0.0601 (0.0615)  time: 1.6952  data: 0.6756  max mem: 19870\n",
      "Epoch: [1]  [ 860/2444]  eta: 0:44:47  lr: 0.020000  loss: 0.6842 (0.6996)  loss_classifier: 0.2414 (0.2463)  loss_box_reg: 0.3010 (0.3181)  loss_objectness: 0.0739 (0.0737)  loss_rpn_box_reg: 0.0601 (0.0615)  time: 1.6862  data: 0.6667  max mem: 19870\n",
      "Epoch: [1]  [ 870/2444]  eta: 0:44:30  lr: 0.020000  loss: 0.6939 (0.6997)  loss_classifier: 0.2386 (0.2463)  loss_box_reg: 0.3083 (0.3181)  loss_objectness: 0.0738 (0.0738)  loss_rpn_box_reg: 0.0632 (0.0615)  time: 1.6912  data: 0.6713  max mem: 19870\n",
      "Epoch: [1]  [ 880/2444]  eta: 0:44:13  lr: 0.020000  loss: 0.6901 (0.6995)  loss_classifier: 0.2386 (0.2462)  loss_box_reg: 0.3083 (0.3180)  loss_objectness: 0.0722 (0.0738)  loss_rpn_box_reg: 0.0632 (0.0616)  time: 1.6964  data: 0.6758  max mem: 19870\n",
      "Epoch: [1]  [ 890/2444]  eta: 0:43:56  lr: 0.020000  loss: 0.6837 (0.6995)  loss_classifier: 0.2410 (0.2462)  loss_box_reg: 0.3096 (0.3179)  loss_objectness: 0.0600 (0.0737)  loss_rpn_box_reg: 0.0632 (0.0616)  time: 1.6958  data: 0.6758  max mem: 19870\n",
      "Epoch: [1]  [ 900/2444]  eta: 0:43:39  lr: 0.020000  loss: 0.6906 (0.6995)  loss_classifier: 0.2500 (0.2462)  loss_box_reg: 0.3154 (0.3179)  loss_objectness: 0.0640 (0.0737)  loss_rpn_box_reg: 0.0634 (0.0616)  time: 1.6954  data: 0.6758  max mem: 19870\n",
      "Epoch: [1]  [ 910/2444]  eta: 0:43:22  lr: 0.020000  loss: 0.6872 (0.6994)  loss_classifier: 0.2477 (0.2462)  loss_box_reg: 0.3040 (0.3178)  loss_objectness: 0.0651 (0.0737)  loss_rpn_box_reg: 0.0671 (0.0616)  time: 1.6903  data: 0.6708  max mem: 19870\n",
      "Epoch: [1]  [ 920/2444]  eta: 0:43:05  lr: 0.020000  loss: 0.6524 (0.6991)  loss_classifier: 0.2314 (0.2461)  loss_box_reg: 0.2962 (0.3177)  loss_objectness: 0.0701 (0.0737)  loss_rpn_box_reg: 0.0582 (0.0615)  time: 1.6874  data: 0.6694  max mem: 19870\n",
      "Epoch: [1]  [ 930/2444]  eta: 0:42:48  lr: 0.020000  loss: 0.6644 (0.6992)  loss_classifier: 0.2399 (0.2462)  loss_box_reg: 0.3106 (0.3177)  loss_objectness: 0.0740 (0.0738)  loss_rpn_box_reg: 0.0573 (0.0615)  time: 1.6991  data: 0.6796  max mem: 19870\n",
      "Epoch: [1]  [ 940/2444]  eta: 0:42:31  lr: 0.020000  loss: 0.6952 (0.6993)  loss_classifier: 0.2501 (0.2462)  loss_box_reg: 0.3232 (0.3177)  loss_objectness: 0.0767 (0.0738)  loss_rpn_box_reg: 0.0623 (0.0615)  time: 1.7047  data: 0.6841  max mem: 19870\n",
      "Epoch: [1]  [ 950/2444]  eta: 0:42:14  lr: 0.020000  loss: 0.6868 (0.6989)  loss_classifier: 0.2476 (0.2461)  loss_box_reg: 0.3139 (0.3175)  loss_objectness: 0.0708 (0.0737)  loss_rpn_box_reg: 0.0627 (0.0615)  time: 1.6934  data: 0.6741  max mem: 19870\n",
      "Epoch: [1]  [ 960/2444]  eta: 0:41:57  lr: 0.020000  loss: 0.6893 (0.6989)  loss_classifier: 0.2482 (0.2461)  loss_box_reg: 0.3067 (0.3175)  loss_objectness: 0.0708 (0.0737)  loss_rpn_box_reg: 0.0615 (0.0616)  time: 1.6974  data: 0.6777  max mem: 19870\n",
      "Epoch: [1]  [ 970/2444]  eta: 0:41:40  lr: 0.020000  loss: 0.6901 (0.6987)  loss_classifier: 0.2409 (0.2459)  loss_box_reg: 0.3153 (0.3174)  loss_objectness: 0.0739 (0.0738)  loss_rpn_box_reg: 0.0606 (0.0616)  time: 1.6985  data: 0.6771  max mem: 19870\n",
      "Epoch: [1]  [ 980/2444]  eta: 0:41:23  lr: 0.020000  loss: 0.6985 (0.6989)  loss_classifier: 0.2338 (0.2460)  loss_box_reg: 0.3209 (0.3175)  loss_objectness: 0.0659 (0.0738)  loss_rpn_box_reg: 0.0607 (0.0616)  time: 1.6939  data: 0.6720  max mem: 19870\n",
      "Epoch: [1]  [ 990/2444]  eta: 0:41:06  lr: 0.020000  loss: 0.6985 (0.6987)  loss_classifier: 0.2367 (0.2459)  loss_box_reg: 0.3127 (0.3173)  loss_objectness: 0.0681 (0.0739)  loss_rpn_box_reg: 0.0607 (0.0616)  time: 1.6951  data: 0.6739  max mem: 19870\n",
      "Epoch: [1]  [1000/2444]  eta: 0:40:49  lr: 0.020000  loss: 0.6902 (0.6987)  loss_classifier: 0.2344 (0.2459)  loss_box_reg: 0.3029 (0.3172)  loss_objectness: 0.0742 (0.0740)  loss_rpn_box_reg: 0.0660 (0.0616)  time: 1.6979  data: 0.6773  max mem: 19870\n",
      "Epoch: [1]  [1010/2444]  eta: 0:40:32  lr: 0.020000  loss: 0.6962 (0.6987)  loss_classifier: 0.2393 (0.2458)  loss_box_reg: 0.3070 (0.3173)  loss_objectness: 0.0679 (0.0740)  loss_rpn_box_reg: 0.0660 (0.0616)  time: 1.7017  data: 0.6820  max mem: 19870\n",
      "Epoch: [1]  [1020/2444]  eta: 0:40:15  lr: 0.020000  loss: 0.6962 (0.6989)  loss_classifier: 0.2438 (0.2459)  loss_box_reg: 0.3133 (0.3173)  loss_objectness: 0.0726 (0.0741)  loss_rpn_box_reg: 0.0595 (0.0617)  time: 1.6964  data: 0.6775  max mem: 19870\n",
      "Epoch: [1]  [1030/2444]  eta: 0:39:59  lr: 0.020000  loss: 0.6787 (0.6987)  loss_classifier: 0.2391 (0.2458)  loss_box_reg: 0.3086 (0.3172)  loss_objectness: 0.0649 (0.0740)  loss_rpn_box_reg: 0.0624 (0.0617)  time: 1.7024  data: 0.6767  max mem: 19870\n",
      "Epoch: [1]  [1040/2444]  eta: 0:39:42  lr: 0.020000  loss: 0.6857 (0.6986)  loss_classifier: 0.2358 (0.2458)  loss_box_reg: 0.3132 (0.3171)  loss_objectness: 0.0616 (0.0740)  loss_rpn_box_reg: 0.0618 (0.0617)  time: 1.7067  data: 0.6818  max mem: 19870\n",
      "Epoch: [1]  [1050/2444]  eta: 0:39:25  lr: 0.020000  loss: 0.6857 (0.6985)  loss_classifier: 0.2346 (0.2457)  loss_box_reg: 0.3041 (0.3170)  loss_objectness: 0.0717 (0.0741)  loss_rpn_box_reg: 0.0600 (0.0617)  time: 1.6965  data: 0.6787  max mem: 19870\n",
      "Epoch: [1]  [1060/2444]  eta: 0:39:08  lr: 0.020000  loss: 0.6906 (0.6983)  loss_classifier: 0.2346 (0.2457)  loss_box_reg: 0.3041 (0.3169)  loss_objectness: 0.0777 (0.0741)  loss_rpn_box_reg: 0.0555 (0.0616)  time: 1.6915  data: 0.6738  max mem: 19870\n",
      "Epoch: [1]  [1070/2444]  eta: 0:38:51  lr: 0.020000  loss: 0.7007 (0.6983)  loss_classifier: 0.2405 (0.2456)  loss_box_reg: 0.3089 (0.3169)  loss_objectness: 0.0767 (0.0741)  loss_rpn_box_reg: 0.0587 (0.0617)  time: 1.6904  data: 0.6734  max mem: 19870\n",
      "Epoch: [1]  [1080/2444]  eta: 0:38:33  lr: 0.020000  loss: 0.6765 (0.6982)  loss_classifier: 0.2345 (0.2456)  loss_box_reg: 0.3012 (0.3168)  loss_objectness: 0.0755 (0.0741)  loss_rpn_box_reg: 0.0608 (0.0617)  time: 1.6895  data: 0.6722  max mem: 19870\n",
      "Epoch: [1]  [1090/2444]  eta: 0:38:16  lr: 0.020000  loss: 0.6726 (0.6981)  loss_classifier: 0.2345 (0.2455)  loss_box_reg: 0.2980 (0.3167)  loss_objectness: 0.0755 (0.0742)  loss_rpn_box_reg: 0.0625 (0.0617)  time: 1.6890  data: 0.6699  max mem: 19870\n",
      "Epoch: [1]  [1100/2444]  eta: 0:37:59  lr: 0.020000  loss: 0.6726 (0.6981)  loss_classifier: 0.2349 (0.2455)  loss_box_reg: 0.3090 (0.3167)  loss_objectness: 0.0718 (0.0742)  loss_rpn_box_reg: 0.0615 (0.0617)  time: 1.6881  data: 0.6683  max mem: 19870\n",
      "Epoch: [1]  [1110/2444]  eta: 0:37:42  lr: 0.020000  loss: 0.7148 (0.6983)  loss_classifier: 0.2480 (0.2456)  loss_box_reg: 0.3225 (0.3168)  loss_objectness: 0.0697 (0.0742)  loss_rpn_box_reg: 0.0619 (0.0618)  time: 1.6976  data: 0.6786  max mem: 19870\n",
      "Epoch: [1]  [1120/2444]  eta: 0:37:26  lr: 0.020000  loss: 0.7148 (0.6984)  loss_classifier: 0.2480 (0.2457)  loss_box_reg: 0.3225 (0.3168)  loss_objectness: 0.0668 (0.0741)  loss_rpn_box_reg: 0.0639 (0.0618)  time: 1.7059  data: 0.6866  max mem: 19870\n",
      "Epoch: [1]  [1130/2444]  eta: 0:37:09  lr: 0.020000  loss: 0.6691 (0.6979)  loss_classifier: 0.2356 (0.2455)  loss_box_reg: 0.3077 (0.3166)  loss_objectness: 0.0630 (0.0740)  loss_rpn_box_reg: 0.0601 (0.0617)  time: 1.6969  data: 0.6773  max mem: 19870\n",
      "Epoch: [1]  [1140/2444]  eta: 0:36:52  lr: 0.020000  loss: 0.6362 (0.6978)  loss_classifier: 0.2235 (0.2454)  loss_box_reg: 0.3059 (0.3165)  loss_objectness: 0.0713 (0.0741)  loss_rpn_box_reg: 0.0601 (0.0617)  time: 1.6977  data: 0.6783  max mem: 19870\n",
      "Epoch: [1]  [1150/2444]  eta: 0:36:35  lr: 0.020000  loss: 0.6991 (0.6977)  loss_classifier: 0.2452 (0.2454)  loss_box_reg: 0.3088 (0.3165)  loss_objectness: 0.0705 (0.0741)  loss_rpn_box_reg: 0.0606 (0.0617)  time: 1.7039  data: 0.6834  max mem: 19870\n",
      "Epoch: [1]  [1160/2444]  eta: 0:36:18  lr: 0.020000  loss: 0.6962 (0.6978)  loss_classifier: 0.2474 (0.2454)  loss_box_reg: 0.3101 (0.3164)  loss_objectness: 0.0758 (0.0742)  loss_rpn_box_reg: 0.0568 (0.0617)  time: 1.6954  data: 0.6744  max mem: 19870\n",
      "Epoch: [1]  [1170/2444]  eta: 0:36:01  lr: 0.020000  loss: 0.7121 (0.6979)  loss_classifier: 0.2452 (0.2454)  loss_box_reg: 0.3209 (0.3164)  loss_objectness: 0.0800 (0.0743)  loss_rpn_box_reg: 0.0623 (0.0618)  time: 1.7000  data: 0.6782  max mem: 19870\n",
      "Epoch: [1]  [1180/2444]  eta: 0:35:44  lr: 0.020000  loss: 0.7121 (0.6978)  loss_classifier: 0.2452 (0.2454)  loss_box_reg: 0.3151 (0.3164)  loss_objectness: 0.0701 (0.0742)  loss_rpn_box_reg: 0.0623 (0.0618)  time: 1.7060  data: 0.6845  max mem: 19870\n",
      "Epoch: [1]  [1190/2444]  eta: 0:35:27  lr: 0.020000  loss: 0.6817 (0.6979)  loss_classifier: 0.2534 (0.2454)  loss_box_reg: 0.3140 (0.3164)  loss_objectness: 0.0663 (0.0743)  loss_rpn_box_reg: 0.0578 (0.0618)  time: 1.6962  data: 0.6759  max mem: 19870\n",
      "Epoch: [1]  [1200/2444]  eta: 0:35:10  lr: 0.020000  loss: 0.6817 (0.6977)  loss_classifier: 0.2365 (0.2454)  loss_box_reg: 0.3147 (0.3163)  loss_objectness: 0.0754 (0.0743)  loss_rpn_box_reg: 0.0584 (0.0618)  time: 1.6897  data: 0.6689  max mem: 19870\n",
      "Epoch: [1]  [1210/2444]  eta: 0:34:53  lr: 0.020000  loss: 0.6796 (0.6977)  loss_classifier: 0.2418 (0.2453)  loss_box_reg: 0.3147 (0.3163)  loss_objectness: 0.0777 (0.0744)  loss_rpn_box_reg: 0.0537 (0.0618)  time: 1.6889  data: 0.6679  max mem: 19870\n",
      "Epoch: [1]  [1220/2444]  eta: 0:34:36  lr: 0.020000  loss: 0.6781 (0.6975)  loss_classifier: 0.2393 (0.2453)  loss_box_reg: 0.3035 (0.3161)  loss_objectness: 0.0736 (0.0744)  loss_rpn_box_reg: 0.0559 (0.0618)  time: 1.6899  data: 0.6699  max mem: 19870\n",
      "Epoch: [1]  [1230/2444]  eta: 0:34:19  lr: 0.020000  loss: 0.6538 (0.6973)  loss_classifier: 0.2323 (0.2452)  loss_box_reg: 0.2959 (0.3160)  loss_objectness: 0.0681 (0.0744)  loss_rpn_box_reg: 0.0619 (0.0617)  time: 1.6979  data: 0.6787  max mem: 19870\n",
      "Epoch: [1]  [1240/2444]  eta: 0:34:02  lr: 0.020000  loss: 0.6491 (0.6971)  loss_classifier: 0.2331 (0.2451)  loss_box_reg: 0.2934 (0.3159)  loss_objectness: 0.0825 (0.0745)  loss_rpn_box_reg: 0.0603 (0.0618)  time: 1.6942  data: 0.6755  max mem: 19870\n",
      "Epoch: [1]  [1250/2444]  eta: 0:33:45  lr: 0.020000  loss: 0.6827 (0.6971)  loss_classifier: 0.2385 (0.2451)  loss_box_reg: 0.2961 (0.3158)  loss_objectness: 0.0815 (0.0745)  loss_rpn_box_reg: 0.0599 (0.0617)  time: 1.6886  data: 0.6704  max mem: 19870\n",
      "Epoch: [1]  [1260/2444]  eta: 0:33:28  lr: 0.020000  loss: 0.6599 (0.6967)  loss_classifier: 0.2323 (0.2449)  loss_box_reg: 0.2986 (0.3156)  loss_objectness: 0.0762 (0.0745)  loss_rpn_box_reg: 0.0592 (0.0617)  time: 1.6881  data: 0.6692  max mem: 19870\n",
      "Epoch: [1]  [1270/2444]  eta: 0:33:11  lr: 0.020000  loss: 0.6465 (0.6964)  loss_classifier: 0.2320 (0.2448)  loss_box_reg: 0.3038 (0.3155)  loss_objectness: 0.0643 (0.0744)  loss_rpn_box_reg: 0.0592 (0.0617)  time: 1.6871  data: 0.6661  max mem: 19870\n",
      "Epoch: [1]  [1280/2444]  eta: 0:32:54  lr: 0.020000  loss: 0.6786 (0.6963)  loss_classifier: 0.2375 (0.2447)  loss_box_reg: 0.3070 (0.3154)  loss_objectness: 0.0631 (0.0744)  loss_rpn_box_reg: 0.0561 (0.0617)  time: 1.6879  data: 0.6663  max mem: 19870\n",
      "Epoch: [1]  [1290/2444]  eta: 0:32:37  lr: 0.020000  loss: 0.6767 (0.6959)  loss_classifier: 0.2309 (0.2446)  loss_box_reg: 0.2982 (0.3152)  loss_objectness: 0.0663 (0.0744)  loss_rpn_box_reg: 0.0561 (0.0617)  time: 1.6849  data: 0.6660  max mem: 19870\n",
      "Epoch: [1]  [1300/2444]  eta: 0:32:20  lr: 0.020000  loss: 0.6767 (0.6959)  loss_classifier: 0.2353 (0.2446)  loss_box_reg: 0.3052 (0.3152)  loss_objectness: 0.0670 (0.0744)  loss_rpn_box_reg: 0.0600 (0.0617)  time: 1.6936  data: 0.6747  max mem: 19870\n",
      "Epoch: [1]  [1310/2444]  eta: 0:32:03  lr: 0.020000  loss: 0.6819 (0.6957)  loss_classifier: 0.2353 (0.2445)  loss_box_reg: 0.3137 (0.3151)  loss_objectness: 0.0679 (0.0744)  loss_rpn_box_reg: 0.0591 (0.0616)  time: 1.6976  data: 0.6781  max mem: 19870\n",
      "Epoch: [1]  [1320/2444]  eta: 0:31:46  lr: 0.020000  loss: 0.6549 (0.6956)  loss_classifier: 0.2286 (0.2444)  loss_box_reg: 0.2961 (0.3151)  loss_objectness: 0.0679 (0.0744)  loss_rpn_box_reg: 0.0510 (0.0616)  time: 1.6910  data: 0.6710  max mem: 19870\n",
      "Epoch: [1]  [1330/2444]  eta: 0:31:29  lr: 0.020000  loss: 0.6797 (0.6954)  loss_classifier: 0.2299 (0.2444)  loss_box_reg: 0.3201 (0.3150)  loss_objectness: 0.0657 (0.0744)  loss_rpn_box_reg: 0.0544 (0.0616)  time: 1.6906  data: 0.6699  max mem: 19870\n",
      "Epoch: [1]  [1340/2444]  eta: 0:31:12  lr: 0.020000  loss: 0.6729 (0.6953)  loss_classifier: 0.2416 (0.2444)  loss_box_reg: 0.3134 (0.3150)  loss_objectness: 0.0632 (0.0744)  loss_rpn_box_reg: 0.0557 (0.0616)  time: 1.6977  data: 0.6778  max mem: 19870\n",
      "Epoch: [1]  [1350/2444]  eta: 0:30:55  lr: 0.020000  loss: 0.6729 (0.6953)  loss_classifier: 0.2434 (0.2444)  loss_box_reg: 0.3035 (0.3149)  loss_objectness: 0.0632 (0.0745)  loss_rpn_box_reg: 0.0592 (0.0616)  time: 1.6975  data: 0.6783  max mem: 19870\n",
      "Epoch: [1]  [1360/2444]  eta: 0:30:38  lr: 0.020000  loss: 0.6822 (0.6952)  loss_classifier: 0.2404 (0.2443)  loss_box_reg: 0.2994 (0.3148)  loss_objectness: 0.0654 (0.0745)  loss_rpn_box_reg: 0.0617 (0.0616)  time: 1.6908  data: 0.6708  max mem: 19870\n",
      "Epoch: [1]  [1370/2444]  eta: 0:30:21  lr: 0.020000  loss: 0.6822 (0.6951)  loss_classifier: 0.2376 (0.2443)  loss_box_reg: 0.2996 (0.3148)  loss_objectness: 0.0679 (0.0745)  loss_rpn_box_reg: 0.0605 (0.0616)  time: 1.6868  data: 0.6666  max mem: 19870\n",
      "Epoch: [1]  [1380/2444]  eta: 0:30:04  lr: 0.020000  loss: 0.6720 (0.6949)  loss_classifier: 0.2370 (0.2442)  loss_box_reg: 0.2957 (0.3147)  loss_objectness: 0.0634 (0.0744)  loss_rpn_box_reg: 0.0601 (0.0616)  time: 1.6882  data: 0.6695  max mem: 19870\n",
      "Epoch: [1]  [1390/2444]  eta: 0:29:47  lr: 0.020000  loss: 0.6668 (0.6949)  loss_classifier: 0.2350 (0.2442)  loss_box_reg: 0.2957 (0.3146)  loss_objectness: 0.0590 (0.0745)  loss_rpn_box_reg: 0.0644 (0.0616)  time: 1.6954  data: 0.6761  max mem: 19870\n",
      "Epoch: [1]  [1400/2444]  eta: 0:29:30  lr: 0.020000  loss: 0.6796 (0.6950)  loss_classifier: 0.2413 (0.2442)  loss_box_reg: 0.3152 (0.3147)  loss_objectness: 0.0668 (0.0745)  loss_rpn_box_reg: 0.0645 (0.0617)  time: 1.6967  data: 0.6773  max mem: 19870\n",
      "Epoch: [1]  [1410/2444]  eta: 0:29:13  lr: 0.020000  loss: 0.6706 (0.6947)  loss_classifier: 0.2347 (0.2441)  loss_box_reg: 0.3101 (0.3146)  loss_objectness: 0.0633 (0.0744)  loss_rpn_box_reg: 0.0587 (0.0616)  time: 1.6898  data: 0.6706  max mem: 19870\n",
      "Epoch: [1]  [1420/2444]  eta: 0:28:56  lr: 0.020000  loss: 0.6705 (0.6947)  loss_classifier: 0.2357 (0.2441)  loss_box_reg: 0.3077 (0.3146)  loss_objectness: 0.0612 (0.0744)  loss_rpn_box_reg: 0.0583 (0.0616)  time: 1.6872  data: 0.6682  max mem: 19870\n",
      "Epoch: [1]  [1430/2444]  eta: 0:28:39  lr: 0.020000  loss: 0.6690 (0.6944)  loss_classifier: 0.2357 (0.2440)  loss_box_reg: 0.2998 (0.3145)  loss_objectness: 0.0654 (0.0743)  loss_rpn_box_reg: 0.0583 (0.0616)  time: 1.6979  data: 0.6725  max mem: 19870\n",
      "Epoch: [1]  [1440/2444]  eta: 0:28:22  lr: 0.020000  loss: 0.6415 (0.6945)  loss_classifier: 0.2254 (0.2440)  loss_box_reg: 0.2856 (0.3144)  loss_objectness: 0.0694 (0.0745)  loss_rpn_box_reg: 0.0572 (0.0616)  time: 1.6990  data: 0.6734  max mem: 19870\n",
      "Epoch: [1]  [1450/2444]  eta: 0:28:05  lr: 0.020000  loss: 0.6349 (0.6941)  loss_classifier: 0.2234 (0.2439)  loss_box_reg: 0.2856 (0.3143)  loss_objectness: 0.0735 (0.0744)  loss_rpn_box_reg: 0.0572 (0.0616)  time: 1.6909  data: 0.6714  max mem: 19870\n",
      "Epoch: [1]  [1460/2444]  eta: 0:27:48  lr: 0.020000  loss: 0.6241 (0.6939)  loss_classifier: 0.2234 (0.2437)  loss_box_reg: 0.3004 (0.3142)  loss_objectness: 0.0680 (0.0744)  loss_rpn_box_reg: 0.0598 (0.0615)  time: 1.6899  data: 0.6702  max mem: 19870\n",
      "Epoch: [1]  [1470/2444]  eta: 0:27:31  lr: 0.020000  loss: 0.6890 (0.6939)  loss_classifier: 0.2315 (0.2438)  loss_box_reg: 0.3015 (0.3141)  loss_objectness: 0.0773 (0.0745)  loss_rpn_box_reg: 0.0638 (0.0616)  time: 1.6991  data: 0.6795  max mem: 19870\n",
      "Epoch: [1]  [1480/2444]  eta: 0:27:14  lr: 0.020000  loss: 0.7050 (0.6937)  loss_classifier: 0.2471 (0.2437)  loss_box_reg: 0.3111 (0.3141)  loss_objectness: 0.0689 (0.0744)  loss_rpn_box_reg: 0.0638 (0.0615)  time: 1.7022  data: 0.6810  max mem: 19870\n",
      "Epoch: [1]  [1490/2444]  eta: 0:26:57  lr: 0.020000  loss: 0.6522 (0.6934)  loss_classifier: 0.2366 (0.2436)  loss_box_reg: 0.2992 (0.3139)  loss_objectness: 0.0568 (0.0743)  loss_rpn_box_reg: 0.0596 (0.0615)  time: 1.6903  data: 0.6687  max mem: 19870\n",
      "Epoch: [1]  [1500/2444]  eta: 0:26:40  lr: 0.020000  loss: 0.6378 (0.6931)  loss_classifier: 0.2326 (0.2436)  loss_box_reg: 0.2986 (0.3139)  loss_objectness: 0.0568 (0.0742)  loss_rpn_box_reg: 0.0538 (0.0615)  time: 1.6891  data: 0.6690  max mem: 19870\n",
      "Epoch: [1]  [1510/2444]  eta: 0:26:23  lr: 0.020000  loss: 0.6378 (0.6927)  loss_classifier: 0.2307 (0.2434)  loss_box_reg: 0.2971 (0.3137)  loss_objectness: 0.0652 (0.0741)  loss_rpn_box_reg: 0.0538 (0.0614)  time: 1.6917  data: 0.6720  max mem: 19870\n",
      "Epoch: [1]  [1520/2444]  eta: 0:26:06  lr: 0.020000  loss: 0.6334 (0.6925)  loss_classifier: 0.2253 (0.2433)  loss_box_reg: 0.2913 (0.3136)  loss_objectness: 0.0652 (0.0741)  loss_rpn_box_reg: 0.0575 (0.0614)  time: 1.6905  data: 0.6708  max mem: 19870\n",
      "Epoch: [1]  [1530/2444]  eta: 0:25:49  lr: 0.020000  loss: 0.6461 (0.6924)  loss_classifier: 0.2253 (0.2433)  loss_box_reg: 0.2854 (0.3135)  loss_objectness: 0.0690 (0.0742)  loss_rpn_box_reg: 0.0564 (0.0614)  time: 1.6947  data: 0.6744  max mem: 19870\n",
      "Epoch: [1]  [1540/2444]  eta: 0:25:32  lr: 0.020000  loss: 0.6461 (0.6920)  loss_classifier: 0.2278 (0.2432)  loss_box_reg: 0.2862 (0.3134)  loss_objectness: 0.0609 (0.0741)  loss_rpn_box_reg: 0.0562 (0.0614)  time: 1.6895  data: 0.6682  max mem: 19870\n",
      "Epoch: [1]  [1550/2444]  eta: 0:25:15  lr: 0.020000  loss: 0.6266 (0.6917)  loss_classifier: 0.2299 (0.2431)  loss_box_reg: 0.2946 (0.3133)  loss_objectness: 0.0550 (0.0740)  loss_rpn_box_reg: 0.0530 (0.0614)  time: 1.6821  data: 0.6620  max mem: 19870\n",
      "Epoch: [1]  [1560/2444]  eta: 0:24:58  lr: 0.020000  loss: 0.6587 (0.6917)  loss_classifier: 0.2328 (0.2431)  loss_box_reg: 0.3119 (0.3133)  loss_objectness: 0.0668 (0.0740)  loss_rpn_box_reg: 0.0540 (0.0614)  time: 1.6945  data: 0.6741  max mem: 19870\n",
      "Epoch: [1]  [1570/2444]  eta: 0:24:41  lr: 0.020000  loss: 0.6827 (0.6916)  loss_classifier: 0.2420 (0.2431)  loss_box_reg: 0.3136 (0.3132)  loss_objectness: 0.0680 (0.0739)  loss_rpn_box_reg: 0.0600 (0.0614)  time: 1.7013  data: 0.6809  max mem: 19870\n",
      "Epoch: [1]  [1580/2444]  eta: 0:24:24  lr: 0.020000  loss: 0.6582 (0.6914)  loss_classifier: 0.2376 (0.2430)  loss_box_reg: 0.2978 (0.3131)  loss_objectness: 0.0680 (0.0739)  loss_rpn_box_reg: 0.0589 (0.0614)  time: 1.6898  data: 0.6704  max mem: 19870\n",
      "Epoch: [1]  [1590/2444]  eta: 0:24:08  lr: 0.020000  loss: 0.6690 (0.6915)  loss_classifier: 0.2329 (0.2431)  loss_box_reg: 0.2956 (0.3131)  loss_objectness: 0.0697 (0.0739)  loss_rpn_box_reg: 0.0581 (0.0614)  time: 1.6964  data: 0.6762  max mem: 19870\n",
      "Epoch: [1]  [1600/2444]  eta: 0:23:50  lr: 0.020000  loss: 0.6751 (0.6914)  loss_classifier: 0.2366 (0.2430)  loss_box_reg: 0.3108 (0.3131)  loss_objectness: 0.0697 (0.0740)  loss_rpn_box_reg: 0.0629 (0.0614)  time: 1.6960  data: 0.6766  max mem: 19870\n",
      "Epoch: [1]  [1610/2444]  eta: 0:23:33  lr: 0.020000  loss: 0.6720 (0.6914)  loss_classifier: 0.2366 (0.2430)  loss_box_reg: 0.3063 (0.3130)  loss_objectness: 0.0716 (0.0740)  loss_rpn_box_reg: 0.0629 (0.0614)  time: 1.6837  data: 0.6650  max mem: 19870\n",
      "Epoch: [1]  [1620/2444]  eta: 0:23:17  lr: 0.020000  loss: 0.6720 (0.6914)  loss_classifier: 0.2393 (0.2430)  loss_box_reg: 0.3052 (0.3130)  loss_objectness: 0.0716 (0.0740)  loss_rpn_box_reg: 0.0633 (0.0614)  time: 1.6960  data: 0.6771  max mem: 19870\n",
      "Epoch: [1]  [1630/2444]  eta: 0:23:00  lr: 0.020000  loss: 0.6365 (0.6912)  loss_classifier: 0.2279 (0.2428)  loss_box_reg: 0.3002 (0.3129)  loss_objectness: 0.0681 (0.0740)  loss_rpn_box_reg: 0.0612 (0.0614)  time: 1.7033  data: 0.6838  max mem: 19870\n",
      "Epoch: [1]  [1640/2444]  eta: 0:22:43  lr: 0.020000  loss: 0.6365 (0.6909)  loss_classifier: 0.2259 (0.2428)  loss_box_reg: 0.2921 (0.3128)  loss_objectness: 0.0681 (0.0740)  loss_rpn_box_reg: 0.0597 (0.0614)  time: 1.7008  data: 0.6747  max mem: 19870\n",
      "Epoch: [1]  [1650/2444]  eta: 0:22:26  lr: 0.020000  loss: 0.6324 (0.6908)  loss_classifier: 0.2259 (0.2427)  loss_box_reg: 0.2988 (0.3127)  loss_objectness: 0.0700 (0.0740)  loss_rpn_box_reg: 0.0559 (0.0614)  time: 1.7013  data: 0.6749  max mem: 19870\n",
      "Epoch: [1]  [1660/2444]  eta: 0:22:09  lr: 0.020000  loss: 0.6444 (0.6909)  loss_classifier: 0.2373 (0.2427)  loss_box_reg: 0.3044 (0.3127)  loss_objectness: 0.0698 (0.0740)  loss_rpn_box_reg: 0.0666 (0.0614)  time: 1.7020  data: 0.6829  max mem: 19870\n",
      "Epoch: [1]  [1670/2444]  eta: 0:21:52  lr: 0.020000  loss: 0.6607 (0.6907)  loss_classifier: 0.2341 (0.2426)  loss_box_reg: 0.3110 (0.3127)  loss_objectness: 0.0656 (0.0740)  loss_rpn_box_reg: 0.0615 (0.0614)  time: 1.6989  data: 0.6812  max mem: 19870\n",
      "Epoch: [1]  [1680/2444]  eta: 0:21:35  lr: 0.020000  loss: 0.6502 (0.6906)  loss_classifier: 0.2270 (0.2426)  loss_box_reg: 0.3016 (0.3126)  loss_objectness: 0.0656 (0.0739)  loss_rpn_box_reg: 0.0593 (0.0614)  time: 1.6967  data: 0.6777  max mem: 19870\n",
      "Epoch: [1]  [1690/2444]  eta: 0:21:18  lr: 0.020000  loss: 0.6822 (0.6906)  loss_classifier: 0.2339 (0.2426)  loss_box_reg: 0.3009 (0.3126)  loss_objectness: 0.0667 (0.0739)  loss_rpn_box_reg: 0.0627 (0.0614)  time: 1.6952  data: 0.6765  max mem: 19870\n",
      "Epoch: [1]  [1700/2444]  eta: 0:21:01  lr: 0.020000  loss: 0.6573 (0.6905)  loss_classifier: 0.2419 (0.2426)  loss_box_reg: 0.3009 (0.3125)  loss_objectness: 0.0616 (0.0739)  loss_rpn_box_reg: 0.0634 (0.0614)  time: 1.6909  data: 0.6731  max mem: 19870\n",
      "Epoch: [1]  [1710/2444]  eta: 0:20:44  lr: 0.020000  loss: 0.6573 (0.6903)  loss_classifier: 0.2333 (0.2425)  loss_box_reg: 0.2985 (0.3125)  loss_objectness: 0.0605 (0.0739)  loss_rpn_box_reg: 0.0555 (0.0614)  time: 1.6916  data: 0.6733  max mem: 19870\n",
      "Epoch: [1]  [1720/2444]  eta: 0:20:27  lr: 0.020000  loss: 0.6682 (0.6904)  loss_classifier: 0.2378 (0.2426)  loss_box_reg: 0.3053 (0.3125)  loss_objectness: 0.0689 (0.0739)  loss_rpn_box_reg: 0.0573 (0.0614)  time: 1.6939  data: 0.6753  max mem: 19870\n",
      "Epoch: [1]  [1730/2444]  eta: 0:20:10  lr: 0.020000  loss: 0.6897 (0.6903)  loss_classifier: 0.2363 (0.2425)  loss_box_reg: 0.3033 (0.3124)  loss_objectness: 0.0747 (0.0739)  loss_rpn_box_reg: 0.0573 (0.0614)  time: 1.6922  data: 0.6726  max mem: 19870\n",
      "Epoch: [1]  [1740/2444]  eta: 0:19:53  lr: 0.020000  loss: 0.6445 (0.6901)  loss_classifier: 0.2315 (0.2425)  loss_box_reg: 0.2934 (0.3123)  loss_objectness: 0.0747 (0.0739)  loss_rpn_box_reg: 0.0557 (0.0614)  time: 1.6929  data: 0.6730  max mem: 19870\n",
      "Epoch: [1]  [1750/2444]  eta: 0:19:36  lr: 0.020000  loss: 0.6657 (0.6900)  loss_classifier: 0.2350 (0.2425)  loss_box_reg: 0.3023 (0.3124)  loss_objectness: 0.0637 (0.0738)  loss_rpn_box_reg: 0.0621 (0.0614)  time: 1.7024  data: 0.6836  max mem: 19870\n",
      "Epoch: [1]  [1760/2444]  eta: 0:19:19  lr: 0.020000  loss: 0.6657 (0.6899)  loss_classifier: 0.2379 (0.2424)  loss_box_reg: 0.3010 (0.3123)  loss_objectness: 0.0633 (0.0738)  loss_rpn_box_reg: 0.0569 (0.0614)  time: 1.6962  data: 0.6767  max mem: 19870\n",
      "Epoch: [1]  [1770/2444]  eta: 0:19:02  lr: 0.020000  loss: 0.6694 (0.6900)  loss_classifier: 0.2431 (0.2425)  loss_box_reg: 0.3098 (0.3123)  loss_objectness: 0.0584 (0.0738)  loss_rpn_box_reg: 0.0569 (0.0614)  time: 1.6888  data: 0.6687  max mem: 19870\n",
      "Epoch: [1]  [1780/2444]  eta: 0:18:45  lr: 0.020000  loss: 0.6803 (0.6899)  loss_classifier: 0.2460 (0.2425)  loss_box_reg: 0.3116 (0.3123)  loss_objectness: 0.0591 (0.0738)  loss_rpn_box_reg: 0.0585 (0.0614)  time: 1.6902  data: 0.6710  max mem: 19870\n",
      "Epoch: [1]  [1790/2444]  eta: 0:18:28  lr: 0.020000  loss: 0.6974 (0.6900)  loss_classifier: 0.2413 (0.2425)  loss_box_reg: 0.3116 (0.3123)  loss_objectness: 0.0602 (0.0738)  loss_rpn_box_reg: 0.0602 (0.0614)  time: 1.6912  data: 0.6722  max mem: 19870\n",
      "Epoch: [1]  [1800/2444]  eta: 0:18:11  lr: 0.020000  loss: 0.7088 (0.6901)  loss_classifier: 0.2531 (0.2426)  loss_box_reg: 0.3146 (0.3123)  loss_objectness: 0.0721 (0.0738)  loss_rpn_box_reg: 0.0619 (0.0614)  time: 1.6987  data: 0.6786  max mem: 19870\n",
      "Epoch: [1]  [1810/2444]  eta: 0:17:54  lr: 0.020000  loss: 0.6582 (0.6899)  loss_classifier: 0.2310 (0.2425)  loss_box_reg: 0.3100 (0.3123)  loss_objectness: 0.0711 (0.0737)  loss_rpn_box_reg: 0.0576 (0.0614)  time: 1.6947  data: 0.6732  max mem: 19870\n",
      "Epoch: [1]  [1820/2444]  eta: 0:17:38  lr: 0.020000  loss: 0.7035 (0.6900)  loss_classifier: 0.2349 (0.2425)  loss_box_reg: 0.3143 (0.3123)  loss_objectness: 0.0690 (0.0737)  loss_rpn_box_reg: 0.0597 (0.0614)  time: 1.6972  data: 0.6767  max mem: 19870\n",
      "Epoch: [1]  [1830/2444]  eta: 0:17:21  lr: 0.020000  loss: 0.7035 (0.6897)  loss_classifier: 0.2349 (0.2424)  loss_box_reg: 0.3098 (0.3122)  loss_objectness: 0.0721 (0.0737)  loss_rpn_box_reg: 0.0601 (0.0614)  time: 1.6999  data: 0.6748  max mem: 19870\n",
      "Epoch: [1]  [1840/2444]  eta: 0:17:04  lr: 0.020000  loss: 0.6733 (0.6898)  loss_classifier: 0.2242 (0.2425)  loss_box_reg: 0.2945 (0.3122)  loss_objectness: 0.0674 (0.0738)  loss_rpn_box_reg: 0.0601 (0.0614)  time: 1.6975  data: 0.6727  max mem: 19870\n",
      "Epoch: [1]  [1850/2444]  eta: 0:16:47  lr: 0.020000  loss: 0.6733 (0.6897)  loss_classifier: 0.2324 (0.2424)  loss_box_reg: 0.3046 (0.3121)  loss_objectness: 0.0663 (0.0738)  loss_rpn_box_reg: 0.0596 (0.0614)  time: 1.7022  data: 0.6831  max mem: 19870\n",
      "Epoch: [1]  [1860/2444]  eta: 0:16:30  lr: 0.020000  loss: 0.6726 (0.6896)  loss_classifier: 0.2324 (0.2424)  loss_box_reg: 0.3003 (0.3121)  loss_objectness: 0.0634 (0.0738)  loss_rpn_box_reg: 0.0547 (0.0613)  time: 1.6953  data: 0.6763  max mem: 19870\n",
      "Epoch: [1]  [1870/2444]  eta: 0:16:13  lr: 0.020000  loss: 0.6918 (0.6898)  loss_classifier: 0.2422 (0.2425)  loss_box_reg: 0.3080 (0.3122)  loss_objectness: 0.0724 (0.0738)  loss_rpn_box_reg: 0.0567 (0.0614)  time: 1.6899  data: 0.6711  max mem: 19870\n",
      "Epoch: [1]  [1880/2444]  eta: 0:15:56  lr: 0.020000  loss: 0.6733 (0.6897)  loss_classifier: 0.2392 (0.2424)  loss_box_reg: 0.3080 (0.3121)  loss_objectness: 0.0667 (0.0739)  loss_rpn_box_reg: 0.0595 (0.0613)  time: 1.6862  data: 0.6661  max mem: 19870\n",
      "Epoch: [1]  [1890/2444]  eta: 0:15:39  lr: 0.020000  loss: 0.6696 (0.6899)  loss_classifier: 0.2306 (0.2424)  loss_box_reg: 0.3070 (0.3122)  loss_objectness: 0.0707 (0.0740)  loss_rpn_box_reg: 0.0619 (0.0614)  time: 1.6876  data: 0.6674  max mem: 19870\n",
      "Epoch: [1]  [1900/2444]  eta: 0:15:22  lr: 0.020000  loss: 0.6955 (0.6900)  loss_classifier: 0.2441 (0.2425)  loss_box_reg: 0.3145 (0.3122)  loss_objectness: 0.0761 (0.0740)  loss_rpn_box_reg: 0.0625 (0.0614)  time: 1.7025  data: 0.6819  max mem: 19870\n",
      "Epoch: [1]  [1910/2444]  eta: 0:15:05  lr: 0.020000  loss: 0.6955 (0.6901)  loss_classifier: 0.2452 (0.2425)  loss_box_reg: 0.3129 (0.3122)  loss_objectness: 0.0747 (0.0740)  loss_rpn_box_reg: 0.0625 (0.0614)  time: 1.7035  data: 0.6825  max mem: 19870\n",
      "Epoch: [1]  [1920/2444]  eta: 0:14:48  lr: 0.020000  loss: 0.6519 (0.6898)  loss_classifier: 0.2326 (0.2424)  loss_box_reg: 0.2925 (0.3121)  loss_objectness: 0.0656 (0.0740)  loss_rpn_box_reg: 0.0581 (0.0614)  time: 1.6897  data: 0.6699  max mem: 19870\n",
      "Epoch: [1]  [1930/2444]  eta: 0:14:31  lr: 0.020000  loss: 0.6331 (0.6897)  loss_classifier: 0.2256 (0.2423)  loss_box_reg: 0.2913 (0.3120)  loss_objectness: 0.0656 (0.0740)  loss_rpn_box_reg: 0.0551 (0.0614)  time: 1.6829  data: 0.6631  max mem: 19870\n",
      "Epoch: [1]  [1940/2444]  eta: 0:14:14  lr: 0.020000  loss: 0.6465 (0.6897)  loss_classifier: 0.2316 (0.2423)  loss_box_reg: 0.2993 (0.3120)  loss_objectness: 0.0720 (0.0740)  loss_rpn_box_reg: 0.0612 (0.0614)  time: 1.6896  data: 0.6689  max mem: 19870\n",
      "Epoch: [1]  [1950/2444]  eta: 0:13:57  lr: 0.020000  loss: 0.6538 (0.6896)  loss_classifier: 0.2374 (0.2423)  loss_box_reg: 0.2984 (0.3119)  loss_objectness: 0.0693 (0.0740)  loss_rpn_box_reg: 0.0622 (0.0614)  time: 1.6937  data: 0.6715  max mem: 19870\n",
      "Epoch: [1]  [1960/2444]  eta: 0:13:40  lr: 0.020000  loss: 0.6536 (0.6894)  loss_classifier: 0.2267 (0.2422)  loss_box_reg: 0.2984 (0.3119)  loss_objectness: 0.0600 (0.0739)  loss_rpn_box_reg: 0.0574 (0.0614)  time: 1.6876  data: 0.6666  max mem: 19870\n",
      "Epoch: [1]  [1970/2444]  eta: 0:13:23  lr: 0.020000  loss: 0.6430 (0.6893)  loss_classifier: 0.2267 (0.2421)  loss_box_reg: 0.3019 (0.3119)  loss_objectness: 0.0559 (0.0739)  loss_rpn_box_reg: 0.0564 (0.0614)  time: 1.6923  data: 0.6720  max mem: 19870\n",
      "Epoch: [1]  [1980/2444]  eta: 0:13:06  lr: 0.020000  loss: 0.6379 (0.6892)  loss_classifier: 0.2211 (0.2421)  loss_box_reg: 0.2987 (0.3118)  loss_objectness: 0.0564 (0.0739)  loss_rpn_box_reg: 0.0611 (0.0614)  time: 1.7027  data: 0.6807  max mem: 19870\n",
      "Epoch: [1]  [1990/2444]  eta: 0:12:49  lr: 0.020000  loss: 0.6536 (0.6892)  loss_classifier: 0.2346 (0.2421)  loss_box_reg: 0.3038 (0.3118)  loss_objectness: 0.0623 (0.0739)  loss_rpn_box_reg: 0.0611 (0.0614)  time: 1.6976  data: 0.6764  max mem: 19870\n",
      "Epoch: [1]  [2000/2444]  eta: 0:12:32  lr: 0.020000  loss: 0.6773 (0.6889)  loss_classifier: 0.2309 (0.2420)  loss_box_reg: 0.3011 (0.3117)  loss_objectness: 0.0655 (0.0738)  loss_rpn_box_reg: 0.0569 (0.0614)  time: 1.6828  data: 0.6624  max mem: 19870\n",
      "Epoch: [1]  [2010/2444]  eta: 0:12:15  lr: 0.020000  loss: 0.6475 (0.6888)  loss_classifier: 0.2279 (0.2419)  loss_box_reg: 0.3004 (0.3117)  loss_objectness: 0.0680 (0.0738)  loss_rpn_box_reg: 0.0563 (0.0614)  time: 1.6859  data: 0.6653  max mem: 19870\n",
      "Epoch: [1]  [2020/2444]  eta: 0:11:58  lr: 0.020000  loss: 0.6353 (0.6887)  loss_classifier: 0.2266 (0.2419)  loss_box_reg: 0.3030 (0.3116)  loss_objectness: 0.0658 (0.0738)  loss_rpn_box_reg: 0.0589 (0.0614)  time: 1.6924  data: 0.6721  max mem: 19870\n",
      "Epoch: [1]  [2030/2444]  eta: 0:11:41  lr: 0.020000  loss: 0.6502 (0.6886)  loss_classifier: 0.2280 (0.2418)  loss_box_reg: 0.3030 (0.3115)  loss_objectness: 0.0735 (0.0739)  loss_rpn_box_reg: 0.0645 (0.0614)  time: 1.6940  data: 0.6748  max mem: 19870\n",
      "Epoch: [1]  [2040/2444]  eta: 0:11:24  lr: 0.020000  loss: 0.6502 (0.6884)  loss_classifier: 0.2238 (0.2418)  loss_box_reg: 0.2984 (0.3115)  loss_objectness: 0.0695 (0.0738)  loss_rpn_box_reg: 0.0657 (0.0614)  time: 1.6982  data: 0.6806  max mem: 19870\n",
      "Epoch: [1]  [2050/2444]  eta: 0:11:07  lr: 0.020000  loss: 0.6714 (0.6885)  loss_classifier: 0.2261 (0.2417)  loss_box_reg: 0.3088 (0.3115)  loss_objectness: 0.0684 (0.0739)  loss_rpn_box_reg: 0.0653 (0.0614)  time: 1.6963  data: 0.6787  max mem: 19870\n",
      "Epoch: [1]  [2060/2444]  eta: 0:10:50  lr: 0.020000  loss: 0.6740 (0.6883)  loss_classifier: 0.2280 (0.2417)  loss_box_reg: 0.3047 (0.3114)  loss_objectness: 0.0680 (0.0738)  loss_rpn_box_reg: 0.0587 (0.0614)  time: 1.6885  data: 0.6701  max mem: 19870\n",
      "Epoch: [1]  [2070/2444]  eta: 0:10:33  lr: 0.020000  loss: 0.6698 (0.6882)  loss_classifier: 0.2349 (0.2417)  loss_box_reg: 0.3006 (0.3114)  loss_objectness: 0.0644 (0.0738)  loss_rpn_box_reg: 0.0526 (0.0613)  time: 1.6846  data: 0.6656  max mem: 19870\n",
      "Epoch: [1]  [2080/2444]  eta: 0:10:17  lr: 0.020000  loss: 0.6698 (0.6881)  loss_classifier: 0.2349 (0.2416)  loss_box_reg: 0.3006 (0.3113)  loss_objectness: 0.0598 (0.0738)  loss_rpn_box_reg: 0.0566 (0.0613)  time: 1.6846  data: 0.6662  max mem: 19870\n",
      "Epoch: [1]  [2090/2444]  eta: 0:10:00  lr: 0.020000  loss: 0.6631 (0.6880)  loss_classifier: 0.2318 (0.2416)  loss_box_reg: 0.2994 (0.3112)  loss_objectness: 0.0657 (0.0738)  loss_rpn_box_reg: 0.0631 (0.0614)  time: 1.6895  data: 0.6709  max mem: 19870\n",
      "Epoch: [1]  [2100/2444]  eta: 0:09:43  lr: 0.020000  loss: 0.6957 (0.6882)  loss_classifier: 0.2388 (0.2417)  loss_box_reg: 0.3202 (0.3113)  loss_objectness: 0.0657 (0.0738)  loss_rpn_box_reg: 0.0671 (0.0614)  time: 1.7005  data: 0.6803  max mem: 19870\n",
      "Epoch: [1]  [2110/2444]  eta: 0:09:26  lr: 0.020000  loss: 0.7006 (0.6883)  loss_classifier: 0.2500 (0.2417)  loss_box_reg: 0.3205 (0.3113)  loss_objectness: 0.0673 (0.0739)  loss_rpn_box_reg: 0.0639 (0.0614)  time: 1.6923  data: 0.6728  max mem: 19870\n",
      "Epoch: [1]  [2120/2444]  eta: 0:09:09  lr: 0.020000  loss: 0.6775 (0.6883)  loss_classifier: 0.2391 (0.2417)  loss_box_reg: 0.2955 (0.3113)  loss_objectness: 0.0733 (0.0739)  loss_rpn_box_reg: 0.0606 (0.0614)  time: 1.6870  data: 0.6685  max mem: 19870\n",
      "Epoch: [1]  [2130/2444]  eta: 0:08:52  lr: 0.020000  loss: 0.6636 (0.6883)  loss_classifier: 0.2372 (0.2417)  loss_box_reg: 0.3047 (0.3112)  loss_objectness: 0.0680 (0.0739)  loss_rpn_box_reg: 0.0620 (0.0614)  time: 1.6949  data: 0.6758  max mem: 19870\n",
      "Epoch: [1]  [2140/2444]  eta: 0:08:35  lr: 0.020000  loss: 0.6784 (0.6884)  loss_classifier: 0.2338 (0.2417)  loss_box_reg: 0.3069 (0.3112)  loss_objectness: 0.0830 (0.0740)  loss_rpn_box_reg: 0.0696 (0.0615)  time: 1.6993  data: 0.6791  max mem: 19870\n",
      "Epoch: [1]  [2150/2444]  eta: 0:08:18  lr: 0.020000  loss: 0.6784 (0.6883)  loss_classifier: 0.2380 (0.2417)  loss_box_reg: 0.3069 (0.3112)  loss_objectness: 0.0693 (0.0739)  loss_rpn_box_reg: 0.0590 (0.0615)  time: 1.6972  data: 0.6773  max mem: 19870\n",
      "Epoch: [1]  [2160/2444]  eta: 0:08:01  lr: 0.020000  loss: 0.6654 (0.6883)  loss_classifier: 0.2406 (0.2418)  loss_box_reg: 0.3072 (0.3112)  loss_objectness: 0.0621 (0.0739)  loss_rpn_box_reg: 0.0587 (0.0615)  time: 1.6937  data: 0.6753  max mem: 19870\n",
      "Epoch: [1]  [2170/2444]  eta: 0:07:44  lr: 0.020000  loss: 0.6593 (0.6883)  loss_classifier: 0.2338 (0.2417)  loss_box_reg: 0.3019 (0.3112)  loss_objectness: 0.0662 (0.0739)  loss_rpn_box_reg: 0.0598 (0.0615)  time: 1.6977  data: 0.6790  max mem: 19870\n",
      "Epoch: [1]  [2180/2444]  eta: 0:07:27  lr: 0.020000  loss: 0.6731 (0.6883)  loss_classifier: 0.2329 (0.2417)  loss_box_reg: 0.3019 (0.3112)  loss_objectness: 0.0748 (0.0739)  loss_rpn_box_reg: 0.0619 (0.0615)  time: 1.6956  data: 0.6769  max mem: 19870\n",
      "Epoch: [1]  [2190/2444]  eta: 0:07:10  lr: 0.020000  loss: 0.6660 (0.6882)  loss_classifier: 0.2177 (0.2417)  loss_box_reg: 0.3007 (0.3111)  loss_objectness: 0.0730 (0.0739)  loss_rpn_box_reg: 0.0653 (0.0615)  time: 1.6877  data: 0.6688  max mem: 19870\n",
      "Epoch: [1]  [2200/2444]  eta: 0:06:53  lr: 0.020000  loss: 0.6365 (0.6880)  loss_classifier: 0.2153 (0.2416)  loss_box_reg: 0.2919 (0.3110)  loss_objectness: 0.0652 (0.0738)  loss_rpn_box_reg: 0.0596 (0.0615)  time: 1.6909  data: 0.6705  max mem: 19870\n",
      "Epoch: [1]  [2210/2444]  eta: 0:06:36  lr: 0.020000  loss: 0.6416 (0.6879)  loss_classifier: 0.2262 (0.2416)  loss_box_reg: 0.3052 (0.3110)  loss_objectness: 0.0595 (0.0738)  loss_rpn_box_reg: 0.0549 (0.0615)  time: 1.7089  data: 0.6887  max mem: 19870\n",
      "Epoch: [1]  [2220/2444]  eta: 0:06:19  lr: 0.020000  loss: 0.6393 (0.6877)  loss_classifier: 0.2262 (0.2415)  loss_box_reg: 0.2946 (0.3110)  loss_objectness: 0.0595 (0.0738)  loss_rpn_box_reg: 0.0548 (0.0615)  time: 1.7016  data: 0.6815  max mem: 19870\n",
      "Epoch: [1]  [2230/2444]  eta: 0:06:02  lr: 0.020000  loss: 0.6414 (0.6875)  loss_classifier: 0.2230 (0.2415)  loss_box_reg: 0.2891 (0.3109)  loss_objectness: 0.0596 (0.0737)  loss_rpn_box_reg: 0.0553 (0.0614)  time: 1.6899  data: 0.6706  max mem: 19870\n",
      "Epoch: [1]  [2240/2444]  eta: 0:05:45  lr: 0.020000  loss: 0.6414 (0.6873)  loss_classifier: 0.2287 (0.2414)  loss_box_reg: 0.2977 (0.3108)  loss_objectness: 0.0558 (0.0736)  loss_rpn_box_reg: 0.0533 (0.0614)  time: 1.6985  data: 0.6806  max mem: 19870\n",
      "Epoch: [1]  [2250/2444]  eta: 0:05:28  lr: 0.020000  loss: 0.6469 (0.6873)  loss_classifier: 0.2325 (0.2414)  loss_box_reg: 0.2988 (0.3108)  loss_objectness: 0.0603 (0.0737)  loss_rpn_box_reg: 0.0595 (0.0614)  time: 1.6946  data: 0.6764  max mem: 19870\n",
      "Epoch: [1]  [2260/2444]  eta: 0:05:11  lr: 0.020000  loss: 0.6759 (0.6873)  loss_classifier: 0.2338 (0.2414)  loss_box_reg: 0.2973 (0.3108)  loss_objectness: 0.0744 (0.0737)  loss_rpn_box_reg: 0.0653 (0.0615)  time: 1.6930  data: 0.6735  max mem: 19870\n",
      "Epoch: [1]  [2270/2444]  eta: 0:04:54  lr: 0.020000  loss: 0.6703 (0.6873)  loss_classifier: 0.2381 (0.2414)  loss_box_reg: 0.2984 (0.3108)  loss_objectness: 0.0703 (0.0737)  loss_rpn_box_reg: 0.0611 (0.0615)  time: 1.6925  data: 0.6728  max mem: 19870\n",
      "Epoch: [1]  [2280/2444]  eta: 0:04:37  lr: 0.020000  loss: 0.7078 (0.6873)  loss_classifier: 0.2406 (0.2414)  loss_box_reg: 0.3083 (0.3108)  loss_objectness: 0.0668 (0.0737)  loss_rpn_box_reg: 0.0611 (0.0615)  time: 1.6891  data: 0.6707  max mem: 19870\n",
      "Epoch: [1]  [2290/2444]  eta: 0:04:21  lr: 0.020000  loss: 0.6738 (0.6872)  loss_classifier: 0.2406 (0.2414)  loss_box_reg: 0.3066 (0.3107)  loss_objectness: 0.0667 (0.0737)  loss_rpn_box_reg: 0.0628 (0.0615)  time: 1.6862  data: 0.6688  max mem: 19870\n",
      "Epoch: [1]  [2300/2444]  eta: 0:04:04  lr: 0.020000  loss: 0.6533 (0.6870)  loss_classifier: 0.2329 (0.2413)  loss_box_reg: 0.2993 (0.3106)  loss_objectness: 0.0607 (0.0736)  loss_rpn_box_reg: 0.0570 (0.0614)  time: 1.6881  data: 0.6706  max mem: 19870\n",
      "Epoch: [1]  [2310/2444]  eta: 0:03:47  lr: 0.020000  loss: 0.6533 (0.6869)  loss_classifier: 0.2319 (0.2413)  loss_box_reg: 0.2993 (0.3106)  loss_objectness: 0.0589 (0.0736)  loss_rpn_box_reg: 0.0544 (0.0614)  time: 1.6948  data: 0.6762  max mem: 19870\n",
      "Epoch: [1]  [2320/2444]  eta: 0:03:30  lr: 0.020000  loss: 0.6374 (0.6868)  loss_classifier: 0.2319 (0.2412)  loss_box_reg: 0.2955 (0.3106)  loss_objectness: 0.0522 (0.0735)  loss_rpn_box_reg: 0.0562 (0.0614)  time: 1.6893  data: 0.6711  max mem: 19870\n",
      "Epoch: [1]  [2330/2444]  eta: 0:03:13  lr: 0.020000  loss: 0.6599 (0.6867)  loss_classifier: 0.2341 (0.2412)  loss_box_reg: 0.2955 (0.3106)  loss_objectness: 0.0629 (0.0735)  loss_rpn_box_reg: 0.0620 (0.0614)  time: 1.6836  data: 0.6656  max mem: 19870\n",
      "Epoch: [1]  [2340/2444]  eta: 0:02:56  lr: 0.020000  loss: 0.6530 (0.6867)  loss_classifier: 0.2301 (0.2412)  loss_box_reg: 0.2968 (0.3105)  loss_objectness: 0.0693 (0.0735)  loss_rpn_box_reg: 0.0622 (0.0614)  time: 1.6914  data: 0.6727  max mem: 19870\n",
      "Epoch: [1]  [2350/2444]  eta: 0:02:39  lr: 0.020000  loss: 0.6530 (0.6867)  loss_classifier: 0.2294 (0.2412)  loss_box_reg: 0.2956 (0.3105)  loss_objectness: 0.0731 (0.0736)  loss_rpn_box_reg: 0.0618 (0.0614)  time: 1.6959  data: 0.6779  max mem: 19870\n",
      "Epoch: [1]  [2360/2444]  eta: 0:02:22  lr: 0.020000  loss: 0.6906 (0.6867)  loss_classifier: 0.2384 (0.2412)  loss_box_reg: 0.3054 (0.3105)  loss_objectness: 0.0754 (0.0736)  loss_rpn_box_reg: 0.0645 (0.0614)  time: 1.6979  data: 0.6797  max mem: 19870\n",
      "Epoch: [1]  [2370/2444]  eta: 0:02:05  lr: 0.020000  loss: 0.6696 (0.6866)  loss_classifier: 0.2360 (0.2411)  loss_box_reg: 0.3046 (0.3105)  loss_objectness: 0.0672 (0.0735)  loss_rpn_box_reg: 0.0653 (0.0614)  time: 1.6936  data: 0.6747  max mem: 19870\n",
      "Epoch: [1]  [2380/2444]  eta: 0:01:48  lr: 0.020000  loss: 0.6300 (0.6863)  loss_classifier: 0.2202 (0.2411)  loss_box_reg: 0.2950 (0.3104)  loss_objectness: 0.0545 (0.0735)  loss_rpn_box_reg: 0.0538 (0.0614)  time: 1.6904  data: 0.6708  max mem: 19870\n",
      "Epoch: [1]  [2390/2444]  eta: 0:01:31  lr: 0.020000  loss: 0.6458 (0.6862)  loss_classifier: 0.2346 (0.2411)  loss_box_reg: 0.2937 (0.3103)  loss_objectness: 0.0574 (0.0735)  loss_rpn_box_reg: 0.0543 (0.0614)  time: 1.6935  data: 0.6741  max mem: 19870\n",
      "Epoch: [1]  [2400/2444]  eta: 0:01:14  lr: 0.020000  loss: 0.6793 (0.6863)  loss_classifier: 0.2427 (0.2410)  loss_box_reg: 0.2969 (0.3103)  loss_objectness: 0.0840 (0.0736)  loss_rpn_box_reg: 0.0583 (0.0614)  time: 1.6936  data: 0.6747  max mem: 19870\n",
      "Epoch: [1]  [2410/2444]  eta: 0:00:57  lr: 0.020000  loss: 0.6576 (0.6862)  loss_classifier: 0.2314 (0.2410)  loss_box_reg: 0.3023 (0.3102)  loss_objectness: 0.0840 (0.0736)  loss_rpn_box_reg: 0.0583 (0.0614)  time: 1.7011  data: 0.6820  max mem: 19870\n",
      "Epoch: [1]  [2420/2444]  eta: 0:00:40  lr: 0.020000  loss: 0.6443 (0.6861)  loss_classifier: 0.2312 (0.2409)  loss_box_reg: 0.2933 (0.3102)  loss_objectness: 0.0699 (0.0736)  loss_rpn_box_reg: 0.0598 (0.0614)  time: 1.7076  data: 0.6882  max mem: 19870\n",
      "Epoch: [1]  [2430/2444]  eta: 0:00:23  lr: 0.020000  loss: 0.6599 (0.6861)  loss_classifier: 0.2303 (0.2409)  loss_box_reg: 0.3046 (0.3102)  loss_objectness: 0.0614 (0.0736)  loss_rpn_box_reg: 0.0596 (0.0614)  time: 1.6978  data: 0.6783  max mem: 19870\n",
      "Epoch: [1]  [2440/2444]  eta: 0:00:06  lr: 0.020000  loss: 0.6649 (0.6861)  loss_classifier: 0.2344 (0.2409)  loss_box_reg: 0.3046 (0.3102)  loss_objectness: 0.0723 (0.0736)  loss_rpn_box_reg: 0.0588 (0.0614)  time: 1.6994  data: 0.6797  max mem: 19870\n",
      "Epoch: [1]  [2443/2444]  eta: 0:00:01  lr: 0.020000  loss: 0.6649 (0.6863)  loss_classifier: 0.2344 (0.2409)  loss_box_reg: 0.3024 (0.3102)  loss_objectness: 0.0723 (0.0737)  loss_rpn_box_reg: 0.0609 (0.0615)  time: 1.6194  data: 0.6477  max mem: 19870\n",
      "Epoch: [1] Total time: 1:09:01 (1.6945 s / it)\n",
      "creating index...\n",
      "index created!\n",
      "Test:  [   0/4952]  eta: 0:07:02  model_time: 0.0468 (0.0468)  evaluator_time: 0.0163 (0.0163)  time: 0.0853  data: 0.0210  max mem: 19870\n",
      "Test:  [ 100/4952]  eta: 0:04:58  model_time: 0.0328 (0.0337)  evaluator_time: 0.0089 (0.0121)  time: 0.0618  data: 0.0168  max mem: 19870\n",
      "Test:  [ 200/4952]  eta: 0:04:48  model_time: 0.0332 (0.0337)  evaluator_time: 0.0057 (0.0111)  time: 0.0606  data: 0.0150  max mem: 19870\n",
      "Test:  [ 300/4952]  eta: 0:04:40  model_time: 0.0334 (0.0336)  evaluator_time: 0.0079 (0.0108)  time: 0.0640  data: 0.0165  max mem: 19870\n",
      "Test:  [ 400/4952]  eta: 0:04:33  model_time: 0.0333 (0.0336)  evaluator_time: 0.0065 (0.0106)  time: 0.0580  data: 0.0148  max mem: 19870\n",
      "Test:  [ 500/4952]  eta: 0:04:26  model_time: 0.0328 (0.0336)  evaluator_time: 0.0072 (0.0104)  time: 0.0596  data: 0.0151  max mem: 19870\n",
      "Test:  [ 600/4952]  eta: 0:04:21  model_time: 0.0328 (0.0336)  evaluator_time: 0.0058 (0.0106)  time: 0.0573  data: 0.0144  max mem: 19870\n",
      "Test:  [ 700/4952]  eta: 0:04:17  model_time: 0.0330 (0.0336)  evaluator_time: 0.0065 (0.0106)  time: 0.0587  data: 0.0151  max mem: 19870\n",
      "Test:  [ 800/4952]  eta: 0:04:11  model_time: 0.0331 (0.0336)  evaluator_time: 0.0089 (0.0107)  time: 0.0617  data: 0.0158  max mem: 19870\n",
      "Test:  [ 900/4952]  eta: 0:04:04  model_time: 0.0330 (0.0336)  evaluator_time: 0.0055 (0.0106)  time: 0.0576  data: 0.0143  max mem: 19870\n",
      "Test:  [1000/4952]  eta: 0:03:58  model_time: 0.0328 (0.0336)  evaluator_time: 0.0068 (0.0105)  time: 0.0555  data: 0.0139  max mem: 19870\n",
      "Test:  [1100/4952]  eta: 0:03:52  model_time: 0.0327 (0.0336)  evaluator_time: 0.0056 (0.0105)  time: 0.0565  data: 0.0145  max mem: 19870\n",
      "Test:  [1200/4952]  eta: 0:03:46  model_time: 0.0330 (0.0336)  evaluator_time: 0.0118 (0.0106)  time: 0.0665  data: 0.0168  max mem: 19870\n",
      "Test:  [1300/4952]  eta: 0:03:40  model_time: 0.0328 (0.0336)  evaluator_time: 0.0062 (0.0106)  time: 0.0606  data: 0.0152  max mem: 19870\n",
      "Test:  [1400/4952]  eta: 0:03:34  model_time: 0.0334 (0.0336)  evaluator_time: 0.0061 (0.0108)  time: 0.0612  data: 0.0153  max mem: 19870\n",
      "Test:  [1500/4952]  eta: 0:03:28  model_time: 0.0327 (0.0336)  evaluator_time: 0.0075 (0.0108)  time: 0.0607  data: 0.0160  max mem: 19870\n",
      "Test:  [1600/4952]  eta: 0:03:23  model_time: 0.0328 (0.0336)  evaluator_time: 0.0071 (0.0108)  time: 0.0627  data: 0.0160  max mem: 19870\n",
      "Test:  [1700/4952]  eta: 0:03:16  model_time: 0.0325 (0.0336)  evaluator_time: 0.0054 (0.0108)  time: 0.0596  data: 0.0152  max mem: 19870\n",
      "Test:  [1800/4952]  eta: 0:03:10  model_time: 0.0327 (0.0336)  evaluator_time: 0.0058 (0.0107)  time: 0.0602  data: 0.0148  max mem: 19870\n",
      "Test:  [1900/4952]  eta: 0:03:04  model_time: 0.0327 (0.0335)  evaluator_time: 0.0068 (0.0107)  time: 0.0581  data: 0.0147  max mem: 19870\n",
      "Test:  [2000/4952]  eta: 0:02:58  model_time: 0.0332 (0.0336)  evaluator_time: 0.0094 (0.0107)  time: 0.0639  data: 0.0163  max mem: 19870\n",
      "Test:  [2100/4952]  eta: 0:02:52  model_time: 0.0330 (0.0336)  evaluator_time: 0.0071 (0.0107)  time: 0.0611  data: 0.0157  max mem: 19870\n",
      "Test:  [2200/4952]  eta: 0:02:45  model_time: 0.0327 (0.0335)  evaluator_time: 0.0063 (0.0106)  time: 0.0623  data: 0.0155  max mem: 19870\n",
      "Test:  [2300/4952]  eta: 0:02:39  model_time: 0.0329 (0.0335)  evaluator_time: 0.0067 (0.0108)  time: 0.0621  data: 0.0156  max mem: 19870\n",
      "Test:  [2400/4952]  eta: 0:02:33  model_time: 0.0329 (0.0335)  evaluator_time: 0.0099 (0.0107)  time: 0.0653  data: 0.0164  max mem: 19870\n",
      "Test:  [2500/4952]  eta: 0:02:27  model_time: 0.0336 (0.0335)  evaluator_time: 0.0067 (0.0107)  time: 0.0636  data: 0.0170  max mem: 19870\n",
      "Test:  [2600/4952]  eta: 0:02:21  model_time: 0.0325 (0.0335)  evaluator_time: 0.0081 (0.0107)  time: 0.0592  data: 0.0143  max mem: 19870\n",
      "Test:  [2700/4952]  eta: 0:02:15  model_time: 0.0325 (0.0335)  evaluator_time: 0.0075 (0.0107)  time: 0.0584  data: 0.0142  max mem: 19870\n",
      "Test:  [2800/4952]  eta: 0:02:09  model_time: 0.0331 (0.0335)  evaluator_time: 0.0077 (0.0107)  time: 0.0609  data: 0.0150  max mem: 19870\n",
      "Test:  [2900/4952]  eta: 0:02:03  model_time: 0.0327 (0.0335)  evaluator_time: 0.0064 (0.0107)  time: 0.0627  data: 0.0165  max mem: 19870\n",
      "Test:  [3000/4952]  eta: 0:01:57  model_time: 0.0331 (0.0335)  evaluator_time: 0.0075 (0.0107)  time: 0.0610  data: 0.0152  max mem: 19870\n",
      "Test:  [3100/4952]  eta: 0:01:51  model_time: 0.0332 (0.0335)  evaluator_time: 0.0071 (0.0106)  time: 0.0619  data: 0.0156  max mem: 19870\n",
      "Test:  [3200/4952]  eta: 0:01:45  model_time: 0.0329 (0.0335)  evaluator_time: 0.0085 (0.0106)  time: 0.0620  data: 0.0158  max mem: 19870\n",
      "Test:  [3300/4952]  eta: 0:01:39  model_time: 0.0327 (0.0335)  evaluator_time: 0.0072 (0.0108)  time: 0.0601  data: 0.0157  max mem: 19870\n",
      "Test:  [3400/4952]  eta: 0:01:33  model_time: 0.0330 (0.0335)  evaluator_time: 0.0051 (0.0108)  time: 0.0610  data: 0.0157  max mem: 19870\n",
      "Test:  [3500/4952]  eta: 0:01:27  model_time: 0.0327 (0.0335)  evaluator_time: 0.0074 (0.0108)  time: 0.0555  data: 0.0137  max mem: 19870\n",
      "Test:  [3600/4952]  eta: 0:01:21  model_time: 0.0340 (0.0335)  evaluator_time: 0.0075 (0.0108)  time: 0.0600  data: 0.0158  max mem: 19870\n",
      "Test:  [3700/4952]  eta: 0:01:15  model_time: 0.0329 (0.0335)  evaluator_time: 0.0055 (0.0107)  time: 0.0563  data: 0.0143  max mem: 19870\n",
      "Test:  [3800/4952]  eta: 0:01:09  model_time: 0.0329 (0.0335)  evaluator_time: 0.0060 (0.0107)  time: 0.0555  data: 0.0137  max mem: 19870\n",
      "Test:  [3900/4952]  eta: 0:01:03  model_time: 0.0328 (0.0335)  evaluator_time: 0.0066 (0.0107)  time: 0.0578  data: 0.0147  max mem: 19870\n",
      "Test:  [4000/4952]  eta: 0:00:57  model_time: 0.0326 (0.0335)  evaluator_time: 0.0094 (0.0107)  time: 0.0593  data: 0.0147  max mem: 19870\n",
      "Test:  [4100/4952]  eta: 0:00:51  model_time: 0.0326 (0.0335)  evaluator_time: 0.0070 (0.0107)  time: 0.0583  data: 0.0148  max mem: 19870\n",
      "Test:  [4200/4952]  eta: 0:00:45  model_time: 0.0326 (0.0335)  evaluator_time: 0.0051 (0.0107)  time: 0.0568  data: 0.0137  max mem: 19870\n",
      "Test:  [4300/4952]  eta: 0:00:39  model_time: 0.0328 (0.0335)  evaluator_time: 0.0060 (0.0107)  time: 0.0567  data: 0.0138  max mem: 19870\n",
      "Test:  [4400/4952]  eta: 0:00:33  model_time: 0.0328 (0.0335)  evaluator_time: 0.0084 (0.0107)  time: 0.0582  data: 0.0148  max mem: 19870\n",
      "Test:  [4500/4952]  eta: 0:00:27  model_time: 0.0329 (0.0335)  evaluator_time: 0.0072 (0.0106)  time: 0.0607  data: 0.0158  max mem: 19870\n",
      "Test:  [4600/4952]  eta: 0:00:21  model_time: 0.0327 (0.0335)  evaluator_time: 0.0068 (0.0107)  time: 0.0571  data: 0.0150  max mem: 19870\n",
      "Test:  [4700/4952]  eta: 0:00:15  model_time: 0.0330 (0.0335)  evaluator_time: 0.0058 (0.0107)  time: 0.0571  data: 0.0148  max mem: 19870\n",
      "Test:  [4800/4952]  eta: 0:00:09  model_time: 0.0331 (0.0335)  evaluator_time: 0.0082 (0.0107)  time: 0.0595  data: 0.0148  max mem: 19870\n",
      "Test:  [4900/4952]  eta: 0:00:03  model_time: 0.0328 (0.0335)  evaluator_time: 0.0051 (0.0107)  time: 0.0560  data: 0.0134  max mem: 19870\n",
      "Test:  [4951/4952]  eta: 0:00:00  model_time: 0.0327 (0.0335)  evaluator_time: 0.0060 (0.0107)  time: 0.0586  data: 0.0139  max mem: 19870\n",
      "Test: Total time: 0:04:57 (0.0601 s / it)\n",
      "Averaged stats: model_time: 0.0327 (0.0335)  evaluator_time: 0.0060 (0.0107)\n",
      "Accumulating evaluation results...\n",
      "DONE (t=8.12s).\n",
      "IoU metric: bbox\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.318\n",
      " Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.535\n",
      " Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.339\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.131\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.315\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.427\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.276\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.457\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.485\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.262\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.490\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.596\n",
      "Epoch: [2]  [   0/2444]  eta: 1:09:26  lr: 0.020000  loss: 0.6384 (0.6384)  loss_classifier: 0.2352 (0.2352)  loss_box_reg: 0.2751 (0.2751)  loss_objectness: 0.0741 (0.0741)  loss_rpn_box_reg: 0.0540 (0.0540)  time: 1.7048  data: 0.6656  max mem: 19870\n",
      "Epoch: [2]  [  10/2444]  eta: 1:08:53  lr: 0.020000  loss: 0.6504 (0.6730)  loss_classifier: 0.2388 (0.2360)  loss_box_reg: 0.2979 (0.3048)  loss_objectness: 0.0655 (0.0695)  loss_rpn_box_reg: 0.0592 (0.0628)  time: 1.6981  data: 0.6769  max mem: 19870\n",
      "Epoch: [2]  [  20/2444]  eta: 1:08:33  lr: 0.020000  loss: 0.6426 (0.6534)  loss_classifier: 0.2379 (0.2320)  loss_box_reg: 0.2973 (0.2956)  loss_objectness: 0.0581 (0.0660)  loss_rpn_box_reg: 0.0581 (0.0598)  time: 1.6967  data: 0.6765  max mem: 19870\n",
      "Epoch: [2]  [  30/2444]  eta: 1:08:13  lr: 0.020000  loss: 0.6424 (0.6548)  loss_classifier: 0.2306 (0.2310)  loss_box_reg: 0.2899 (0.2962)  loss_objectness: 0.0621 (0.0674)  loss_rpn_box_reg: 0.0566 (0.0601)  time: 1.6946  data: 0.6730  max mem: 19870\n",
      "Epoch: [2]  [  40/2444]  eta: 1:07:55  lr: 0.020000  loss: 0.6451 (0.6569)  loss_classifier: 0.2250 (0.2307)  loss_box_reg: 0.3065 (0.2985)  loss_objectness: 0.0648 (0.0682)  loss_rpn_box_reg: 0.0591 (0.0594)  time: 1.6938  data: 0.6732  max mem: 19870\n",
      "Epoch: [2]  [  50/2444]  eta: 1:07:38  lr: 0.020000  loss: 0.6543 (0.6564)  loss_classifier: 0.2193 (0.2292)  loss_box_reg: 0.2994 (0.2990)  loss_objectness: 0.0595 (0.0686)  loss_rpn_box_reg: 0.0593 (0.0596)  time: 1.6945  data: 0.6758  max mem: 19870\n",
      "Epoch: [2]  [  60/2444]  eta: 1:07:21  lr: 0.020000  loss: 0.6639 (0.6610)  loss_classifier: 0.2296 (0.2318)  loss_box_reg: 0.3001 (0.3001)  loss_objectness: 0.0667 (0.0690)  loss_rpn_box_reg: 0.0619 (0.0601)  time: 1.6953  data: 0.6770  max mem: 19870\n",
      "Epoch: [2]  [  70/2444]  eta: 1:07:07  lr: 0.020000  loss: 0.6787 (0.6647)  loss_classifier: 0.2361 (0.2332)  loss_box_reg: 0.3078 (0.3002)  loss_objectness: 0.0735 (0.0710)  loss_rpn_box_reg: 0.0615 (0.0603)  time: 1.6997  data: 0.6807  max mem: 19870\n",
      "Epoch: [2]  [  80/2444]  eta: 1:06:49  lr: 0.020000  loss: 0.6774 (0.6646)  loss_classifier: 0.2359 (0.2331)  loss_box_reg: 0.3052 (0.3015)  loss_objectness: 0.0655 (0.0700)  loss_rpn_box_reg: 0.0593 (0.0600)  time: 1.6978  data: 0.6795  max mem: 19870\n",
      "Epoch: [2]  [  90/2444]  eta: 1:06:33  lr: 0.020000  loss: 0.6885 (0.6690)  loss_classifier: 0.2385 (0.2342)  loss_box_reg: 0.3065 (0.3035)  loss_objectness: 0.0628 (0.0704)  loss_rpn_box_reg: 0.0593 (0.0608)  time: 1.6958  data: 0.6779  max mem: 19870\n",
      "Epoch: [2]  [ 100/2444]  eta: 1:06:18  lr: 0.020000  loss: 0.6828 (0.6688)  loss_classifier: 0.2385 (0.2345)  loss_box_reg: 0.3058 (0.3035)  loss_objectness: 0.0640 (0.0701)  loss_rpn_box_reg: 0.0643 (0.0608)  time: 1.7018  data: 0.6830  max mem: 19870\n",
      "Epoch: [2]  [ 110/2444]  eta: 1:06:02  lr: 0.020000  loss: 0.6788 (0.6714)  loss_classifier: 0.2424 (0.2351)  loss_box_reg: 0.3056 (0.3037)  loss_objectness: 0.0640 (0.0712)  loss_rpn_box_reg: 0.0652 (0.0614)  time: 1.7043  data: 0.6851  max mem: 19870\n",
      "Epoch: [2]  [ 120/2444]  eta: 1:05:44  lr: 0.020000  loss: 0.6359 (0.6672)  loss_classifier: 0.2327 (0.2340)  loss_box_reg: 0.2904 (0.3018)  loss_objectness: 0.0654 (0.0704)  loss_rpn_box_reg: 0.0605 (0.0611)  time: 1.6970  data: 0.6761  max mem: 19870\n",
      "Epoch: [2]  [ 130/2444]  eta: 1:05:25  lr: 0.020000  loss: 0.6322 (0.6662)  loss_classifier: 0.2216 (0.2336)  loss_box_reg: 0.2733 (0.3017)  loss_objectness: 0.0650 (0.0699)  loss_rpn_box_reg: 0.0559 (0.0611)  time: 1.6886  data: 0.6670  max mem: 19870\n",
      "Epoch: [2]  [ 140/2444]  eta: 1:05:08  lr: 0.020000  loss: 0.6540 (0.6657)  loss_classifier: 0.2293 (0.2331)  loss_box_reg: 0.2982 (0.3014)  loss_objectness: 0.0654 (0.0701)  loss_rpn_box_reg: 0.0557 (0.0611)  time: 1.6928  data: 0.6726  max mem: 19870\n",
      "Epoch: [2]  [ 150/2444]  eta: 1:04:52  lr: 0.020000  loss: 0.6842 (0.6665)  loss_classifier: 0.2333 (0.2336)  loss_box_reg: 0.3113 (0.3020)  loss_objectness: 0.0716 (0.0699)  loss_rpn_box_reg: 0.0617 (0.0611)  time: 1.7001  data: 0.6803  max mem: 19870\n",
      "Epoch: [2]  [ 160/2444]  eta: 1:04:36  lr: 0.020000  loss: 0.6842 (0.6693)  loss_classifier: 0.2472 (0.2346)  loss_box_reg: 0.3076 (0.3025)  loss_objectness: 0.0730 (0.0708)  loss_rpn_box_reg: 0.0617 (0.0614)  time: 1.7028  data: 0.6830  max mem: 19870\n",
      "Epoch: [2]  [ 170/2444]  eta: 1:04:17  lr: 0.020000  loss: 0.6588 (0.6673)  loss_classifier: 0.2363 (0.2343)  loss_box_reg: 0.2984 (0.3017)  loss_objectness: 0.0700 (0.0703)  loss_rpn_box_reg: 0.0577 (0.0611)  time: 1.6943  data: 0.6729  max mem: 19870\n",
      "Epoch: [2]  [ 180/2444]  eta: 1:04:00  lr: 0.020000  loss: 0.6283 (0.6659)  loss_classifier: 0.2313 (0.2340)  loss_box_reg: 0.2849 (0.3010)  loss_objectness: 0.0650 (0.0701)  loss_rpn_box_reg: 0.0551 (0.0608)  time: 1.6888  data: 0.6678  max mem: 19870\n",
      "Epoch: [2]  [ 190/2444]  eta: 1:03:45  lr: 0.020000  loss: 0.6448 (0.6661)  loss_classifier: 0.2314 (0.2342)  loss_box_reg: 0.2905 (0.3012)  loss_objectness: 0.0633 (0.0699)  loss_rpn_box_reg: 0.0592 (0.0609)  time: 1.7029  data: 0.6825  max mem: 19870\n",
      "Epoch: [2]  [ 200/2444]  eta: 1:03:28  lr: 0.020000  loss: 0.6193 (0.6650)  loss_classifier: 0.2114 (0.2335)  loss_box_reg: 0.2777 (0.3005)  loss_objectness: 0.0683 (0.0701)  loss_rpn_box_reg: 0.0600 (0.0609)  time: 1.7048  data: 0.6827  max mem: 19870\n",
      "Epoch: [2]  [ 210/2444]  eta: 1:03:10  lr: 0.020000  loss: 0.6336 (0.6645)  loss_classifier: 0.2117 (0.2333)  loss_box_reg: 0.2905 (0.3002)  loss_objectness: 0.0711 (0.0700)  loss_rpn_box_reg: 0.0610 (0.0610)  time: 1.6927  data: 0.6722  max mem: 19870\n",
      "Epoch: [2]  [ 220/2444]  eta: 1:02:53  lr: 0.020000  loss: 0.6479 (0.6642)  loss_classifier: 0.2303 (0.2331)  loss_box_reg: 0.2940 (0.2999)  loss_objectness: 0.0637 (0.0702)  loss_rpn_box_reg: 0.0610 (0.0610)  time: 1.6904  data: 0.6710  max mem: 19870\n",
      "Epoch: [2]  [ 230/2444]  eta: 1:02:34  lr: 0.020000  loss: 0.6742 (0.6648)  loss_classifier: 0.2333 (0.2333)  loss_box_reg: 0.2994 (0.2999)  loss_objectness: 0.0676 (0.0704)  loss_rpn_box_reg: 0.0580 (0.0611)  time: 1.6876  data: 0.6677  max mem: 19870\n",
      "Epoch: [2]  [ 240/2444]  eta: 1:02:18  lr: 0.020000  loss: 0.6409 (0.6631)  loss_classifier: 0.2264 (0.2329)  loss_box_reg: 0.2990 (0.2996)  loss_objectness: 0.0555 (0.0697)  loss_rpn_box_reg: 0.0580 (0.0608)  time: 1.6897  data: 0.6689  max mem: 19870\n",
      "Epoch: [2]  [ 250/2444]  eta: 1:02:02  lr: 0.020000  loss: 0.6404 (0.6636)  loss_classifier: 0.2264 (0.2332)  loss_box_reg: 0.2990 (0.2997)  loss_objectness: 0.0546 (0.0697)  loss_rpn_box_reg: 0.0579 (0.0610)  time: 1.7034  data: 0.6822  max mem: 19870\n",
      "Epoch: [2]  [ 260/2444]  eta: 1:01:44  lr: 0.020000  loss: 0.6512 (0.6629)  loss_classifier: 0.2331 (0.2330)  loss_box_reg: 0.2999 (0.2995)  loss_objectness: 0.0597 (0.0697)  loss_rpn_box_reg: 0.0578 (0.0608)  time: 1.7009  data: 0.6805  max mem: 19870\n",
      "Epoch: [2]  [ 270/2444]  eta: 1:01:29  lr: 0.020000  loss: 0.6603 (0.6643)  loss_classifier: 0.2334 (0.2334)  loss_box_reg: 0.2954 (0.2999)  loss_objectness: 0.0631 (0.0701)  loss_rpn_box_reg: 0.0595 (0.0609)  time: 1.7050  data: 0.6840  max mem: 19870\n",
      "Epoch: [2]  [ 280/2444]  eta: 1:01:13  lr: 0.020000  loss: 0.6673 (0.6652)  loss_classifier: 0.2357 (0.2335)  loss_box_reg: 0.2954 (0.3002)  loss_objectness: 0.0759 (0.0705)  loss_rpn_box_reg: 0.0595 (0.0610)  time: 1.7154  data: 0.6937  max mem: 19870\n",
      "Epoch: [2]  [ 290/2444]  eta: 1:00:56  lr: 0.020000  loss: 0.6664 (0.6649)  loss_classifier: 0.2252 (0.2334)  loss_box_reg: 0.2959 (0.3001)  loss_objectness: 0.0755 (0.0706)  loss_rpn_box_reg: 0.0546 (0.0608)  time: 1.7028  data: 0.6823  max mem: 19870\n",
      "Epoch: [2]  [ 300/2444]  eta: 1:00:39  lr: 0.020000  loss: 0.6682 (0.6647)  loss_classifier: 0.2296 (0.2332)  loss_box_reg: 0.3047 (0.3003)  loss_objectness: 0.0659 (0.0703)  loss_rpn_box_reg: 0.0587 (0.0609)  time: 1.6954  data: 0.6758  max mem: 19870\n",
      "Epoch: [2]  [ 310/2444]  eta: 1:00:22  lr: 0.020000  loss: 0.6581 (0.6646)  loss_classifier: 0.2327 (0.2331)  loss_box_reg: 0.3029 (0.3002)  loss_objectness: 0.0620 (0.0703)  loss_rpn_box_reg: 0.0609 (0.0609)  time: 1.6962  data: 0.6766  max mem: 19870\n",
      "Epoch: [2]  [ 320/2444]  eta: 1:00:05  lr: 0.020000  loss: 0.6444 (0.6643)  loss_classifier: 0.2313 (0.2332)  loss_box_reg: 0.2942 (0.3001)  loss_objectness: 0.0605 (0.0702)  loss_rpn_box_reg: 0.0579 (0.0608)  time: 1.6980  data: 0.6777  max mem: 19870\n",
      "Epoch: [2]  [ 330/2444]  eta: 0:59:49  lr: 0.020000  loss: 0.6444 (0.6646)  loss_classifier: 0.2299 (0.2333)  loss_box_reg: 0.2946 (0.3003)  loss_objectness: 0.0605 (0.0701)  loss_rpn_box_reg: 0.0606 (0.0609)  time: 1.7069  data: 0.6866  max mem: 19870\n",
      "Epoch: [2]  [ 340/2444]  eta: 0:59:32  lr: 0.020000  loss: 0.6634 (0.6644)  loss_classifier: 0.2290 (0.2332)  loss_box_reg: 0.3030 (0.3002)  loss_objectness: 0.0639 (0.0700)  loss_rpn_box_reg: 0.0629 (0.0610)  time: 1.7032  data: 0.6836  max mem: 19870\n",
      "Epoch: [2]  [ 350/2444]  eta: 0:59:15  lr: 0.020000  loss: 0.6634 (0.6645)  loss_classifier: 0.2311 (0.2334)  loss_box_reg: 0.3008 (0.3001)  loss_objectness: 0.0639 (0.0700)  loss_rpn_box_reg: 0.0603 (0.0610)  time: 1.6930  data: 0.6737  max mem: 19870\n",
      "Epoch: [2]  [ 360/2444]  eta: 0:58:58  lr: 0.020000  loss: 0.6528 (0.6647)  loss_classifier: 0.2304 (0.2334)  loss_box_reg: 0.2883 (0.3001)  loss_objectness: 0.0630 (0.0701)  loss_rpn_box_reg: 0.0567 (0.0611)  time: 1.6945  data: 0.6757  max mem: 19870\n",
      "Epoch: [2]  [ 370/2444]  eta: 0:58:41  lr: 0.020000  loss: 0.6569 (0.6658)  loss_classifier: 0.2272 (0.2337)  loss_box_reg: 0.2929 (0.3004)  loss_objectness: 0.0813 (0.0705)  loss_rpn_box_reg: 0.0612 (0.0612)  time: 1.7032  data: 0.6848  max mem: 19870\n",
      "Epoch: [2]  [ 380/2444]  eta: 0:58:24  lr: 0.020000  loss: 0.7406 (0.6671)  loss_classifier: 0.2560 (0.2341)  loss_box_reg: 0.3087 (0.3006)  loss_objectness: 0.0813 (0.0710)  loss_rpn_box_reg: 0.0695 (0.0614)  time: 1.7051  data: 0.6860  max mem: 19870\n",
      "Epoch: [2]  [ 390/2444]  eta: 0:58:07  lr: 0.020000  loss: 0.6730 (0.6668)  loss_classifier: 0.2349 (0.2341)  loss_box_reg: 0.3042 (0.3005)  loss_objectness: 0.0715 (0.0709)  loss_rpn_box_reg: 0.0608 (0.0613)  time: 1.6951  data: 0.6753  max mem: 19870\n",
      "Epoch: [2]  [ 400/2444]  eta: 0:57:50  lr: 0.020000  loss: 0.7005 (0.6682)  loss_classifier: 0.2432 (0.2345)  loss_box_reg: 0.3056 (0.3008)  loss_objectness: 0.0753 (0.0715)  loss_rpn_box_reg: 0.0570 (0.0614)  time: 1.6970  data: 0.6766  max mem: 19870\n",
      "Epoch: [2]  [ 410/2444]  eta: 0:57:33  lr: 0.020000  loss: 0.7005 (0.6680)  loss_classifier: 0.2432 (0.2346)  loss_box_reg: 0.3070 (0.3009)  loss_objectness: 0.0701 (0.0712)  loss_rpn_box_reg: 0.0564 (0.0613)  time: 1.6957  data: 0.6745  max mem: 19870\n",
      "Epoch: [2]  [ 420/2444]  eta: 0:57:16  lr: 0.020000  loss: 0.6502 (0.6678)  loss_classifier: 0.2271 (0.2344)  loss_box_reg: 0.2963 (0.3009)  loss_objectness: 0.0641 (0.0713)  loss_rpn_box_reg: 0.0549 (0.0612)  time: 1.6976  data: 0.6772  max mem: 19870\n",
      "Epoch: [2]  [ 430/2444]  eta: 0:57:00  lr: 0.020000  loss: 0.6595 (0.6682)  loss_classifier: 0.2338 (0.2345)  loss_box_reg: 0.2955 (0.3010)  loss_objectness: 0.0704 (0.0714)  loss_rpn_box_reg: 0.0586 (0.0612)  time: 1.7041  data: 0.6851  max mem: 19870\n",
      "Epoch: [2]  [ 440/2444]  eta: 0:56:42  lr: 0.020000  loss: 0.6241 (0.6676)  loss_classifier: 0.2326 (0.2344)  loss_box_reg: 0.2953 (0.3007)  loss_objectness: 0.0627 (0.0713)  loss_rpn_box_reg: 0.0562 (0.0611)  time: 1.6965  data: 0.6771  max mem: 19870\n",
      "Epoch: [2]  [ 450/2444]  eta: 0:56:25  lr: 0.020000  loss: 0.6241 (0.6669)  loss_classifier: 0.2259 (0.2341)  loss_box_reg: 0.2830 (0.3003)  loss_objectness: 0.0647 (0.0715)  loss_rpn_box_reg: 0.0561 (0.0610)  time: 1.6880  data: 0.6688  max mem: 19870\n",
      "Epoch: [2]  [ 460/2444]  eta: 0:56:07  lr: 0.020000  loss: 0.6466 (0.6670)  loss_classifier: 0.2261 (0.2342)  loss_box_reg: 0.2863 (0.3002)  loss_objectness: 0.0716 (0.0716)  loss_rpn_box_reg: 0.0561 (0.0610)  time: 1.6874  data: 0.6678  max mem: 19870\n",
      "Epoch: [2]  [ 470/2444]  eta: 0:55:50  lr: 0.020000  loss: 0.6600 (0.6668)  loss_classifier: 0.2261 (0.2340)  loss_box_reg: 0.2979 (0.3003)  loss_objectness: 0.0668 (0.0714)  loss_rpn_box_reg: 0.0614 (0.0610)  time: 1.6937  data: 0.6730  max mem: 19870\n",
      "Epoch: [2]  [ 480/2444]  eta: 0:55:33  lr: 0.020000  loss: 0.6701 (0.6673)  loss_classifier: 0.2313 (0.2341)  loss_box_reg: 0.3056 (0.3005)  loss_objectness: 0.0715 (0.0715)  loss_rpn_box_reg: 0.0637 (0.0613)  time: 1.6948  data: 0.6752  max mem: 19870\n",
      "Epoch: [2]  [ 490/2444]  eta: 0:55:17  lr: 0.020000  loss: 0.6719 (0.6674)  loss_classifier: 0.2393 (0.2341)  loss_box_reg: 0.2987 (0.3004)  loss_objectness: 0.0749 (0.0716)  loss_rpn_box_reg: 0.0644 (0.0613)  time: 1.7052  data: 0.6777  max mem: 19870\n",
      "Epoch: [2]  [ 500/2444]  eta: 0:55:00  lr: 0.020000  loss: 0.6582 (0.6674)  loss_classifier: 0.2379 (0.2341)  loss_box_reg: 0.2964 (0.3005)  loss_objectness: 0.0659 (0.0716)  loss_rpn_box_reg: 0.0565 (0.0612)  time: 1.7094  data: 0.6815  max mem: 19870\n",
      "Epoch: [2]  [ 510/2444]  eta: 0:54:43  lr: 0.020000  loss: 0.6644 (0.6674)  loss_classifier: 0.2299 (0.2339)  loss_box_reg: 0.2998 (0.3005)  loss_objectness: 0.0669 (0.0718)  loss_rpn_box_reg: 0.0565 (0.0611)  time: 1.6964  data: 0.6770  max mem: 19870\n",
      "Epoch: [2]  [ 520/2444]  eta: 0:54:26  lr: 0.020000  loss: 0.6661 (0.6678)  loss_classifier: 0.2294 (0.2340)  loss_box_reg: 0.2998 (0.3005)  loss_objectness: 0.0831 (0.0720)  loss_rpn_box_reg: 0.0618 (0.0612)  time: 1.6985  data: 0.6794  max mem: 19870\n",
      "Epoch: [2]  [ 530/2444]  eta: 0:54:09  lr: 0.020000  loss: 0.6611 (0.6674)  loss_classifier: 0.2288 (0.2339)  loss_box_reg: 0.2943 (0.3005)  loss_objectness: 0.0718 (0.0719)  loss_rpn_box_reg: 0.0561 (0.0611)  time: 1.6926  data: 0.6742  max mem: 19870\n",
      "Epoch: [2]  [ 540/2444]  eta: 0:53:52  lr: 0.020000  loss: 0.6715 (0.6679)  loss_classifier: 0.2286 (0.2340)  loss_box_reg: 0.2955 (0.3005)  loss_objectness: 0.0718 (0.0722)  loss_rpn_box_reg: 0.0561 (0.0612)  time: 1.6885  data: 0.6692  max mem: 19870\n",
      "Epoch: [2]  [ 550/2444]  eta: 0:53:34  lr: 0.020000  loss: 0.6791 (0.6679)  loss_classifier: 0.2352 (0.2340)  loss_box_reg: 0.2955 (0.3004)  loss_objectness: 0.0749 (0.0724)  loss_rpn_box_reg: 0.0596 (0.0611)  time: 1.6930  data: 0.6744  max mem: 19870\n",
      "Epoch: [2]  [ 560/2444]  eta: 0:53:17  lr: 0.020000  loss: 0.6622 (0.6680)  loss_classifier: 0.2368 (0.2341)  loss_box_reg: 0.2959 (0.3003)  loss_objectness: 0.0693 (0.0725)  loss_rpn_box_reg: 0.0596 (0.0612)  time: 1.6920  data: 0.6740  max mem: 19870\n",
      "Epoch: [2]  [ 570/2444]  eta: 0:53:00  lr: 0.020000  loss: 0.6579 (0.6680)  loss_classifier: 0.2280 (0.2340)  loss_box_reg: 0.2979 (0.3002)  loss_objectness: 0.0693 (0.0725)  loss_rpn_box_reg: 0.0658 (0.0612)  time: 1.6945  data: 0.6758  max mem: 19870\n",
      "Epoch: [2]  [ 580/2444]  eta: 0:52:43  lr: 0.020000  loss: 0.6502 (0.6677)  loss_classifier: 0.2254 (0.2339)  loss_box_reg: 0.2885 (0.3000)  loss_objectness: 0.0704 (0.0726)  loss_rpn_box_reg: 0.0613 (0.0612)  time: 1.6984  data: 0.6803  max mem: 19870\n",
      "Epoch: [2]  [ 590/2444]  eta: 0:52:27  lr: 0.020000  loss: 0.6541 (0.6681)  loss_classifier: 0.2295 (0.2341)  loss_box_reg: 0.2907 (0.3002)  loss_objectness: 0.0751 (0.0727)  loss_rpn_box_reg: 0.0588 (0.0612)  time: 1.7131  data: 0.6945  max mem: 19870\n",
      "Epoch: [2]  [ 600/2444]  eta: 0:52:10  lr: 0.020000  loss: 0.6694 (0.6681)  loss_classifier: 0.2295 (0.2340)  loss_box_reg: 0.2994 (0.3002)  loss_objectness: 0.0712 (0.0727)  loss_rpn_box_reg: 0.0607 (0.0612)  time: 1.7134  data: 0.6948  max mem: 19870\n",
      "Epoch: [2]  [ 610/2444]  eta: 0:51:54  lr: 0.020000  loss: 0.6473 (0.6682)  loss_classifier: 0.2317 (0.2341)  loss_box_reg: 0.2985 (0.3003)  loss_objectness: 0.0669 (0.0726)  loss_rpn_box_reg: 0.0617 (0.0612)  time: 1.7021  data: 0.6835  max mem: 19870\n",
      "Epoch: [2]  [ 620/2444]  eta: 0:51:37  lr: 0.020000  loss: 0.6643 (0.6688)  loss_classifier: 0.2391 (0.2342)  loss_box_reg: 0.3077 (0.3006)  loss_objectness: 0.0669 (0.0726)  loss_rpn_box_reg: 0.0635 (0.0613)  time: 1.7041  data: 0.6850  max mem: 19870\n",
      "Epoch: [2]  [ 630/2444]  eta: 0:51:20  lr: 0.020000  loss: 0.6594 (0.6686)  loss_classifier: 0.2276 (0.2342)  loss_box_reg: 0.3077 (0.3006)  loss_objectness: 0.0613 (0.0725)  loss_rpn_box_reg: 0.0614 (0.0613)  time: 1.7018  data: 0.6834  max mem: 19870\n",
      "Epoch: [2]  [ 640/2444]  eta: 0:51:03  lr: 0.020000  loss: 0.6401 (0.6681)  loss_classifier: 0.2275 (0.2340)  loss_box_reg: 0.3007 (0.3005)  loss_objectness: 0.0588 (0.0724)  loss_rpn_box_reg: 0.0588 (0.0613)  time: 1.6943  data: 0.6754  max mem: 19870\n",
      "Epoch: [2]  [ 650/2444]  eta: 0:50:46  lr: 0.020000  loss: 0.6501 (0.6683)  loss_classifier: 0.2236 (0.2340)  loss_box_reg: 0.3007 (0.3005)  loss_objectness: 0.0648 (0.0725)  loss_rpn_box_reg: 0.0598 (0.0613)  time: 1.6942  data: 0.6740  max mem: 19870\n",
      "Epoch: [2]  [ 660/2444]  eta: 0:50:29  lr: 0.020000  loss: 0.6931 (0.6684)  loss_classifier: 0.2354 (0.2340)  loss_box_reg: 0.3030 (0.3005)  loss_objectness: 0.0764 (0.0725)  loss_rpn_box_reg: 0.0581 (0.0613)  time: 1.6966  data: 0.6757  max mem: 19870\n",
      "Epoch: [2]  [ 670/2444]  eta: 0:50:12  lr: 0.020000  loss: 0.6464 (0.6688)  loss_classifier: 0.2327 (0.2341)  loss_box_reg: 0.3025 (0.3006)  loss_objectness: 0.0767 (0.0726)  loss_rpn_box_reg: 0.0599 (0.0614)  time: 1.7063  data: 0.6851  max mem: 19870\n",
      "Epoch: [2]  [ 680/2444]  eta: 0:49:55  lr: 0.020000  loss: 0.6868 (0.6693)  loss_classifier: 0.2342 (0.2343)  loss_box_reg: 0.3093 (0.3008)  loss_objectness: 0.0693 (0.0727)  loss_rpn_box_reg: 0.0665 (0.0615)  time: 1.7105  data: 0.6891  max mem: 19870\n",
      "Epoch: [2]  [ 690/2444]  eta: 0:49:38  lr: 0.020000  loss: 0.6971 (0.6694)  loss_classifier: 0.2382 (0.2343)  loss_box_reg: 0.3069 (0.3008)  loss_objectness: 0.0697 (0.0728)  loss_rpn_box_reg: 0.0614 (0.0615)  time: 1.7024  data: 0.6818  max mem: 19870\n",
      "Epoch: [2]  [ 700/2444]  eta: 0:49:21  lr: 0.020000  loss: 0.6369 (0.6690)  loss_classifier: 0.2216 (0.2342)  loss_box_reg: 0.2984 (0.3006)  loss_objectness: 0.0653 (0.0727)  loss_rpn_box_reg: 0.0591 (0.0615)  time: 1.6922  data: 0.6716  max mem: 19870\n",
      "Epoch: [2]  [ 710/2444]  eta: 0:49:04  lr: 0.020000  loss: 0.6306 (0.6691)  loss_classifier: 0.2216 (0.2341)  loss_box_reg: 0.2951 (0.3007)  loss_objectness: 0.0625 (0.0729)  loss_rpn_box_reg: 0.0598 (0.0615)  time: 1.6892  data: 0.6687  max mem: 19870\n",
      "Epoch: [2]  [ 720/2444]  eta: 0:48:47  lr: 0.020000  loss: 0.6750 (0.6694)  loss_classifier: 0.2413 (0.2342)  loss_box_reg: 0.3084 (0.3009)  loss_objectness: 0.0718 (0.0728)  loss_rpn_box_reg: 0.0587 (0.0615)  time: 1.6937  data: 0.6737  max mem: 19870\n",
      "Epoch: [2]  [ 730/2444]  eta: 0:48:31  lr: 0.020000  loss: 0.6750 (0.6697)  loss_classifier: 0.2423 (0.2343)  loss_box_reg: 0.3164 (0.3010)  loss_objectness: 0.0678 (0.0728)  loss_rpn_box_reg: 0.0598 (0.0615)  time: 1.7093  data: 0.6876  max mem: 19870\n",
      "Epoch: [2]  [ 740/2444]  eta: 0:48:13  lr: 0.020000  loss: 0.6621 (0.6694)  loss_classifier: 0.2285 (0.2342)  loss_box_reg: 0.2931 (0.3009)  loss_objectness: 0.0626 (0.0728)  loss_rpn_box_reg: 0.0598 (0.0615)  time: 1.7064  data: 0.6844  max mem: 19870\n",
      "Epoch: [2]  [ 750/2444]  eta: 0:47:56  lr: 0.020000  loss: 0.6783 (0.6701)  loss_classifier: 0.2366 (0.2344)  loss_box_reg: 0.2994 (0.3011)  loss_objectness: 0.0688 (0.0730)  loss_rpn_box_reg: 0.0607 (0.0616)  time: 1.6952  data: 0.6746  max mem: 19870\n",
      "Epoch: [2]  [ 760/2444]  eta: 0:47:39  lr: 0.020000  loss: 0.7058 (0.6699)  loss_classifier: 0.2409 (0.2344)  loss_box_reg: 0.3048 (0.3010)  loss_objectness: 0.0683 (0.0728)  loss_rpn_box_reg: 0.0640 (0.0616)  time: 1.6967  data: 0.6774  max mem: 19870\n",
      "Epoch: [2]  [ 770/2444]  eta: 0:47:22  lr: 0.020000  loss: 0.6679 (0.6700)  loss_classifier: 0.2376 (0.2345)  loss_box_reg: 0.3006 (0.3010)  loss_objectness: 0.0683 (0.0729)  loss_rpn_box_reg: 0.0640 (0.0617)  time: 1.6946  data: 0.6741  max mem: 19870\n",
      "Epoch: [2]  [ 780/2444]  eta: 0:47:05  lr: 0.020000  loss: 0.6722 (0.6703)  loss_classifier: 0.2389 (0.2346)  loss_box_reg: 0.3055 (0.3012)  loss_objectness: 0.0731 (0.0728)  loss_rpn_box_reg: 0.0620 (0.0616)  time: 1.6994  data: 0.6785  max mem: 19870\n",
      "Epoch: [2]  [ 790/2444]  eta: 0:46:48  lr: 0.020000  loss: 0.6773 (0.6708)  loss_classifier: 0.2389 (0.2348)  loss_box_reg: 0.3173 (0.3014)  loss_objectness: 0.0737 (0.0730)  loss_rpn_box_reg: 0.0607 (0.0617)  time: 1.7011  data: 0.6814  max mem: 19870\n",
      "Epoch: [2]  [ 800/2444]  eta: 0:46:31  lr: 0.020000  loss: 0.6737 (0.6706)  loss_classifier: 0.2325 (0.2347)  loss_box_reg: 0.2935 (0.3012)  loss_objectness: 0.0756 (0.0731)  loss_rpn_box_reg: 0.0604 (0.0617)  time: 1.6990  data: 0.6781  max mem: 19870\n",
      "Epoch: [2]  [ 810/2444]  eta: 0:46:14  lr: 0.020000  loss: 0.6383 (0.6703)  loss_classifier: 0.2274 (0.2347)  loss_box_reg: 0.2909 (0.3011)  loss_objectness: 0.0670 (0.0729)  loss_rpn_box_reg: 0.0565 (0.0615)  time: 1.6959  data: 0.6751  max mem: 19870\n",
      "Epoch: [2]  [ 820/2444]  eta: 0:45:57  lr: 0.020000  loss: 0.6440 (0.6703)  loss_classifier: 0.2264 (0.2347)  loss_box_reg: 0.2979 (0.3011)  loss_objectness: 0.0619 (0.0730)  loss_rpn_box_reg: 0.0532 (0.0615)  time: 1.6976  data: 0.6762  max mem: 19870\n",
      "Epoch: [2]  [ 830/2444]  eta: 0:45:40  lr: 0.020000  loss: 0.6440 (0.6700)  loss_classifier: 0.2271 (0.2346)  loss_box_reg: 0.2971 (0.3011)  loss_objectness: 0.0657 (0.0729)  loss_rpn_box_reg: 0.0531 (0.0614)  time: 1.6954  data: 0.6739  max mem: 19870\n",
      "Epoch: [2]  [ 840/2444]  eta: 0:45:23  lr: 0.020000  loss: 0.6848 (0.6706)  loss_classifier: 0.2354 (0.2347)  loss_box_reg: 0.3048 (0.3012)  loss_objectness: 0.0750 (0.0732)  loss_rpn_box_reg: 0.0638 (0.0615)  time: 1.6982  data: 0.6784  max mem: 19870\n",
      "Epoch: [2]  [ 850/2444]  eta: 0:45:06  lr: 0.020000  loss: 0.6848 (0.6705)  loss_classifier: 0.2356 (0.2346)  loss_box_reg: 0.3048 (0.3011)  loss_objectness: 0.0789 (0.0732)  loss_rpn_box_reg: 0.0657 (0.0615)  time: 1.6991  data: 0.6806  max mem: 19870\n",
      "Epoch: [2]  [ 860/2444]  eta: 0:44:49  lr: 0.020000  loss: 0.6570 (0.6704)  loss_classifier: 0.2345 (0.2346)  loss_box_reg: 0.2985 (0.3012)  loss_objectness: 0.0658 (0.0731)  loss_rpn_box_reg: 0.0592 (0.0615)  time: 1.6945  data: 0.6745  max mem: 19870\n",
      "Epoch: [2]  [ 870/2444]  eta: 0:44:33  lr: 0.020000  loss: 0.6773 (0.6708)  loss_classifier: 0.2432 (0.2347)  loss_box_reg: 0.3040 (0.3013)  loss_objectness: 0.0633 (0.0732)  loss_rpn_box_reg: 0.0626 (0.0616)  time: 1.7000  data: 0.6786  max mem: 19870\n",
      "Epoch: [2]  [ 880/2444]  eta: 0:44:16  lr: 0.020000  loss: 0.6865 (0.6712)  loss_classifier: 0.2453 (0.2348)  loss_box_reg: 0.3074 (0.3013)  loss_objectness: 0.0738 (0.0733)  loss_rpn_box_reg: 0.0655 (0.0617)  time: 1.7073  data: 0.6881  max mem: 19870\n",
      "Epoch: [2]  [ 890/2444]  eta: 0:43:59  lr: 0.020000  loss: 0.6842 (0.6713)  loss_classifier: 0.2353 (0.2349)  loss_box_reg: 0.3082 (0.3014)  loss_objectness: 0.0718 (0.0733)  loss_rpn_box_reg: 0.0600 (0.0617)  time: 1.7065  data: 0.6860  max mem: 19870\n",
      "Epoch: [2]  [ 900/2444]  eta: 0:43:42  lr: 0.020000  loss: 0.6856 (0.6712)  loss_classifier: 0.2287 (0.2348)  loss_box_reg: 0.3127 (0.3014)  loss_objectness: 0.0661 (0.0732)  loss_rpn_box_reg: 0.0603 (0.0617)  time: 1.7045  data: 0.6761  max mem: 19870\n",
      "Epoch: [2]  [ 910/2444]  eta: 0:43:25  lr: 0.020000  loss: 0.6616 (0.6716)  loss_classifier: 0.2287 (0.2349)  loss_box_reg: 0.3170 (0.3017)  loss_objectness: 0.0682 (0.0732)  loss_rpn_box_reg: 0.0618 (0.0618)  time: 1.7100  data: 0.6824  max mem: 19870\n",
      "Epoch: [2]  [ 920/2444]  eta: 0:43:08  lr: 0.020000  loss: 0.6616 (0.6716)  loss_classifier: 0.2381 (0.2349)  loss_box_reg: 0.3137 (0.3016)  loss_objectness: 0.0592 (0.0733)  loss_rpn_box_reg: 0.0582 (0.0618)  time: 1.7003  data: 0.6809  max mem: 19870\n",
      "Epoch: [2]  [ 930/2444]  eta: 0:42:51  lr: 0.020000  loss: 0.6340 (0.6714)  loss_classifier: 0.2255 (0.2349)  loss_box_reg: 0.2887 (0.3015)  loss_objectness: 0.0602 (0.0732)  loss_rpn_box_reg: 0.0578 (0.0618)  time: 1.6977  data: 0.6782  max mem: 19870\n",
      "Epoch: [2]  [ 940/2444]  eta: 0:42:34  lr: 0.020000  loss: 0.6438 (0.6713)  loss_classifier: 0.2265 (0.2348)  loss_box_reg: 0.2895 (0.3015)  loss_objectness: 0.0666 (0.0732)  loss_rpn_box_reg: 0.0556 (0.0618)  time: 1.7018  data: 0.6798  max mem: 19870\n",
      "Epoch: [2]  [ 950/2444]  eta: 0:42:17  lr: 0.020000  loss: 0.6564 (0.6712)  loss_classifier: 0.2315 (0.2349)  loss_box_reg: 0.2952 (0.3014)  loss_objectness: 0.0600 (0.0732)  loss_rpn_box_reg: 0.0544 (0.0617)  time: 1.7007  data: 0.6784  max mem: 19870\n",
      "Epoch: [2]  [ 960/2444]  eta: 0:42:00  lr: 0.020000  loss: 0.6588 (0.6714)  loss_classifier: 0.2350 (0.2349)  loss_box_reg: 0.2996 (0.3014)  loss_objectness: 0.0630 (0.0733)  loss_rpn_box_reg: 0.0652 (0.0618)  time: 1.7032  data: 0.6830  max mem: 19870\n",
      "Epoch: [2]  [ 970/2444]  eta: 0:41:43  lr: 0.020000  loss: 0.6979 (0.6716)  loss_classifier: 0.2406 (0.2349)  loss_box_reg: 0.3120 (0.3015)  loss_objectness: 0.0723 (0.0733)  loss_rpn_box_reg: 0.0680 (0.0619)  time: 1.6980  data: 0.6788  max mem: 19870\n",
      "Epoch: [2]  [ 980/2444]  eta: 0:41:26  lr: 0.020000  loss: 0.6771 (0.6715)  loss_classifier: 0.2406 (0.2349)  loss_box_reg: 0.2982 (0.3014)  loss_objectness: 0.0712 (0.0733)  loss_rpn_box_reg: 0.0592 (0.0618)  time: 1.6927  data: 0.6730  max mem: 19870\n",
      "Epoch: [2]  [ 990/2444]  eta: 0:41:09  lr: 0.020000  loss: 0.6771 (0.6718)  loss_classifier: 0.2415 (0.2350)  loss_box_reg: 0.2980 (0.3015)  loss_objectness: 0.0712 (0.0735)  loss_rpn_box_reg: 0.0582 (0.0618)  time: 1.6981  data: 0.6787  max mem: 19870\n",
      "Epoch: [2]  [1000/2444]  eta: 0:40:52  lr: 0.020000  loss: 0.6830 (0.6719)  loss_classifier: 0.2430 (0.2351)  loss_box_reg: 0.3068 (0.3016)  loss_objectness: 0.0620 (0.0735)  loss_rpn_box_reg: 0.0568 (0.0618)  time: 1.7063  data: 0.6866  max mem: 19870\n",
      "Epoch: [2]  [1010/2444]  eta: 0:40:35  lr: 0.020000  loss: 0.6824 (0.6721)  loss_classifier: 0.2416 (0.2351)  loss_box_reg: 0.3022 (0.3016)  loss_objectness: 0.0675 (0.0735)  loss_rpn_box_reg: 0.0599 (0.0619)  time: 1.7025  data: 0.6832  max mem: 19870\n",
      "Epoch: [2]  [1020/2444]  eta: 0:40:19  lr: 0.020000  loss: 0.6893 (0.6722)  loss_classifier: 0.2428 (0.2351)  loss_box_reg: 0.2951 (0.3017)  loss_objectness: 0.0726 (0.0735)  loss_rpn_box_reg: 0.0611 (0.0619)  time: 1.6988  data: 0.6801  max mem: 19870\n",
      "Epoch: [2]  [1030/2444]  eta: 0:40:01  lr: 0.020000  loss: 0.6674 (0.6720)  loss_classifier: 0.2291 (0.2350)  loss_box_reg: 0.2966 (0.3016)  loss_objectness: 0.0701 (0.0735)  loss_rpn_box_reg: 0.0640 (0.0619)  time: 1.6981  data: 0.6773  max mem: 19870\n",
      "Epoch: [2]  [1040/2444]  eta: 0:39:45  lr: 0.020000  loss: 0.6674 (0.6723)  loss_classifier: 0.2300 (0.2351)  loss_box_reg: 0.3054 (0.3017)  loss_objectness: 0.0701 (0.0736)  loss_rpn_box_reg: 0.0640 (0.0619)  time: 1.7080  data: 0.6878  max mem: 19870\n",
      "Epoch: [2]  [1050/2444]  eta: 0:39:28  lr: 0.020000  loss: 0.6657 (0.6722)  loss_classifier: 0.2389 (0.2351)  loss_box_reg: 0.2958 (0.3016)  loss_objectness: 0.0717 (0.0736)  loss_rpn_box_reg: 0.0640 (0.0619)  time: 1.7061  data: 0.6874  max mem: 19870\n",
      "Epoch: [2]  [1060/2444]  eta: 0:39:11  lr: 0.020000  loss: 0.6496 (0.6719)  loss_classifier: 0.2287 (0.2350)  loss_box_reg: 0.2947 (0.3016)  loss_objectness: 0.0669 (0.0735)  loss_rpn_box_reg: 0.0551 (0.0618)  time: 1.6955  data: 0.6761  max mem: 19870\n",
      "Epoch: [2]  [1070/2444]  eta: 0:38:54  lr: 0.020000  loss: 0.6590 (0.6721)  loss_classifier: 0.2352 (0.2351)  loss_box_reg: 0.3013 (0.3016)  loss_objectness: 0.0727 (0.0736)  loss_rpn_box_reg: 0.0586 (0.0619)  time: 1.6990  data: 0.6783  max mem: 19870\n",
      "Epoch: [2]  [1080/2444]  eta: 0:38:37  lr: 0.020000  loss: 0.6879 (0.6720)  loss_classifier: 0.2447 (0.2350)  loss_box_reg: 0.3062 (0.3016)  loss_objectness: 0.0735 (0.0735)  loss_rpn_box_reg: 0.0632 (0.0619)  time: 1.6984  data: 0.6789  max mem: 19870\n",
      "Epoch: [2]  [1090/2444]  eta: 0:38:19  lr: 0.020000  loss: 0.6702 (0.6720)  loss_classifier: 0.2385 (0.2350)  loss_box_reg: 0.2990 (0.3016)  loss_objectness: 0.0644 (0.0736)  loss_rpn_box_reg: 0.0612 (0.0619)  time: 1.6877  data: 0.6698  max mem: 19870\n",
      "Epoch: [2]  [1100/2444]  eta: 0:38:02  lr: 0.020000  loss: 0.6326 (0.6716)  loss_classifier: 0.2177 (0.2348)  loss_box_reg: 0.2839 (0.3014)  loss_objectness: 0.0624 (0.0735)  loss_rpn_box_reg: 0.0549 (0.0618)  time: 1.6798  data: 0.6611  max mem: 19870\n",
      "Epoch: [2]  [1110/2444]  eta: 0:37:46  lr: 0.020000  loss: 0.6454 (0.6718)  loss_classifier: 0.2257 (0.2349)  loss_box_reg: 0.2956 (0.3016)  loss_objectness: 0.0701 (0.0735)  loss_rpn_box_reg: 0.0552 (0.0619)  time: 1.6982  data: 0.6730  max mem: 19870\n",
      "Epoch: [2]  [1120/2444]  eta: 0:37:28  lr: 0.020000  loss: 0.6750 (0.6719)  loss_classifier: 0.2292 (0.2349)  loss_box_reg: 0.3118 (0.3016)  loss_objectness: 0.0729 (0.0735)  loss_rpn_box_reg: 0.0628 (0.0619)  time: 1.7019  data: 0.6770  max mem: 19870\n",
      "Epoch: [2]  [1130/2444]  eta: 0:37:11  lr: 0.020000  loss: 0.6687 (0.6720)  loss_classifier: 0.2293 (0.2350)  loss_box_reg: 0.2981 (0.3016)  loss_objectness: 0.0700 (0.0735)  loss_rpn_box_reg: 0.0618 (0.0619)  time: 1.6933  data: 0.6740  max mem: 19870\n",
      "Epoch: [2]  [1140/2444]  eta: 0:36:54  lr: 0.020000  loss: 0.6687 (0.6720)  loss_classifier: 0.2340 (0.2350)  loss_box_reg: 0.2941 (0.3016)  loss_objectness: 0.0700 (0.0735)  loss_rpn_box_reg: 0.0618 (0.0619)  time: 1.6981  data: 0.6781  max mem: 19870\n",
      "Epoch: [2]  [1150/2444]  eta: 0:36:37  lr: 0.020000  loss: 0.6474 (0.6719)  loss_classifier: 0.2340 (0.2349)  loss_box_reg: 0.2941 (0.3016)  loss_objectness: 0.0672 (0.0735)  loss_rpn_box_reg: 0.0532 (0.0619)  time: 1.6914  data: 0.6718  max mem: 19870\n",
      "Epoch: [2]  [1160/2444]  eta: 0:36:20  lr: 0.020000  loss: 0.6474 (0.6718)  loss_classifier: 0.2245 (0.2349)  loss_box_reg: 0.2941 (0.3015)  loss_objectness: 0.0666 (0.0735)  loss_rpn_box_reg: 0.0524 (0.0618)  time: 1.6893  data: 0.6681  max mem: 19870\n",
      "Epoch: [2]  [1170/2444]  eta: 0:36:03  lr: 0.020000  loss: 0.6513 (0.6719)  loss_classifier: 0.2309 (0.2350)  loss_box_reg: 0.2944 (0.3016)  loss_objectness: 0.0646 (0.0735)  loss_rpn_box_reg: 0.0592 (0.0618)  time: 1.6983  data: 0.6773  max mem: 19870\n",
      "Epoch: [2]  [1180/2444]  eta: 0:35:46  lr: 0.020000  loss: 0.6564 (0.6720)  loss_classifier: 0.2345 (0.2351)  loss_box_reg: 0.2977 (0.3016)  loss_objectness: 0.0643 (0.0735)  loss_rpn_box_reg: 0.0592 (0.0619)  time: 1.7022  data: 0.6828  max mem: 19870\n",
      "Epoch: [2]  [1190/2444]  eta: 0:35:29  lr: 0.020000  loss: 0.6875 (0.6721)  loss_classifier: 0.2382 (0.2351)  loss_box_reg: 0.3038 (0.3017)  loss_objectness: 0.0643 (0.0735)  loss_rpn_box_reg: 0.0665 (0.0619)  time: 1.7030  data: 0.6832  max mem: 19870\n",
      "Epoch: [2]  [1200/2444]  eta: 0:35:12  lr: 0.020000  loss: 0.6497 (0.6718)  loss_classifier: 0.2290 (0.2350)  loss_box_reg: 0.2982 (0.3016)  loss_objectness: 0.0630 (0.0734)  loss_rpn_box_reg: 0.0583 (0.0618)  time: 1.6991  data: 0.6798  max mem: 19870\n",
      "Epoch: [2]  [1210/2444]  eta: 0:34:55  lr: 0.020000  loss: 0.6391 (0.6718)  loss_classifier: 0.2269 (0.2350)  loss_box_reg: 0.2967 (0.3017)  loss_objectness: 0.0605 (0.0733)  loss_rpn_box_reg: 0.0596 (0.0618)  time: 1.6913  data: 0.6708  max mem: 19870\n",
      "Epoch: [2]  [1220/2444]  eta: 0:34:38  lr: 0.020000  loss: 0.6373 (0.6717)  loss_classifier: 0.2214 (0.2349)  loss_box_reg: 0.2967 (0.3016)  loss_objectness: 0.0605 (0.0733)  loss_rpn_box_reg: 0.0596 (0.0618)  time: 1.6930  data: 0.6716  max mem: 19870\n",
      "Epoch: [2]  [1230/2444]  eta: 0:34:21  lr: 0.020000  loss: 0.6259 (0.6714)  loss_classifier: 0.2209 (0.2348)  loss_box_reg: 0.2823 (0.3015)  loss_objectness: 0.0688 (0.0733)  loss_rpn_box_reg: 0.0595 (0.0618)  time: 1.6900  data: 0.6698  max mem: 19870\n",
      "Epoch: [2]  [1240/2444]  eta: 0:34:04  lr: 0.020000  loss: 0.6562 (0.6713)  loss_classifier: 0.2240 (0.2347)  loss_box_reg: 0.2927 (0.3014)  loss_objectness: 0.0708 (0.0734)  loss_rpn_box_reg: 0.0604 (0.0618)  time: 1.6900  data: 0.6712  max mem: 19870\n",
      "Epoch: [2]  [1250/2444]  eta: 0:33:47  lr: 0.020000  loss: 0.6652 (0.6712)  loss_classifier: 0.2299 (0.2347)  loss_box_reg: 0.2927 (0.3013)  loss_objectness: 0.0736 (0.0734)  loss_rpn_box_reg: 0.0646 (0.0618)  time: 1.6998  data: 0.6809  max mem: 19870\n",
      "Epoch: [2]  [1260/2444]  eta: 0:33:30  lr: 0.020000  loss: 0.6868 (0.6714)  loss_classifier: 0.2407 (0.2348)  loss_box_reg: 0.3067 (0.3013)  loss_objectness: 0.0703 (0.0734)  loss_rpn_box_reg: 0.0648 (0.0619)  time: 1.7055  data: 0.6856  max mem: 19870\n",
      "Epoch: [2]  [1270/2444]  eta: 0:33:13  lr: 0.020000  loss: 0.6677 (0.6709)  loss_classifier: 0.2261 (0.2346)  loss_box_reg: 0.2965 (0.3012)  loss_objectness: 0.0580 (0.0733)  loss_rpn_box_reg: 0.0598 (0.0618)  time: 1.7041  data: 0.6843  max mem: 19870\n",
      "Epoch: [2]  [1280/2444]  eta: 0:32:56  lr: 0.020000  loss: 0.6465 (0.6710)  loss_classifier: 0.2261 (0.2346)  loss_box_reg: 0.2965 (0.3012)  loss_objectness: 0.0577 (0.0733)  loss_rpn_box_reg: 0.0552 (0.0618)  time: 1.7009  data: 0.6816  max mem: 19870\n",
      "Epoch: [2]  [1290/2444]  eta: 0:32:39  lr: 0.020000  loss: 0.6748 (0.6708)  loss_classifier: 0.2299 (0.2345)  loss_box_reg: 0.3013 (0.3011)  loss_objectness: 0.0681 (0.0733)  loss_rpn_box_reg: 0.0618 (0.0618)  time: 1.6985  data: 0.6795  max mem: 19870\n",
      "Epoch: [2]  [1300/2444]  eta: 0:32:23  lr: 0.020000  loss: 0.6748 (0.6709)  loss_classifier: 0.2268 (0.2345)  loss_box_reg: 0.2940 (0.3011)  loss_objectness: 0.0783 (0.0734)  loss_rpn_box_reg: 0.0671 (0.0619)  time: 1.7011  data: 0.6820  max mem: 19870\n",
      "Epoch: [2]  [1310/2444]  eta: 0:32:06  lr: 0.020000  loss: 0.6583 (0.6707)  loss_classifier: 0.2233 (0.2344)  loss_box_reg: 0.2940 (0.3010)  loss_objectness: 0.0816 (0.0735)  loss_rpn_box_reg: 0.0627 (0.0618)  time: 1.7060  data: 0.6806  max mem: 19870\n",
      "Epoch: [2]  [1320/2444]  eta: 0:31:48  lr: 0.020000  loss: 0.6381 (0.6706)  loss_classifier: 0.2233 (0.2343)  loss_box_reg: 0.2893 (0.3009)  loss_objectness: 0.0732 (0.0736)  loss_rpn_box_reg: 0.0565 (0.0618)  time: 1.6943  data: 0.6685  max mem: 19870\n",
      "Epoch: [2]  [1330/2444]  eta: 0:31:31  lr: 0.020000  loss: 0.6381 (0.6704)  loss_classifier: 0.2314 (0.2343)  loss_box_reg: 0.2923 (0.3009)  loss_objectness: 0.0615 (0.0735)  loss_rpn_box_reg: 0.0541 (0.0618)  time: 1.6862  data: 0.6671  max mem: 19870\n",
      "Epoch: [2]  [1340/2444]  eta: 0:31:14  lr: 0.020000  loss: 0.6684 (0.6706)  loss_classifier: 0.2349 (0.2343)  loss_box_reg: 0.2923 (0.3008)  loss_objectness: 0.0774 (0.0736)  loss_rpn_box_reg: 0.0582 (0.0618)  time: 1.6894  data: 0.6715  max mem: 19870\n",
      "Epoch: [2]  [1350/2444]  eta: 0:30:57  lr: 0.020000  loss: 0.6800 (0.6706)  loss_classifier: 0.2358 (0.2343)  loss_box_reg: 0.2921 (0.3009)  loss_objectness: 0.0774 (0.0736)  loss_rpn_box_reg: 0.0603 (0.0618)  time: 1.6959  data: 0.6777  max mem: 19870\n",
      "Epoch: [2]  [1360/2444]  eta: 0:30:40  lr: 0.020000  loss: 0.6800 (0.6707)  loss_classifier: 0.2375 (0.2344)  loss_box_reg: 0.3041 (0.3010)  loss_objectness: 0.0713 (0.0736)  loss_rpn_box_reg: 0.0603 (0.0618)  time: 1.7036  data: 0.6847  max mem: 19870\n",
      "Epoch: [2]  [1370/2444]  eta: 0:30:24  lr: 0.020000  loss: 0.6762 (0.6709)  loss_classifier: 0.2290 (0.2344)  loss_box_reg: 0.3064 (0.3010)  loss_objectness: 0.0633 (0.0737)  loss_rpn_box_reg: 0.0657 (0.0618)  time: 1.7041  data: 0.6845  max mem: 19870\n",
      "Epoch: [2]  [1380/2444]  eta: 0:30:07  lr: 0.020000  loss: 0.6801 (0.6709)  loss_classifier: 0.2304 (0.2344)  loss_box_reg: 0.3052 (0.3010)  loss_objectness: 0.0683 (0.0737)  loss_rpn_box_reg: 0.0657 (0.0619)  time: 1.7025  data: 0.6822  max mem: 19870\n",
      "Epoch: [2]  [1390/2444]  eta: 0:29:50  lr: 0.020000  loss: 0.6474 (0.6707)  loss_classifier: 0.2235 (0.2343)  loss_box_reg: 0.2934 (0.3009)  loss_objectness: 0.0632 (0.0737)  loss_rpn_box_reg: 0.0568 (0.0618)  time: 1.6981  data: 0.6778  max mem: 19870\n",
      "Epoch: [2]  [1400/2444]  eta: 0:29:33  lr: 0.020000  loss: 0.6256 (0.6703)  loss_classifier: 0.2212 (0.2341)  loss_box_reg: 0.2845 (0.3008)  loss_objectness: 0.0550 (0.0736)  loss_rpn_box_reg: 0.0532 (0.0618)  time: 1.6916  data: 0.6715  max mem: 19870\n",
      "Epoch: [2]  [1410/2444]  eta: 0:29:16  lr: 0.020000  loss: 0.6154 (0.6703)  loss_classifier: 0.2218 (0.2342)  loss_box_reg: 0.2887 (0.3008)  loss_objectness: 0.0621 (0.0736)  loss_rpn_box_reg: 0.0583 (0.0618)  time: 1.6986  data: 0.6787  max mem: 19870\n",
      "Epoch: [2]  [1420/2444]  eta: 0:28:59  lr: 0.020000  loss: 0.6477 (0.6704)  loss_classifier: 0.2270 (0.2342)  loss_box_reg: 0.2922 (0.3008)  loss_objectness: 0.0790 (0.0736)  loss_rpn_box_reg: 0.0623 (0.0619)  time: 1.7065  data: 0.6853  max mem: 19870\n",
      "Epoch: [2]  [1430/2444]  eta: 0:28:42  lr: 0.020000  loss: 0.6477 (0.6703)  loss_classifier: 0.2260 (0.2341)  loss_box_reg: 0.2966 (0.3007)  loss_objectness: 0.0789 (0.0736)  loss_rpn_box_reg: 0.0607 (0.0618)  time: 1.6997  data: 0.6784  max mem: 19870\n",
      "Epoch: [2]  [1440/2444]  eta: 0:28:25  lr: 0.020000  loss: 0.6451 (0.6702)  loss_classifier: 0.2260 (0.2341)  loss_box_reg: 0.2854 (0.3006)  loss_objectness: 0.0608 (0.0737)  loss_rpn_box_reg: 0.0535 (0.0618)  time: 1.6946  data: 0.6745  max mem: 19870\n",
      "Epoch: [2]  [1450/2444]  eta: 0:28:08  lr: 0.020000  loss: 0.6451 (0.6702)  loss_classifier: 0.2225 (0.2341)  loss_box_reg: 0.2854 (0.3006)  loss_objectness: 0.0667 (0.0737)  loss_rpn_box_reg: 0.0595 (0.0618)  time: 1.6993  data: 0.6789  max mem: 19870\n",
      "Epoch: [2]  [1460/2444]  eta: 0:27:51  lr: 0.020000  loss: 0.6383 (0.6698)  loss_classifier: 0.2186 (0.2339)  loss_box_reg: 0.2877 (0.3005)  loss_objectness: 0.0652 (0.0737)  loss_rpn_box_reg: 0.0571 (0.0617)  time: 1.6887  data: 0.6692  max mem: 19870\n",
      "Epoch: [2]  [1470/2444]  eta: 0:27:33  lr: 0.020000  loss: 0.6223 (0.6695)  loss_classifier: 0.2139 (0.2338)  loss_box_reg: 0.2877 (0.3004)  loss_objectness: 0.0614 (0.0736)  loss_rpn_box_reg: 0.0526 (0.0617)  time: 1.6779  data: 0.6601  max mem: 19870\n",
      "Epoch: [2]  [1480/2444]  eta: 0:27:16  lr: 0.020000  loss: 0.6285 (0.6692)  loss_classifier: 0.2131 (0.2337)  loss_box_reg: 0.2874 (0.3002)  loss_objectness: 0.0571 (0.0736)  loss_rpn_box_reg: 0.0557 (0.0617)  time: 1.6901  data: 0.6709  max mem: 19870\n",
      "Epoch: [2]  [1490/2444]  eta: 0:26:59  lr: 0.020000  loss: 0.6130 (0.6688)  loss_classifier: 0.2161 (0.2336)  loss_box_reg: 0.2776 (0.3001)  loss_objectness: 0.0567 (0.0735)  loss_rpn_box_reg: 0.0557 (0.0616)  time: 1.6905  data: 0.6707  max mem: 19870\n",
      "Epoch: [2]  [1500/2444]  eta: 0:26:42  lr: 0.020000  loss: 0.6505 (0.6690)  loss_classifier: 0.2278 (0.2336)  loss_box_reg: 0.2955 (0.3002)  loss_objectness: 0.0622 (0.0736)  loss_rpn_box_reg: 0.0608 (0.0616)  time: 1.6910  data: 0.6714  max mem: 19870\n",
      "Epoch: [2]  [1510/2444]  eta: 0:26:25  lr: 0.020000  loss: 0.6231 (0.6687)  loss_classifier: 0.2310 (0.2335)  loss_box_reg: 0.2930 (0.3000)  loss_objectness: 0.0686 (0.0735)  loss_rpn_box_reg: 0.0598 (0.0616)  time: 1.6881  data: 0.6689  max mem: 19870\n",
      "Epoch: [2]  [1520/2444]  eta: 0:26:08  lr: 0.020000  loss: 0.6194 (0.6684)  loss_classifier: 0.2162 (0.2334)  loss_box_reg: 0.2811 (0.2999)  loss_objectness: 0.0714 (0.0735)  loss_rpn_box_reg: 0.0539 (0.0616)  time: 1.6951  data: 0.6709  max mem: 19870\n",
      "Epoch: [2]  [1530/2444]  eta: 0:25:51  lr: 0.020000  loss: 0.6303 (0.6683)  loss_classifier: 0.2285 (0.2334)  loss_box_reg: 0.2929 (0.2999)  loss_objectness: 0.0595 (0.0734)  loss_rpn_box_reg: 0.0512 (0.0616)  time: 1.7070  data: 0.6816  max mem: 19870\n",
      "Epoch: [2]  [1540/2444]  eta: 0:25:34  lr: 0.020000  loss: 0.6309 (0.6682)  loss_classifier: 0.2286 (0.2334)  loss_box_reg: 0.2957 (0.2998)  loss_objectness: 0.0506 (0.0734)  loss_rpn_box_reg: 0.0545 (0.0616)  time: 1.6934  data: 0.6739  max mem: 19870\n",
      "Epoch: [2]  [1550/2444]  eta: 0:25:17  lr: 0.020000  loss: 0.6587 (0.6683)  loss_classifier: 0.2286 (0.2335)  loss_box_reg: 0.2877 (0.2998)  loss_objectness: 0.0660 (0.0734)  loss_rpn_box_reg: 0.0596 (0.0616)  time: 1.6917  data: 0.6708  max mem: 19870\n",
      "Epoch: [2]  [1560/2444]  eta: 0:25:00  lr: 0.020000  loss: 0.6767 (0.6684)  loss_classifier: 0.2392 (0.2335)  loss_box_reg: 0.3110 (0.2999)  loss_objectness: 0.0689 (0.0733)  loss_rpn_box_reg: 0.0585 (0.0616)  time: 1.6985  data: 0.6774  max mem: 19870\n",
      "Epoch: [2]  [1570/2444]  eta: 0:24:43  lr: 0.020000  loss: 0.6577 (0.6682)  loss_classifier: 0.2243 (0.2334)  loss_box_reg: 0.3088 (0.2999)  loss_objectness: 0.0689 (0.0733)  loss_rpn_box_reg: 0.0585 (0.0616)  time: 1.6950  data: 0.6757  max mem: 19870\n",
      "Epoch: [2]  [1580/2444]  eta: 0:24:26  lr: 0.020000  loss: 0.6575 (0.6682)  loss_classifier: 0.2192 (0.2334)  loss_box_reg: 0.3004 (0.2999)  loss_objectness: 0.0698 (0.0734)  loss_rpn_box_reg: 0.0593 (0.0616)  time: 1.6898  data: 0.6709  max mem: 19870\n",
      "Epoch: [2]  [1590/2444]  eta: 0:24:09  lr: 0.020000  loss: 0.6744 (0.6684)  loss_classifier: 0.2332 (0.2335)  loss_box_reg: 0.3026 (0.2999)  loss_objectness: 0.0741 (0.0734)  loss_rpn_box_reg: 0.0688 (0.0616)  time: 1.6993  data: 0.6794  max mem: 19870\n",
      "Epoch: [2]  [1600/2444]  eta: 0:23:53  lr: 0.020000  loss: 0.6827 (0.6685)  loss_classifier: 0.2312 (0.2334)  loss_box_reg: 0.3048 (0.2999)  loss_objectness: 0.0807 (0.0735)  loss_rpn_box_reg: 0.0661 (0.0617)  time: 1.7050  data: 0.6850  max mem: 19870\n",
      "Epoch: [2]  [1610/2444]  eta: 0:23:36  lr: 0.020000  loss: 0.6816 (0.6686)  loss_classifier: 0.2353 (0.2335)  loss_box_reg: 0.3048 (0.2999)  loss_objectness: 0.0725 (0.0734)  loss_rpn_box_reg: 0.0592 (0.0617)  time: 1.6997  data: 0.6808  max mem: 19870\n",
      "Epoch: [2]  [1620/2444]  eta: 0:23:19  lr: 0.020000  loss: 0.6688 (0.6685)  loss_classifier: 0.2353 (0.2335)  loss_box_reg: 0.3044 (0.2999)  loss_objectness: 0.0725 (0.0735)  loss_rpn_box_reg: 0.0570 (0.0616)  time: 1.6930  data: 0.6731  max mem: 19870\n",
      "Epoch: [2]  [1630/2444]  eta: 0:23:01  lr: 0.020000  loss: 0.6477 (0.6684)  loss_classifier: 0.2294 (0.2334)  loss_box_reg: 0.2780 (0.2998)  loss_objectness: 0.0716 (0.0736)  loss_rpn_box_reg: 0.0558 (0.0616)  time: 1.6859  data: 0.6654  max mem: 19870\n",
      "Epoch: [2]  [1640/2444]  eta: 0:22:45  lr: 0.020000  loss: 0.6593 (0.6685)  loss_classifier: 0.2323 (0.2334)  loss_box_reg: 0.2936 (0.2998)  loss_objectness: 0.0725 (0.0736)  loss_rpn_box_reg: 0.0629 (0.0616)  time: 1.6941  data: 0.6745  max mem: 19870\n",
      "Epoch: [2]  [1650/2444]  eta: 0:22:28  lr: 0.020000  loss: 0.6579 (0.6685)  loss_classifier: 0.2395 (0.2335)  loss_box_reg: 0.3042 (0.2999)  loss_objectness: 0.0633 (0.0735)  loss_rpn_box_reg: 0.0616 (0.0616)  time: 1.7010  data: 0.6825  max mem: 19870\n",
      "Epoch: [2]  [1660/2444]  eta: 0:22:11  lr: 0.020000  loss: 0.6572 (0.6686)  loss_classifier: 0.2372 (0.2335)  loss_box_reg: 0.3042 (0.2999)  loss_objectness: 0.0586 (0.0736)  loss_rpn_box_reg: 0.0644 (0.0617)  time: 1.7038  data: 0.6853  max mem: 19870\n",
      "Epoch: [2]  [1670/2444]  eta: 0:21:54  lr: 0.020000  loss: 0.6622 (0.6685)  loss_classifier: 0.2272 (0.2334)  loss_box_reg: 0.2972 (0.2999)  loss_objectness: 0.0792 (0.0736)  loss_rpn_box_reg: 0.0647 (0.0617)  time: 1.6993  data: 0.6808  max mem: 19870\n",
      "Epoch: [2]  [1680/2444]  eta: 0:21:37  lr: 0.020000  loss: 0.6234 (0.6683)  loss_classifier: 0.2166 (0.2334)  loss_box_reg: 0.2798 (0.2997)  loss_objectness: 0.0639 (0.0735)  loss_rpn_box_reg: 0.0537 (0.0617)  time: 1.6931  data: 0.6740  max mem: 19870\n",
      "Epoch: [2]  [1690/2444]  eta: 0:21:20  lr: 0.020000  loss: 0.6059 (0.6680)  loss_classifier: 0.2195 (0.2333)  loss_box_reg: 0.2783 (0.2996)  loss_objectness: 0.0630 (0.0735)  loss_rpn_box_reg: 0.0535 (0.0616)  time: 1.6881  data: 0.6685  max mem: 19870\n",
      "Epoch: [2]  [1700/2444]  eta: 0:21:03  lr: 0.020000  loss: 0.6251 (0.6680)  loss_classifier: 0.2241 (0.2333)  loss_box_reg: 0.2883 (0.2996)  loss_objectness: 0.0638 (0.0735)  loss_rpn_box_reg: 0.0560 (0.0616)  time: 1.6850  data: 0.6667  max mem: 19870\n",
      "Epoch: [2]  [1710/2444]  eta: 0:20:46  lr: 0.020000  loss: 0.6539 (0.6682)  loss_classifier: 0.2308 (0.2333)  loss_box_reg: 0.2971 (0.2996)  loss_objectness: 0.0678 (0.0736)  loss_rpn_box_reg: 0.0619 (0.0617)  time: 1.6983  data: 0.6789  max mem: 19870\n",
      "Epoch: [2]  [1720/2444]  eta: 0:20:29  lr: 0.020000  loss: 0.6441 (0.6679)  loss_classifier: 0.2239 (0.2332)  loss_box_reg: 0.2930 (0.2996)  loss_objectness: 0.0642 (0.0735)  loss_rpn_box_reg: 0.0614 (0.0616)  time: 1.6958  data: 0.6751  max mem: 19870\n",
      "Epoch: [2]  [1730/2444]  eta: 0:20:12  lr: 0.020000  loss: 0.6312 (0.6679)  loss_classifier: 0.2155 (0.2332)  loss_box_reg: 0.2873 (0.2996)  loss_objectness: 0.0642 (0.0735)  loss_rpn_box_reg: 0.0613 (0.0617)  time: 1.6924  data: 0.6727  max mem: 19870\n",
      "Epoch: [2]  [1740/2444]  eta: 0:19:55  lr: 0.020000  loss: 0.7042 (0.6680)  loss_classifier: 0.2396 (0.2332)  loss_box_reg: 0.2994 (0.2996)  loss_objectness: 0.0715 (0.0735)  loss_rpn_box_reg: 0.0644 (0.0617)  time: 1.6943  data: 0.6736  max mem: 19870\n",
      "Epoch: [2]  [1750/2444]  eta: 0:19:38  lr: 0.020000  loss: 0.6517 (0.6678)  loss_classifier: 0.2273 (0.2331)  loss_box_reg: 0.2973 (0.2995)  loss_objectness: 0.0715 (0.0735)  loss_rpn_box_reg: 0.0603 (0.0616)  time: 1.6856  data: 0.6637  max mem: 19870\n",
      "Epoch: [2]  [1760/2444]  eta: 0:19:21  lr: 0.020000  loss: 0.6517 (0.6679)  loss_classifier: 0.2273 (0.2331)  loss_box_reg: 0.2973 (0.2995)  loss_objectness: 0.0691 (0.0736)  loss_rpn_box_reg: 0.0579 (0.0616)  time: 1.6872  data: 0.6672  max mem: 19870\n",
      "Epoch: [2]  [1770/2444]  eta: 0:19:04  lr: 0.020000  loss: 0.6434 (0.6677)  loss_classifier: 0.2255 (0.2331)  loss_box_reg: 0.2933 (0.2994)  loss_objectness: 0.0642 (0.0735)  loss_rpn_box_reg: 0.0604 (0.0616)  time: 1.6902  data: 0.6718  max mem: 19870\n",
      "Epoch: [2]  [1780/2444]  eta: 0:18:47  lr: 0.020000  loss: 0.6551 (0.6677)  loss_classifier: 0.2248 (0.2331)  loss_box_reg: 0.2908 (0.2994)  loss_objectness: 0.0663 (0.0736)  loss_rpn_box_reg: 0.0608 (0.0616)  time: 1.6928  data: 0.6733  max mem: 19870\n",
      "Epoch: [2]  [1790/2444]  eta: 0:18:30  lr: 0.020000  loss: 0.6734 (0.6677)  loss_classifier: 0.2286 (0.2330)  loss_box_reg: 0.2929 (0.2994)  loss_objectness: 0.0785 (0.0736)  loss_rpn_box_reg: 0.0624 (0.0617)  time: 1.6995  data: 0.6780  max mem: 19870\n",
      "Epoch: [2]  [1800/2444]  eta: 0:18:13  lr: 0.020000  loss: 0.6453 (0.6676)  loss_classifier: 0.2256 (0.2330)  loss_box_reg: 0.2920 (0.2994)  loss_objectness: 0.0608 (0.0735)  loss_rpn_box_reg: 0.0605 (0.0617)  time: 1.6984  data: 0.6759  max mem: 19870\n",
      "Epoch: [2]  [1810/2444]  eta: 0:17:56  lr: 0.020000  loss: 0.6453 (0.6675)  loss_classifier: 0.2252 (0.2330)  loss_box_reg: 0.2807 (0.2993)  loss_objectness: 0.0731 (0.0736)  loss_rpn_box_reg: 0.0580 (0.0617)  time: 1.7006  data: 0.6773  max mem: 19870\n",
      "Epoch: [2]  [1820/2444]  eta: 0:17:39  lr: 0.020000  loss: 0.6442 (0.6674)  loss_classifier: 0.2297 (0.2329)  loss_box_reg: 0.2783 (0.2992)  loss_objectness: 0.0731 (0.0736)  loss_rpn_box_reg: 0.0588 (0.0617)  time: 1.6986  data: 0.6763  max mem: 19870\n",
      "Epoch: [2]  [1830/2444]  eta: 0:17:22  lr: 0.020000  loss: 0.6709 (0.6675)  loss_classifier: 0.2297 (0.2329)  loss_box_reg: 0.2955 (0.2993)  loss_objectness: 0.0768 (0.0737)  loss_rpn_box_reg: 0.0594 (0.0617)  time: 1.6948  data: 0.6740  max mem: 19870\n",
      "Epoch: [2]  [1840/2444]  eta: 0:17:05  lr: 0.020000  loss: 0.6320 (0.6672)  loss_classifier: 0.2217 (0.2328)  loss_box_reg: 0.2900 (0.2991)  loss_objectness: 0.0711 (0.0736)  loss_rpn_box_reg: 0.0567 (0.0616)  time: 1.6926  data: 0.6720  max mem: 19870\n",
      "Epoch: [2]  [1850/2444]  eta: 0:16:48  lr: 0.020000  loss: 0.6320 (0.6673)  loss_classifier: 0.2223 (0.2328)  loss_box_reg: 0.2788 (0.2991)  loss_objectness: 0.0690 (0.0737)  loss_rpn_box_reg: 0.0563 (0.0616)  time: 1.6880  data: 0.6674  max mem: 19870\n",
      "Epoch: [2]  [1860/2444]  eta: 0:16:31  lr: 0.020000  loss: 0.6655 (0.6672)  loss_classifier: 0.2307 (0.2328)  loss_box_reg: 0.2963 (0.2991)  loss_objectness: 0.0746 (0.0737)  loss_rpn_box_reg: 0.0563 (0.0616)  time: 1.6902  data: 0.6711  max mem: 19870\n",
      "Epoch: [2]  [1870/2444]  eta: 0:16:14  lr: 0.020000  loss: 0.6207 (0.6670)  loss_classifier: 0.2187 (0.2328)  loss_box_reg: 0.2879 (0.2991)  loss_objectness: 0.0610 (0.0736)  loss_rpn_box_reg: 0.0541 (0.0616)  time: 1.6900  data: 0.6710  max mem: 19870\n",
      "Epoch: [2]  [1880/2444]  eta: 0:15:57  lr: 0.020000  loss: 0.6314 (0.6672)  loss_classifier: 0.2304 (0.2328)  loss_box_reg: 0.2943 (0.2991)  loss_objectness: 0.0676 (0.0737)  loss_rpn_box_reg: 0.0564 (0.0616)  time: 1.6897  data: 0.6700  max mem: 19870\n",
      "Epoch: [2]  [1890/2444]  eta: 0:15:40  lr: 0.020000  loss: 0.6617 (0.6672)  loss_classifier: 0.2312 (0.2328)  loss_box_reg: 0.2995 (0.2991)  loss_objectness: 0.0701 (0.0737)  loss_rpn_box_reg: 0.0588 (0.0616)  time: 1.6971  data: 0.6769  max mem: 19870\n",
      "Epoch: [2]  [1900/2444]  eta: 0:15:23  lr: 0.020000  loss: 0.7149 (0.6675)  loss_classifier: 0.2444 (0.2329)  loss_box_reg: 0.3020 (0.2992)  loss_objectness: 0.0793 (0.0739)  loss_rpn_box_reg: 0.0625 (0.0616)  time: 1.7094  data: 0.6887  max mem: 19870\n",
      "Epoch: [2]  [1910/2444]  eta: 0:15:06  lr: 0.020000  loss: 0.7149 (0.6677)  loss_classifier: 0.2495 (0.2330)  loss_box_reg: 0.3122 (0.2993)  loss_objectness: 0.0783 (0.0739)  loss_rpn_box_reg: 0.0634 (0.0616)  time: 1.7176  data: 0.6965  max mem: 19870\n",
      "Epoch: [2]  [1920/2444]  eta: 0:14:49  lr: 0.020000  loss: 0.6772 (0.6677)  loss_classifier: 0.2394 (0.2330)  loss_box_reg: 0.3120 (0.2992)  loss_objectness: 0.0689 (0.0739)  loss_rpn_box_reg: 0.0596 (0.0616)  time: 1.7225  data: 0.7004  max mem: 19870\n",
      "Epoch: [2]  [1930/2444]  eta: 0:14:32  lr: 0.020000  loss: 0.6688 (0.6680)  loss_classifier: 0.2339 (0.2331)  loss_box_reg: 0.3059 (0.2993)  loss_objectness: 0.0739 (0.0739)  loss_rpn_box_reg: 0.0572 (0.0616)  time: 1.7327  data: 0.7107  max mem: 19870\n",
      "Epoch: [2]  [1940/2444]  eta: 0:14:15  lr: 0.020000  loss: 0.6888 (0.6680)  loss_classifier: 0.2333 (0.2331)  loss_box_reg: 0.2976 (0.2993)  loss_objectness: 0.0739 (0.0740)  loss_rpn_box_reg: 0.0567 (0.0616)  time: 1.7257  data: 0.7040  max mem: 19870\n",
      "Epoch: [2]  [1950/2444]  eta: 0:13:58  lr: 0.020000  loss: 0.6483 (0.6680)  loss_classifier: 0.2321 (0.2331)  loss_box_reg: 0.2871 (0.2993)  loss_objectness: 0.0736 (0.0740)  loss_rpn_box_reg: 0.0605 (0.0616)  time: 1.7430  data: 0.7179  max mem: 19870\n",
      "Epoch: [2]  [1960/2444]  eta: 0:13:42  lr: 0.020000  loss: 0.6483 (0.6679)  loss_classifier: 0.2321 (0.2330)  loss_box_reg: 0.2821 (0.2992)  loss_objectness: 0.0718 (0.0741)  loss_rpn_box_reg: 0.0552 (0.0616)  time: 1.7440  data: 0.7180  max mem: 19870\n",
      "Epoch: [2]  [1970/2444]  eta: 0:13:25  lr: 0.020000  loss: 0.6549 (0.6679)  loss_classifier: 0.2252 (0.2330)  loss_box_reg: 0.2868 (0.2992)  loss_objectness: 0.0718 (0.0740)  loss_rpn_box_reg: 0.0552 (0.0616)  time: 1.7016  data: 0.6800  max mem: 19870\n",
      "Epoch: [2]  [1980/2444]  eta: 0:13:08  lr: 0.020000  loss: 0.6489 (0.6676)  loss_classifier: 0.2299 (0.2330)  loss_box_reg: 0.2936 (0.2991)  loss_objectness: 0.0647 (0.0740)  loss_rpn_box_reg: 0.0564 (0.0615)  time: 1.6878  data: 0.6692  max mem: 19870\n",
      "Epoch: [2]  [1990/2444]  eta: 0:12:51  lr: 0.020000  loss: 0.6232 (0.6675)  loss_classifier: 0.2299 (0.2329)  loss_box_reg: 0.2876 (0.2991)  loss_objectness: 0.0578 (0.0739)  loss_rpn_box_reg: 0.0531 (0.0615)  time: 1.6980  data: 0.6777  max mem: 19870\n",
      "Epoch: [2]  [2000/2444]  eta: 0:12:34  lr: 0.020000  loss: 0.6548 (0.6676)  loss_classifier: 0.2378 (0.2330)  loss_box_reg: 0.2890 (0.2991)  loss_objectness: 0.0615 (0.0740)  loss_rpn_box_reg: 0.0550 (0.0615)  time: 1.7007  data: 0.6792  max mem: 19870\n",
      "Epoch: [2]  [2010/2444]  eta: 0:12:17  lr: 0.020000  loss: 0.6632 (0.6674)  loss_classifier: 0.2369 (0.2330)  loss_box_reg: 0.2890 (0.2990)  loss_objectness: 0.0642 (0.0739)  loss_rpn_box_reg: 0.0567 (0.0615)  time: 1.6929  data: 0.6735  max mem: 19870\n",
      "Epoch: [2]  [2020/2444]  eta: 0:12:00  lr: 0.020000  loss: 0.6386 (0.6671)  loss_classifier: 0.2212 (0.2329)  loss_box_reg: 0.2824 (0.2989)  loss_objectness: 0.0608 (0.0739)  loss_rpn_box_reg: 0.0563 (0.0615)  time: 1.6926  data: 0.6732  max mem: 19870\n",
      "Epoch: [2]  [2030/2444]  eta: 0:11:43  lr: 0.020000  loss: 0.5980 (0.6670)  loss_classifier: 0.2162 (0.2328)  loss_box_reg: 0.2783 (0.2989)  loss_objectness: 0.0551 (0.0738)  loss_rpn_box_reg: 0.0539 (0.0615)  time: 1.6927  data: 0.6723  max mem: 19870\n",
      "Epoch: [2]  [2040/2444]  eta: 0:11:26  lr: 0.020000  loss: 0.6259 (0.6669)  loss_classifier: 0.2191 (0.2328)  loss_box_reg: 0.2842 (0.2988)  loss_objectness: 0.0655 (0.0738)  loss_rpn_box_reg: 0.0547 (0.0614)  time: 1.6960  data: 0.6762  max mem: 19870\n",
      "Epoch: [2]  [2050/2444]  eta: 0:11:09  lr: 0.020000  loss: 0.6437 (0.6668)  loss_classifier: 0.2201 (0.2328)  loss_box_reg: 0.2866 (0.2988)  loss_objectness: 0.0694 (0.0738)  loss_rpn_box_reg: 0.0572 (0.0615)  time: 1.6994  data: 0.6809  max mem: 19870\n",
      "Epoch: [2]  [2060/2444]  eta: 0:10:52  lr: 0.020000  loss: 0.6437 (0.6668)  loss_classifier: 0.2255 (0.2328)  loss_box_reg: 0.2894 (0.2988)  loss_objectness: 0.0586 (0.0738)  loss_rpn_box_reg: 0.0590 (0.0614)  time: 1.7055  data: 0.6862  max mem: 19870\n",
      "Epoch: [2]  [2070/2444]  eta: 0:10:35  lr: 0.020000  loss: 0.6407 (0.6667)  loss_classifier: 0.2255 (0.2327)  loss_box_reg: 0.2876 (0.2987)  loss_objectness: 0.0615 (0.0738)  loss_rpn_box_reg: 0.0570 (0.0614)  time: 1.6983  data: 0.6797  max mem: 19870\n",
      "Epoch: [2]  [2080/2444]  eta: 0:10:18  lr: 0.020000  loss: 0.6393 (0.6667)  loss_classifier: 0.2244 (0.2328)  loss_box_reg: 0.2876 (0.2987)  loss_objectness: 0.0715 (0.0737)  loss_rpn_box_reg: 0.0570 (0.0614)  time: 1.6911  data: 0.6735  max mem: 19870\n",
      "Epoch: [2]  [2090/2444]  eta: 0:10:01  lr: 0.020000  loss: 0.6517 (0.6666)  loss_classifier: 0.2240 (0.2328)  loss_box_reg: 0.2928 (0.2987)  loss_objectness: 0.0667 (0.0737)  loss_rpn_box_reg: 0.0587 (0.0614)  time: 1.6970  data: 0.6784  max mem: 19870\n",
      "Epoch: [2]  [2100/2444]  eta: 0:09:44  lr: 0.020000  loss: 0.6627 (0.6668)  loss_classifier: 0.2348 (0.2328)  loss_box_reg: 0.2979 (0.2987)  loss_objectness: 0.0679 (0.0738)  loss_rpn_box_reg: 0.0650 (0.0614)  time: 1.6969  data: 0.6777  max mem: 19870\n",
      "Epoch: [2]  [2110/2444]  eta: 0:09:27  lr: 0.020000  loss: 0.6746 (0.6670)  loss_classifier: 0.2431 (0.2329)  loss_box_reg: 0.3013 (0.2988)  loss_objectness: 0.0771 (0.0739)  loss_rpn_box_reg: 0.0651 (0.0615)  time: 1.7010  data: 0.6831  max mem: 19870\n",
      "Epoch: [2]  [2120/2444]  eta: 0:09:10  lr: 0.020000  loss: 0.6651 (0.6670)  loss_classifier: 0.2420 (0.2329)  loss_box_reg: 0.3049 (0.2988)  loss_objectness: 0.0731 (0.0738)  loss_rpn_box_reg: 0.0609 (0.0615)  time: 1.7049  data: 0.6855  max mem: 19870\n",
      "Epoch: [2]  [2130/2444]  eta: 0:08:53  lr: 0.020000  loss: 0.6445 (0.6668)  loss_classifier: 0.2271 (0.2328)  loss_box_reg: 0.2949 (0.2987)  loss_objectness: 0.0644 (0.0738)  loss_rpn_box_reg: 0.0593 (0.0614)  time: 1.6979  data: 0.6779  max mem: 19870\n",
      "Epoch: [2]  [2140/2444]  eta: 0:08:36  lr: 0.020000  loss: 0.6183 (0.6667)  loss_classifier: 0.2271 (0.2328)  loss_box_reg: 0.2848 (0.2987)  loss_objectness: 0.0555 (0.0738)  loss_rpn_box_reg: 0.0565 (0.0614)  time: 1.6946  data: 0.6766  max mem: 19870\n",
      "Epoch: [2]  [2150/2444]  eta: 0:08:19  lr: 0.020000  loss: 0.6686 (0.6668)  loss_classifier: 0.2300 (0.2328)  loss_box_reg: 0.3009 (0.2987)  loss_objectness: 0.0679 (0.0738)  loss_rpn_box_reg: 0.0647 (0.0614)  time: 1.7014  data: 0.6820  max mem: 19870\n",
      "Epoch: [2]  [2160/2444]  eta: 0:08:02  lr: 0.020000  loss: 0.7031 (0.6669)  loss_classifier: 0.2300 (0.2328)  loss_box_reg: 0.3038 (0.2988)  loss_objectness: 0.0810 (0.0738)  loss_rpn_box_reg: 0.0671 (0.0615)  time: 1.7116  data: 0.6904  max mem: 19870\n",
      "Epoch: [2]  [2170/2444]  eta: 0:07:45  lr: 0.020000  loss: 0.6566 (0.6668)  loss_classifier: 0.2269 (0.2328)  loss_box_reg: 0.3044 (0.2987)  loss_objectness: 0.0633 (0.0738)  loss_rpn_box_reg: 0.0647 (0.0615)  time: 1.7035  data: 0.6830  max mem: 19870\n",
      "Epoch: [2]  [2180/2444]  eta: 0:07:28  lr: 0.020000  loss: 0.6566 (0.6668)  loss_classifier: 0.2264 (0.2328)  loss_box_reg: 0.2951 (0.2988)  loss_objectness: 0.0639 (0.0738)  loss_rpn_box_reg: 0.0578 (0.0615)  time: 1.7021  data: 0.6821  max mem: 19870\n",
      "Epoch: [2]  [2190/2444]  eta: 0:07:11  lr: 0.020000  loss: 0.6866 (0.6669)  loss_classifier: 0.2348 (0.2328)  loss_box_reg: 0.2951 (0.2988)  loss_objectness: 0.0772 (0.0739)  loss_rpn_box_reg: 0.0616 (0.0615)  time: 1.7124  data: 0.6920  max mem: 19870\n",
      "Epoch: [2]  [2200/2444]  eta: 0:06:54  lr: 0.020000  loss: 0.6670 (0.6668)  loss_classifier: 0.2356 (0.2328)  loss_box_reg: 0.2891 (0.2987)  loss_objectness: 0.0614 (0.0738)  loss_rpn_box_reg: 0.0566 (0.0615)  time: 1.7032  data: 0.6831  max mem: 19870\n",
      "Epoch: [2]  [2210/2444]  eta: 0:06:37  lr: 0.020000  loss: 0.6643 (0.6669)  loss_classifier: 0.2372 (0.2328)  loss_box_reg: 0.2902 (0.2988)  loss_objectness: 0.0614 (0.0738)  loss_rpn_box_reg: 0.0582 (0.0615)  time: 1.7141  data: 0.6928  max mem: 19870\n",
      "Epoch: [2]  [2220/2444]  eta: 0:06:20  lr: 0.020000  loss: 0.6732 (0.6669)  loss_classifier: 0.2343 (0.2328)  loss_box_reg: 0.2960 (0.2988)  loss_objectness: 0.0665 (0.0738)  loss_rpn_box_reg: 0.0712 (0.0615)  time: 1.7228  data: 0.7005  max mem: 19870\n",
      "Epoch: [2]  [2230/2444]  eta: 0:06:03  lr: 0.020000  loss: 0.6594 (0.6668)  loss_classifier: 0.2260 (0.2328)  loss_box_reg: 0.2951 (0.2987)  loss_objectness: 0.0689 (0.0738)  loss_rpn_box_reg: 0.0594 (0.0615)  time: 1.7073  data: 0.6859  max mem: 19870\n",
      "Epoch: [2]  [2240/2444]  eta: 0:05:46  lr: 0.020000  loss: 0.6493 (0.6667)  loss_classifier: 0.2242 (0.2328)  loss_box_reg: 0.2870 (0.2987)  loss_objectness: 0.0718 (0.0738)  loss_rpn_box_reg: 0.0594 (0.0615)  time: 1.6937  data: 0.6738  max mem: 19870\n",
      "Epoch: [2]  [2250/2444]  eta: 0:05:29  lr: 0.020000  loss: 0.6493 (0.6666)  loss_classifier: 0.2212 (0.2327)  loss_box_reg: 0.2859 (0.2986)  loss_objectness: 0.0711 (0.0737)  loss_rpn_box_reg: 0.0642 (0.0615)  time: 1.6992  data: 0.6787  max mem: 19870\n",
      "Epoch: [2]  [2260/2444]  eta: 0:05:12  lr: 0.020000  loss: 0.6554 (0.6666)  loss_classifier: 0.2246 (0.2327)  loss_box_reg: 0.2917 (0.2987)  loss_objectness: 0.0612 (0.0738)  loss_rpn_box_reg: 0.0627 (0.0615)  time: 1.7234  data: 0.7016  max mem: 19870\n",
      "Epoch: [2]  [2270/2444]  eta: 0:04:55  lr: 0.020000  loss: 0.6655 (0.6666)  loss_classifier: 0.2349 (0.2327)  loss_box_reg: 0.2917 (0.2986)  loss_objectness: 0.0770 (0.0738)  loss_rpn_box_reg: 0.0609 (0.0615)  time: 1.7159  data: 0.6947  max mem: 19870\n",
      "Epoch: [2]  [2280/2444]  eta: 0:04:38  lr: 0.020000  loss: 0.6609 (0.6667)  loss_classifier: 0.2333 (0.2327)  loss_box_reg: 0.2892 (0.2987)  loss_objectness: 0.0680 (0.0738)  loss_rpn_box_reg: 0.0609 (0.0615)  time: 1.7068  data: 0.6872  max mem: 19870\n",
      "Epoch: [2]  [2290/2444]  eta: 0:04:21  lr: 0.020000  loss: 0.6158 (0.6665)  loss_classifier: 0.2249 (0.2327)  loss_box_reg: 0.2911 (0.2986)  loss_objectness: 0.0593 (0.0737)  loss_rpn_box_reg: 0.0582 (0.0615)  time: 1.7039  data: 0.6860  max mem: 19870\n",
      "Epoch: [2]  [2300/2444]  eta: 0:04:04  lr: 0.020000  loss: 0.6449 (0.6666)  loss_classifier: 0.2275 (0.2327)  loss_box_reg: 0.2963 (0.2987)  loss_objectness: 0.0661 (0.0737)  loss_rpn_box_reg: 0.0625 (0.0615)  time: 1.7133  data: 0.6941  max mem: 19870\n",
      "Epoch: [2]  [2310/2444]  eta: 0:03:47  lr: 0.020000  loss: 0.6489 (0.6665)  loss_classifier: 0.2275 (0.2327)  loss_box_reg: 0.2901 (0.2986)  loss_objectness: 0.0693 (0.0737)  loss_rpn_box_reg: 0.0635 (0.0615)  time: 1.7246  data: 0.7038  max mem: 19870\n",
      "Epoch: [2]  [2320/2444]  eta: 0:03:30  lr: 0.020000  loss: 0.6361 (0.6665)  loss_classifier: 0.2197 (0.2327)  loss_box_reg: 0.2941 (0.2986)  loss_objectness: 0.0634 (0.0737)  loss_rpn_box_reg: 0.0581 (0.0615)  time: 1.7098  data: 0.6891  max mem: 19870\n",
      "Epoch: [2]  [2330/2444]  eta: 0:03:13  lr: 0.020000  loss: 0.6457 (0.6664)  loss_classifier: 0.2299 (0.2326)  loss_box_reg: 0.2960 (0.2986)  loss_objectness: 0.0575 (0.0737)  loss_rpn_box_reg: 0.0586 (0.0615)  time: 1.7119  data: 0.6913  max mem: 19870\n",
      "Epoch: [2]  [2340/2444]  eta: 0:02:56  lr: 0.020000  loss: 0.6483 (0.6664)  loss_classifier: 0.2223 (0.2326)  loss_box_reg: 0.2862 (0.2986)  loss_objectness: 0.0654 (0.0737)  loss_rpn_box_reg: 0.0581 (0.0615)  time: 1.7049  data: 0.6850  max mem: 19870\n",
      "Epoch: [2]  [2350/2444]  eta: 0:02:39  lr: 0.020000  loss: 0.6405 (0.6663)  loss_classifier: 0.2215 (0.2326)  loss_box_reg: 0.2862 (0.2986)  loss_objectness: 0.0732 (0.0737)  loss_rpn_box_reg: 0.0595 (0.0615)  time: 1.6960  data: 0.6769  max mem: 19870\n",
      "Epoch: [2]  [2360/2444]  eta: 0:02:22  lr: 0.020000  loss: 0.6305 (0.6663)  loss_classifier: 0.2191 (0.2326)  loss_box_reg: 0.2838 (0.2986)  loss_objectness: 0.0717 (0.0737)  loss_rpn_box_reg: 0.0619 (0.0615)  time: 1.7053  data: 0.6853  max mem: 19870\n",
      "Epoch: [2]  [2370/2444]  eta: 0:02:05  lr: 0.020000  loss: 0.6708 (0.6663)  loss_classifier: 0.2308 (0.2326)  loss_box_reg: 0.2990 (0.2986)  loss_objectness: 0.0701 (0.0737)  loss_rpn_box_reg: 0.0613 (0.0615)  time: 1.7267  data: 0.7051  max mem: 19870\n",
      "Epoch: [2]  [2380/2444]  eta: 0:01:48  lr: 0.020000  loss: 0.6708 (0.6664)  loss_classifier: 0.2308 (0.2326)  loss_box_reg: 0.3013 (0.2986)  loss_objectness: 0.0733 (0.0737)  loss_rpn_box_reg: 0.0625 (0.0615)  time: 1.7293  data: 0.7074  max mem: 19870\n",
      "Epoch: [2]  [2390/2444]  eta: 0:01:31  lr: 0.020000  loss: 0.6504 (0.6662)  loss_classifier: 0.2240 (0.2325)  loss_box_reg: 0.2808 (0.2985)  loss_objectness: 0.0733 (0.0737)  loss_rpn_box_reg: 0.0602 (0.0615)  time: 1.7109  data: 0.6906  max mem: 19870\n",
      "Epoch: [2]  [2400/2444]  eta: 0:01:14  lr: 0.020000  loss: 0.6301 (0.6661)  loss_classifier: 0.2240 (0.2325)  loss_box_reg: 0.2808 (0.2985)  loss_objectness: 0.0664 (0.0736)  loss_rpn_box_reg: 0.0579 (0.0615)  time: 1.7025  data: 0.6834  max mem: 19870\n",
      "Epoch: [2]  [2410/2444]  eta: 0:00:57  lr: 0.020000  loss: 0.6359 (0.6660)  loss_classifier: 0.2265 (0.2325)  loss_box_reg: 0.2985 (0.2985)  loss_objectness: 0.0624 (0.0736)  loss_rpn_box_reg: 0.0567 (0.0615)  time: 1.7079  data: 0.6875  max mem: 19870\n",
      "Epoch: [2]  [2420/2444]  eta: 0:00:40  lr: 0.020000  loss: 0.6378 (0.6659)  loss_classifier: 0.2210 (0.2324)  loss_box_reg: 0.2956 (0.2984)  loss_objectness: 0.0618 (0.0736)  loss_rpn_box_reg: 0.0528 (0.0614)  time: 1.7072  data: 0.6856  max mem: 19870\n",
      "Epoch: [2]  [2430/2444]  eta: 0:00:23  lr: 0.020000  loss: 0.6338 (0.6658)  loss_classifier: 0.2210 (0.2324)  loss_box_reg: 0.2828 (0.2984)  loss_objectness: 0.0672 (0.0736)  loss_rpn_box_reg: 0.0555 (0.0614)  time: 1.7043  data: 0.6829  max mem: 19870\n",
      "Epoch: [2]  [2440/2444]  eta: 0:00:06  lr: 0.020000  loss: 0.6365 (0.6658)  loss_classifier: 0.2199 (0.2324)  loss_box_reg: 0.2918 (0.2984)  loss_objectness: 0.0745 (0.0736)  loss_rpn_box_reg: 0.0592 (0.0614)  time: 1.7114  data: 0.6907  max mem: 19870\n",
      "Epoch: [2]  [2443/2444]  eta: 0:00:01  lr: 0.020000  loss: 0.6444 (0.6655)  loss_classifier: 0.2241 (0.2323)  loss_box_reg: 0.2962 (0.2983)  loss_objectness: 0.0683 (0.0735)  loss_rpn_box_reg: 0.0581 (0.0614)  time: 1.6313  data: 0.6582  max mem: 19870\n",
      "Epoch: [2] Total time: 1:09:12 (1.6992 s / it)\n",
      "creating index...\n",
      "index created!\n",
      "Test:  [   0/4952]  eta: 0:06:16  model_time: 0.0382 (0.0382)  evaluator_time: 0.0160 (0.0160)  time: 0.0760  data: 0.0207  max mem: 19870\n",
      "Test:  [ 100/4952]  eta: 0:04:50  model_time: 0.0331 (0.0339)  evaluator_time: 0.0090 (0.0101)  time: 0.0616  data: 0.0171  max mem: 19870\n",
      "Test:  [ 200/4952]  eta: 0:04:48  model_time: 0.0334 (0.0339)  evaluator_time: 0.0056 (0.0103)  time: 0.0611  data: 0.0153  max mem: 19870\n",
      "Test:  [ 300/4952]  eta: 0:04:41  model_time: 0.0331 (0.0338)  evaluator_time: 0.0082 (0.0102)  time: 0.0635  data: 0.0166  max mem: 19870\n",
      "Test:  [ 400/4952]  eta: 0:04:36  model_time: 0.0330 (0.0338)  evaluator_time: 0.0071 (0.0106)  time: 0.0586  data: 0.0153  max mem: 19870\n",
      "Test:  [ 500/4952]  eta: 0:04:29  model_time: 0.0336 (0.0338)  evaluator_time: 0.0069 (0.0104)  time: 0.0606  data: 0.0155  max mem: 19870\n",
      "Test:  [ 600/4952]  eta: 0:04:24  model_time: 0.0332 (0.0337)  evaluator_time: 0.0061 (0.0106)  time: 0.0589  data: 0.0152  max mem: 19870\n",
      "Test:  [ 700/4952]  eta: 0:04:21  model_time: 0.0336 (0.0338)  evaluator_time: 0.0066 (0.0108)  time: 0.0605  data: 0.0166  max mem: 19870\n",
      "Test:  [ 800/4952]  eta: 0:04:15  model_time: 0.0331 (0.0338)  evaluator_time: 0.0092 (0.0109)  time: 0.0615  data: 0.0159  max mem: 19870\n",
      "Test:  [ 900/4952]  eta: 0:04:07  model_time: 0.0331 (0.0338)  evaluator_time: 0.0055 (0.0108)  time: 0.0575  data: 0.0144  max mem: 19870\n",
      "Test:  [1000/4952]  eta: 0:04:01  model_time: 0.0328 (0.0338)  evaluator_time: 0.0061 (0.0106)  time: 0.0555  data: 0.0139  max mem: 19870\n",
      "Test:  [1100/4952]  eta: 0:03:55  model_time: 0.0331 (0.0338)  evaluator_time: 0.0062 (0.0107)  time: 0.0566  data: 0.0149  max mem: 19870\n",
      "Test:  [1200/4952]  eta: 0:03:49  model_time: 0.0338 (0.0338)  evaluator_time: 0.0117 (0.0107)  time: 0.0669  data: 0.0171  max mem: 19870\n",
      "Test:  [1300/4952]  eta: 0:03:43  model_time: 0.0329 (0.0338)  evaluator_time: 0.0061 (0.0107)  time: 0.0609  data: 0.0157  max mem: 19870\n",
      "Test:  [1400/4952]  eta: 0:03:37  model_time: 0.0335 (0.0338)  evaluator_time: 0.0060 (0.0107)  time: 0.0605  data: 0.0153  max mem: 19870\n",
      "Test:  [1500/4952]  eta: 0:03:31  model_time: 0.0328 (0.0338)  evaluator_time: 0.0070 (0.0107)  time: 0.0610  data: 0.0162  max mem: 19870\n",
      "Test:  [1600/4952]  eta: 0:03:25  model_time: 0.0333 (0.0338)  evaluator_time: 0.0073 (0.0107)  time: 0.0627  data: 0.0161  max mem: 19870\n",
      "Test:  [1700/4952]  eta: 0:03:18  model_time: 0.0330 (0.0338)  evaluator_time: 0.0054 (0.0107)  time: 0.0600  data: 0.0153  max mem: 19870\n",
      "Test:  [1800/4952]  eta: 0:03:12  model_time: 0.0335 (0.0338)  evaluator_time: 0.0061 (0.0107)  time: 0.0760  data: 0.0297  max mem: 19870\n",
      "Test:  [1900/4952]  eta: 0:03:06  model_time: 0.0333 (0.0338)  evaluator_time: 0.0069 (0.0107)  time: 0.0585  data: 0.0149  max mem: 19870\n",
      "Test:  [2000/4952]  eta: 0:03:00  model_time: 0.0331 (0.0338)  evaluator_time: 0.0095 (0.0107)  time: 0.0643  data: 0.0164  max mem: 19870\n",
      "Test:  [2100/4952]  eta: 0:02:54  model_time: 0.0331 (0.0338)  evaluator_time: 0.0069 (0.0106)  time: 0.0610  data: 0.0159  max mem: 19870\n",
      "Test:  [2200/4952]  eta: 0:02:47  model_time: 0.0329 (0.0337)  evaluator_time: 0.0064 (0.0106)  time: 0.0628  data: 0.0158  max mem: 19870\n",
      "Test:  [2300/4952]  eta: 0:02:41  model_time: 0.0332 (0.0337)  evaluator_time: 0.0066 (0.0105)  time: 0.0625  data: 0.0159  max mem: 19870\n",
      "Test:  [2400/4952]  eta: 0:02:35  model_time: 0.0336 (0.0337)  evaluator_time: 0.0096 (0.0105)  time: 0.0656  data: 0.0166  max mem: 19870\n",
      "Test:  [2500/4952]  eta: 0:02:29  model_time: 0.0332 (0.0337)  evaluator_time: 0.0066 (0.0105)  time: 0.0630  data: 0.0171  max mem: 19870\n",
      "Test:  [2600/4952]  eta: 0:02:22  model_time: 0.0338 (0.0337)  evaluator_time: 0.0069 (0.0105)  time: 0.0613  data: 0.0148  max mem: 19870\n",
      "Test:  [2700/4952]  eta: 0:02:16  model_time: 0.0337 (0.0337)  evaluator_time: 0.0071 (0.0105)  time: 0.0602  data: 0.0146  max mem: 19870\n",
      "Test:  [2800/4952]  eta: 0:02:11  model_time: 0.0341 (0.0338)  evaluator_time: 0.0076 (0.0107)  time: 0.0621  data: 0.0156  max mem: 19870\n",
      "Test:  [2900/4952]  eta: 0:02:05  model_time: 0.0349 (0.0338)  evaluator_time: 0.0068 (0.0107)  time: 0.0656  data: 0.0173  max mem: 19870\n",
      "Test:  [3000/4952]  eta: 0:01:59  model_time: 0.0345 (0.0338)  evaluator_time: 0.0080 (0.0107)  time: 0.0633  data: 0.0159  max mem: 19870\n",
      "Test:  [3100/4952]  eta: 0:01:53  model_time: 0.0348 (0.0339)  evaluator_time: 0.0071 (0.0106)  time: 0.0646  data: 0.0164  max mem: 19870\n",
      "Test:  [3200/4952]  eta: 0:01:47  model_time: 0.0353 (0.0339)  evaluator_time: 0.0078 (0.0106)  time: 0.0658  data: 0.0170  max mem: 19870\n",
      "Test:  [3300/4952]  eta: 0:01:41  model_time: 0.0338 (0.0340)  evaluator_time: 0.0073 (0.0107)  time: 0.0621  data: 0.0163  max mem: 19870\n",
      "Test:  [3400/4952]  eta: 0:01:35  model_time: 0.0340 (0.0340)  evaluator_time: 0.0054 (0.0107)  time: 0.0627  data: 0.0160  max mem: 19870\n",
      "Test:  [3500/4952]  eta: 0:01:28  model_time: 0.0337 (0.0340)  evaluator_time: 0.0072 (0.0107)  time: 0.0572  data: 0.0143  max mem: 19870\n",
      "Test:  [3600/4952]  eta: 0:01:22  model_time: 0.0345 (0.0340)  evaluator_time: 0.0077 (0.0107)  time: 0.0603  data: 0.0161  max mem: 19870\n",
      "Test:  [3700/4952]  eta: 0:01:16  model_time: 0.0332 (0.0340)  evaluator_time: 0.0057 (0.0107)  time: 0.0573  data: 0.0146  max mem: 19870\n",
      "Test:  [3800/4952]  eta: 0:01:10  model_time: 0.0340 (0.0340)  evaluator_time: 0.0059 (0.0107)  time: 0.0570  data: 0.0141  max mem: 19870\n",
      "Test:  [3900/4952]  eta: 0:01:04  model_time: 0.0336 (0.0340)  evaluator_time: 0.0071 (0.0108)  time: 0.0591  data: 0.0152  max mem: 19870\n",
      "Test:  [4000/4952]  eta: 0:00:58  model_time: 0.0338 (0.0340)  evaluator_time: 0.0091 (0.0108)  time: 0.0617  data: 0.0154  max mem: 19870\n",
      "Test:  [4100/4952]  eta: 0:00:52  model_time: 0.0340 (0.0341)  evaluator_time: 0.0063 (0.0108)  time: 0.0607  data: 0.0156  max mem: 19870\n",
      "Test:  [4200/4952]  eta: 0:00:46  model_time: 0.0338 (0.0341)  evaluator_time: 0.0052 (0.0108)  time: 0.0584  data: 0.0140  max mem: 19870\n",
      "Test:  [4300/4952]  eta: 0:00:40  model_time: 0.0337 (0.0341)  evaluator_time: 0.0063 (0.0107)  time: 0.0586  data: 0.0144  max mem: 19870\n",
      "Test:  [4400/4952]  eta: 0:00:33  model_time: 0.0342 (0.0341)  evaluator_time: 0.0084 (0.0107)  time: 0.0604  data: 0.0155  max mem: 19870\n",
      "Test:  [4500/4952]  eta: 0:00:27  model_time: 0.0338 (0.0341)  evaluator_time: 0.0075 (0.0107)  time: 0.0624  data: 0.0160  max mem: 19870\n",
      "Test:  [4600/4952]  eta: 0:00:21  model_time: 0.0343 (0.0341)  evaluator_time: 0.0068 (0.0107)  time: 0.0596  data: 0.0157  max mem: 19870\n",
      "Test:  [4700/4952]  eta: 0:00:15  model_time: 0.0337 (0.0341)  evaluator_time: 0.0058 (0.0107)  time: 0.0576  data: 0.0150  max mem: 19870\n",
      "Test:  [4800/4952]  eta: 0:00:09  model_time: 0.0336 (0.0341)  evaluator_time: 0.0077 (0.0107)  time: 0.0605  data: 0.0153  max mem: 19870\n",
      "Test:  [4900/4952]  eta: 0:00:03  model_time: 0.0341 (0.0341)  evaluator_time: 0.0053 (0.0107)  time: 0.0577  data: 0.0141  max mem: 19870\n",
      "Test:  [4951/4952]  eta: 0:00:00  model_time: 0.0339 (0.0341)  evaluator_time: 0.0061 (0.0107)  time: 0.0599  data: 0.0145  max mem: 19870\n",
      "Test: Total time: 0:05:03 (0.0613 s / it)\n",
      "Averaged stats: model_time: 0.0339 (0.0341)  evaluator_time: 0.0061 (0.0107)\n",
      "Accumulating evaluation results...\n",
      "DONE (t=8.32s).\n",
      "IoU metric: bbox\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.324\n",
      " Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.540\n",
      " Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.344\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.134\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.319\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.436\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.280\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.461\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.489\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.263\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.495\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.602\n",
      "Epoch: [3]  [   0/2444]  eta: 1:13:01  lr: 0.002000  loss: 0.7296 (0.7296)  loss_classifier: 0.2299 (0.2299)  loss_box_reg: 0.3141 (0.3141)  loss_objectness: 0.1244 (0.1244)  loss_rpn_box_reg: 0.0613 (0.0613)  time: 1.7929  data: 0.7381  max mem: 19870\n",
      "Epoch: [3]  [  10/2444]  eta: 1:11:06  lr: 0.002000  loss: 0.6798 (0.6962)  loss_classifier: 0.2422 (0.2432)  loss_box_reg: 0.3051 (0.3088)  loss_objectness: 0.0611 (0.0817)  loss_rpn_box_reg: 0.0613 (0.0624)  time: 1.7530  data: 0.7263  max mem: 19870\n",
      "Epoch: [3]  [  20/2444]  eta: 1:10:42  lr: 0.002000  loss: 0.6659 (0.6732)  loss_classifier: 0.2304 (0.2333)  loss_box_reg: 0.2963 (0.3026)  loss_objectness: 0.0624 (0.0766)  loss_rpn_box_reg: 0.0583 (0.0608)  time: 1.7482  data: 0.7239  max mem: 19870\n",
      "Epoch: [3]  [  30/2444]  eta: 1:10:12  lr: 0.002000  loss: 0.6489 (0.6697)  loss_classifier: 0.2275 (0.2336)  loss_box_reg: 0.2958 (0.3006)  loss_objectness: 0.0624 (0.0739)  loss_rpn_box_reg: 0.0580 (0.0615)  time: 1.7405  data: 0.7152  max mem: 19870\n",
      "Epoch: [3]  [  40/2444]  eta: 1:09:51  lr: 0.002000  loss: 0.6410 (0.6622)  loss_classifier: 0.2287 (0.2323)  loss_box_reg: 0.2946 (0.2980)  loss_objectness: 0.0567 (0.0708)  loss_rpn_box_reg: 0.0584 (0.0611)  time: 1.7363  data: 0.7107  max mem: 19870\n",
      "Epoch: [3]  [  50/2444]  eta: 1:09:25  lr: 0.002000  loss: 0.6134 (0.6523)  loss_classifier: 0.2256 (0.2305)  loss_box_reg: 0.2812 (0.2933)  loss_objectness: 0.0537 (0.0687)  loss_rpn_box_reg: 0.0552 (0.0599)  time: 1.7321  data: 0.7072  max mem: 19870\n",
      "Epoch: [3]  [  60/2444]  eta: 1:09:02  lr: 0.002000  loss: 0.6340 (0.6526)  loss_classifier: 0.2236 (0.2305)  loss_box_reg: 0.2889 (0.2932)  loss_objectness: 0.0547 (0.0688)  loss_rpn_box_reg: 0.0541 (0.0601)  time: 1.7262  data: 0.7013  max mem: 19870\n",
      "Epoch: [3]  [  70/2444]  eta: 1:08:46  lr: 0.002000  loss: 0.6678 (0.6558)  loss_classifier: 0.2356 (0.2317)  loss_box_reg: 0.2967 (0.2949)  loss_objectness: 0.0661 (0.0690)  loss_rpn_box_reg: 0.0612 (0.0602)  time: 1.7339  data: 0.7093  max mem: 19870\n",
      "Epoch: [3]  [  80/2444]  eta: 1:08:32  lr: 0.002000  loss: 0.6678 (0.6547)  loss_classifier: 0.2354 (0.2302)  loss_box_reg: 0.2957 (0.2934)  loss_objectness: 0.0696 (0.0704)  loss_rpn_box_reg: 0.0652 (0.0607)  time: 1.7457  data: 0.7136  max mem: 19870\n",
      "Epoch: [3]  [  90/2444]  eta: 1:08:10  lr: 0.002000  loss: 0.6627 (0.6543)  loss_classifier: 0.2270 (0.2302)  loss_box_reg: 0.2957 (0.2938)  loss_objectness: 0.0707 (0.0704)  loss_rpn_box_reg: 0.0589 (0.0600)  time: 1.7364  data: 0.7015  max mem: 19870\n",
      "Epoch: [3]  [ 100/2444]  eta: 1:07:55  lr: 0.002000  loss: 0.6801 (0.6587)  loss_classifier: 0.2404 (0.2321)  loss_box_reg: 0.3030 (0.2953)  loss_objectness: 0.0635 (0.0705)  loss_rpn_box_reg: 0.0603 (0.0608)  time: 1.7335  data: 0.7061  max mem: 19870\n",
      "Epoch: [3]  [ 110/2444]  eta: 1:07:33  lr: 0.002000  loss: 0.6889 (0.6572)  loss_classifier: 0.2356 (0.2310)  loss_box_reg: 0.3005 (0.2944)  loss_objectness: 0.0718 (0.0713)  loss_rpn_box_reg: 0.0636 (0.0604)  time: 1.7308  data: 0.7046  max mem: 19870\n",
      "Epoch: [3]  [ 120/2444]  eta: 1:07:16  lr: 0.002000  loss: 0.6335 (0.6570)  loss_classifier: 0.2319 (0.2315)  loss_box_reg: 0.2923 (0.2946)  loss_objectness: 0.0668 (0.0707)  loss_rpn_box_reg: 0.0564 (0.0602)  time: 1.7273  data: 0.7015  max mem: 19870\n",
      "Epoch: [3]  [ 130/2444]  eta: 1:06:53  lr: 0.002000  loss: 0.6445 (0.6552)  loss_classifier: 0.2329 (0.2308)  loss_box_reg: 0.2923 (0.2941)  loss_objectness: 0.0634 (0.0700)  loss_rpn_box_reg: 0.0564 (0.0603)  time: 1.7229  data: 0.6983  max mem: 19870\n",
      "Epoch: [3]  [ 140/2444]  eta: 1:06:37  lr: 0.002000  loss: 0.6463 (0.6552)  loss_classifier: 0.2329 (0.2305)  loss_box_reg: 0.2939 (0.2945)  loss_objectness: 0.0592 (0.0700)  loss_rpn_box_reg: 0.0586 (0.0602)  time: 1.7245  data: 0.7002  max mem: 19870\n",
      "Epoch: [3]  [ 150/2444]  eta: 1:06:19  lr: 0.002000  loss: 0.6472 (0.6534)  loss_classifier: 0.2327 (0.2298)  loss_box_reg: 0.2989 (0.2936)  loss_objectness: 0.0563 (0.0699)  loss_rpn_box_reg: 0.0550 (0.0601)  time: 1.7358  data: 0.7117  max mem: 19870\n",
      "Epoch: [3]  [ 160/2444]  eta: 1:05:59  lr: 0.002000  loss: 0.6093 (0.6501)  loss_classifier: 0.2187 (0.2290)  loss_box_reg: 0.2819 (0.2925)  loss_objectness: 0.0539 (0.0691)  loss_rpn_box_reg: 0.0515 (0.0594)  time: 1.7231  data: 0.6986  max mem: 19870\n",
      "Epoch: [3]  [ 170/2444]  eta: 1:05:40  lr: 0.002000  loss: 0.6070 (0.6494)  loss_classifier: 0.2194 (0.2286)  loss_box_reg: 0.2687 (0.2916)  loss_objectness: 0.0622 (0.0696)  loss_rpn_box_reg: 0.0538 (0.0596)  time: 1.7187  data: 0.6951  max mem: 19870\n",
      "Epoch: [3]  [ 180/2444]  eta: 1:05:24  lr: 0.002000  loss: 0.6700 (0.6526)  loss_classifier: 0.2252 (0.2293)  loss_box_reg: 0.2949 (0.2926)  loss_objectness: 0.0765 (0.0704)  loss_rpn_box_reg: 0.0593 (0.0602)  time: 1.7334  data: 0.7090  max mem: 19870\n",
      "Epoch: [3]  [ 190/2444]  eta: 1:05:07  lr: 0.002000  loss: 0.6783 (0.6514)  loss_classifier: 0.2264 (0.2289)  loss_box_reg: 0.3000 (0.2923)  loss_objectness: 0.0741 (0.0701)  loss_rpn_box_reg: 0.0596 (0.0601)  time: 1.7386  data: 0.7138  max mem: 19870\n",
      "Epoch: [3]  [ 200/2444]  eta: 1:04:49  lr: 0.002000  loss: 0.6543 (0.6516)  loss_classifier: 0.2242 (0.2290)  loss_box_reg: 0.3000 (0.2928)  loss_objectness: 0.0538 (0.0699)  loss_rpn_box_reg: 0.0582 (0.0599)  time: 1.7305  data: 0.7064  max mem: 19870\n",
      "Epoch: [3]  [ 210/2444]  eta: 1:04:32  lr: 0.002000  loss: 0.6544 (0.6526)  loss_classifier: 0.2365 (0.2293)  loss_box_reg: 0.3031 (0.2933)  loss_objectness: 0.0613 (0.0700)  loss_rpn_box_reg: 0.0571 (0.0601)  time: 1.7350  data: 0.7109  max mem: 19870\n",
      "Epoch: [3]  [ 220/2444]  eta: 1:04:16  lr: 0.002000  loss: 0.6610 (0.6547)  loss_classifier: 0.2365 (0.2298)  loss_box_reg: 0.3015 (0.2938)  loss_objectness: 0.0729 (0.0707)  loss_rpn_box_reg: 0.0595 (0.0604)  time: 1.7411  data: 0.7165  max mem: 19870\n",
      "Epoch: [3]  [ 230/2444]  eta: 1:03:57  lr: 0.002000  loss: 0.6630 (0.6561)  loss_classifier: 0.2295 (0.2302)  loss_box_reg: 0.3016 (0.2945)  loss_objectness: 0.0729 (0.0708)  loss_rpn_box_reg: 0.0667 (0.0606)  time: 1.7327  data: 0.7068  max mem: 19870\n",
      "Epoch: [3]  [ 240/2444]  eta: 1:03:40  lr: 0.002000  loss: 0.6615 (0.6569)  loss_classifier: 0.2233 (0.2303)  loss_box_reg: 0.2979 (0.2948)  loss_objectness: 0.0718 (0.0710)  loss_rpn_box_reg: 0.0595 (0.0607)  time: 1.7266  data: 0.7006  max mem: 19870\n",
      "Epoch: [3]  [ 250/2444]  eta: 1:03:21  lr: 0.002000  loss: 0.6485 (0.6565)  loss_classifier: 0.2233 (0.2303)  loss_box_reg: 0.2995 (0.2949)  loss_objectness: 0.0715 (0.0708)  loss_rpn_box_reg: 0.0566 (0.0605)  time: 1.7255  data: 0.6998  max mem: 19870\n",
      "Epoch: [3]  [ 260/2444]  eta: 1:03:05  lr: 0.002000  loss: 0.6671 (0.6589)  loss_classifier: 0.2320 (0.2309)  loss_box_reg: 0.3058 (0.2960)  loss_objectness: 0.0704 (0.0713)  loss_rpn_box_reg: 0.0587 (0.0607)  time: 1.7345  data: 0.7102  max mem: 19870\n",
      "Epoch: [3]  [ 270/2444]  eta: 1:02:47  lr: 0.002000  loss: 0.6730 (0.6588)  loss_classifier: 0.2403 (0.2307)  loss_box_reg: 0.3037 (0.2960)  loss_objectness: 0.0782 (0.0714)  loss_rpn_box_reg: 0.0583 (0.0607)  time: 1.7348  data: 0.7113  max mem: 19870\n",
      "Epoch: [3]  [ 280/2444]  eta: 1:02:29  lr: 0.002000  loss: 0.6447 (0.6584)  loss_classifier: 0.2228 (0.2306)  loss_box_reg: 0.2969 (0.2960)  loss_objectness: 0.0541 (0.0711)  loss_rpn_box_reg: 0.0572 (0.0607)  time: 1.7244  data: 0.7015  max mem: 19870\n",
      "Epoch: [3]  [ 290/2444]  eta: 1:02:13  lr: 0.002000  loss: 0.6447 (0.6578)  loss_classifier: 0.2184 (0.2303)  loss_box_reg: 0.2963 (0.2956)  loss_objectness: 0.0578 (0.0711)  loss_rpn_box_reg: 0.0605 (0.0608)  time: 1.7378  data: 0.7144  max mem: 19870\n",
      "Epoch: [3]  [ 300/2444]  eta: 1:01:55  lr: 0.002000  loss: 0.6367 (0.6569)  loss_classifier: 0.2184 (0.2299)  loss_box_reg: 0.2871 (0.2951)  loss_objectness: 0.0712 (0.0711)  loss_rpn_box_reg: 0.0620 (0.0607)  time: 1.7362  data: 0.7120  max mem: 19870\n",
      "Epoch: [3]  [ 310/2444]  eta: 1:01:38  lr: 0.002000  loss: 0.6367 (0.6565)  loss_classifier: 0.2104 (0.2298)  loss_box_reg: 0.2836 (0.2949)  loss_objectness: 0.0712 (0.0712)  loss_rpn_box_reg: 0.0532 (0.0606)  time: 1.7276  data: 0.7036  max mem: 19870\n",
      "Epoch: [3]  [ 320/2444]  eta: 1:01:19  lr: 0.002000  loss: 0.6157 (0.6552)  loss_classifier: 0.2186 (0.2294)  loss_box_reg: 0.2814 (0.2945)  loss_objectness: 0.0645 (0.0710)  loss_rpn_box_reg: 0.0511 (0.0603)  time: 1.7198  data: 0.6961  max mem: 19870\n",
      "Epoch: [3]  [ 330/2444]  eta: 1:01:02  lr: 0.002000  loss: 0.6311 (0.6568)  loss_classifier: 0.2249 (0.2299)  loss_box_reg: 0.2901 (0.2950)  loss_objectness: 0.0698 (0.0715)  loss_rpn_box_reg: 0.0564 (0.0604)  time: 1.7286  data: 0.7040  max mem: 19870\n",
      "Epoch: [3]  [ 340/2444]  eta: 1:00:45  lr: 0.002000  loss: 0.6748 (0.6570)  loss_classifier: 0.2339 (0.2299)  loss_box_reg: 0.3136 (0.2951)  loss_objectness: 0.0698 (0.0715)  loss_rpn_box_reg: 0.0628 (0.0606)  time: 1.7386  data: 0.7132  max mem: 19870\n",
      "Epoch: [3]  [ 350/2444]  eta: 1:00:28  lr: 0.002000  loss: 0.6351 (0.6565)  loss_classifier: 0.2298 (0.2298)  loss_box_reg: 0.2911 (0.2950)  loss_objectness: 0.0616 (0.0713)  loss_rpn_box_reg: 0.0622 (0.0605)  time: 1.7315  data: 0.7063  max mem: 19870\n",
      "Epoch: [3]  [ 360/2444]  eta: 1:00:10  lr: 0.002000  loss: 0.6325 (0.6562)  loss_classifier: 0.2247 (0.2299)  loss_box_reg: 0.2911 (0.2950)  loss_objectness: 0.0574 (0.0709)  loss_rpn_box_reg: 0.0561 (0.0604)  time: 1.7304  data: 0.7048  max mem: 19870\n",
      "Epoch: [3]  [ 370/2444]  eta: 0:59:52  lr: 0.002000  loss: 0.6568 (0.6559)  loss_classifier: 0.2236 (0.2295)  loss_box_reg: 0.2810 (0.2946)  loss_objectness: 0.0692 (0.0713)  loss_rpn_box_reg: 0.0610 (0.0605)  time: 1.7215  data: 0.6962  max mem: 19870\n",
      "Epoch: [3]  [ 380/2444]  eta: 0:59:34  lr: 0.002000  loss: 0.6625 (0.6560)  loss_classifier: 0.2233 (0.2294)  loss_box_reg: 0.2877 (0.2945)  loss_objectness: 0.0839 (0.0714)  loss_rpn_box_reg: 0.0618 (0.0606)  time: 1.7258  data: 0.7012  max mem: 19870\n",
      "Epoch: [3]  [ 390/2444]  eta: 0:59:17  lr: 0.002000  loss: 0.6719 (0.6565)  loss_classifier: 0.2340 (0.2294)  loss_box_reg: 0.2938 (0.2948)  loss_objectness: 0.0690 (0.0716)  loss_rpn_box_reg: 0.0632 (0.0607)  time: 1.7358  data: 0.7117  max mem: 19870\n",
      "Epoch: [3]  [ 400/2444]  eta: 0:59:00  lr: 0.002000  loss: 0.6353 (0.6560)  loss_classifier: 0.2294 (0.2292)  loss_box_reg: 0.2911 (0.2944)  loss_objectness: 0.0690 (0.0716)  loss_rpn_box_reg: 0.0614 (0.0608)  time: 1.7335  data: 0.7106  max mem: 19870\n",
      "Epoch: [3]  [ 410/2444]  eta: 0:58:42  lr: 0.002000  loss: 0.6079 (0.6555)  loss_classifier: 0.2155 (0.2289)  loss_box_reg: 0.2842 (0.2941)  loss_objectness: 0.0649 (0.0717)  loss_rpn_box_reg: 0.0614 (0.0608)  time: 1.7243  data: 0.7014  max mem: 19870\n",
      "Epoch: [3]  [ 420/2444]  eta: 0:58:25  lr: 0.002000  loss: 0.6491 (0.6557)  loss_classifier: 0.2155 (0.2288)  loss_box_reg: 0.2883 (0.2942)  loss_objectness: 0.0707 (0.0719)  loss_rpn_box_reg: 0.0612 (0.0608)  time: 1.7242  data: 0.6996  max mem: 19870\n",
      "Epoch: [3]  [ 430/2444]  eta: 0:58:07  lr: 0.002000  loss: 0.6681 (0.6565)  loss_classifier: 0.2353 (0.2291)  loss_box_reg: 0.3063 (0.2944)  loss_objectness: 0.0709 (0.0721)  loss_rpn_box_reg: 0.0612 (0.0609)  time: 1.7274  data: 0.7014  max mem: 19870\n",
      "Epoch: [3]  [ 440/2444]  eta: 0:57:50  lr: 0.002000  loss: 0.6751 (0.6570)  loss_classifier: 0.2364 (0.2293)  loss_box_reg: 0.3056 (0.2946)  loss_objectness: 0.0645 (0.0721)  loss_rpn_box_reg: 0.0639 (0.0610)  time: 1.7302  data: 0.7049  max mem: 19870\n",
      "Epoch: [3]  [ 450/2444]  eta: 0:57:33  lr: 0.002000  loss: 0.6471 (0.6559)  loss_classifier: 0.2282 (0.2290)  loss_box_reg: 0.2934 (0.2942)  loss_objectness: 0.0611 (0.0719)  loss_rpn_box_reg: 0.0599 (0.0609)  time: 1.7366  data: 0.7123  max mem: 19870\n",
      "Epoch: [3]  [ 460/2444]  eta: 0:57:15  lr: 0.002000  loss: 0.6049 (0.6554)  loss_classifier: 0.2183 (0.2288)  loss_box_reg: 0.2818 (0.2941)  loss_objectness: 0.0665 (0.0716)  loss_rpn_box_reg: 0.0599 (0.0608)  time: 1.7336  data: 0.7086  max mem: 19870\n",
      "Epoch: [3]  [ 470/2444]  eta: 0:56:58  lr: 0.002000  loss: 0.6593 (0.6558)  loss_classifier: 0.2258 (0.2288)  loss_box_reg: 0.2908 (0.2942)  loss_objectness: 0.0690 (0.0719)  loss_rpn_box_reg: 0.0591 (0.0608)  time: 1.7335  data: 0.7081  max mem: 19870\n",
      "Epoch: [3]  [ 480/2444]  eta: 0:56:41  lr: 0.002000  loss: 0.6641 (0.6556)  loss_classifier: 0.2309 (0.2289)  loss_box_reg: 0.2859 (0.2939)  loss_objectness: 0.0763 (0.0720)  loss_rpn_box_reg: 0.0598 (0.0608)  time: 1.7410  data: 0.7078  max mem: 19870\n",
      "Epoch: [3]  [ 490/2444]  eta: 0:56:24  lr: 0.002000  loss: 0.6306 (0.6550)  loss_classifier: 0.2127 (0.2285)  loss_box_reg: 0.2712 (0.2935)  loss_objectness: 0.0763 (0.0721)  loss_rpn_box_reg: 0.0629 (0.0609)  time: 1.7334  data: 0.7002  max mem: 19870\n",
      "Epoch: [3]  [ 500/2444]  eta: 0:56:07  lr: 0.002000  loss: 0.6624 (0.6559)  loss_classifier: 0.2152 (0.2287)  loss_box_reg: 0.2893 (0.2940)  loss_objectness: 0.0697 (0.0722)  loss_rpn_box_reg: 0.0641 (0.0610)  time: 1.7330  data: 0.7077  max mem: 19870\n",
      "Epoch: [3]  [ 510/2444]  eta: 0:55:50  lr: 0.002000  loss: 0.6624 (0.6561)  loss_classifier: 0.2362 (0.2287)  loss_box_reg: 0.2951 (0.2940)  loss_objectness: 0.0697 (0.0724)  loss_rpn_box_reg: 0.0609 (0.0610)  time: 1.7384  data: 0.7123  max mem: 19870\n",
      "Epoch: [3]  [ 520/2444]  eta: 0:55:32  lr: 0.002000  loss: 0.6608 (0.6566)  loss_classifier: 0.2251 (0.2288)  loss_box_reg: 0.2931 (0.2941)  loss_objectness: 0.0800 (0.0726)  loss_rpn_box_reg: 0.0614 (0.0611)  time: 1.7333  data: 0.7059  max mem: 20046\n",
      "Epoch: [3]  [ 530/2444]  eta: 0:55:15  lr: 0.002000  loss: 0.6721 (0.6564)  loss_classifier: 0.2304 (0.2288)  loss_box_reg: 0.2932 (0.2941)  loss_objectness: 0.0686 (0.0725)  loss_rpn_box_reg: 0.0580 (0.0611)  time: 1.7274  data: 0.7004  max mem: 20046\n",
      "Epoch: [3]  [ 540/2444]  eta: 0:54:57  lr: 0.002000  loss: 0.6563 (0.6566)  loss_classifier: 0.2280 (0.2288)  loss_box_reg: 0.2932 (0.2941)  loss_objectness: 0.0715 (0.0727)  loss_rpn_box_reg: 0.0566 (0.0611)  time: 1.7204  data: 0.6956  max mem: 20046\n",
      "Epoch: [3]  [ 550/2444]  eta: 0:54:40  lr: 0.002000  loss: 0.6711 (0.6571)  loss_classifier: 0.2304 (0.2288)  loss_box_reg: 0.3028 (0.2943)  loss_objectness: 0.0727 (0.0728)  loss_rpn_box_reg: 0.0620 (0.0612)  time: 1.7273  data: 0.7025  max mem: 20046\n",
      "Epoch: [3]  [ 560/2444]  eta: 0:54:22  lr: 0.002000  loss: 0.6667 (0.6569)  loss_classifier: 0.2304 (0.2288)  loss_box_reg: 0.3015 (0.2942)  loss_objectness: 0.0652 (0.0729)  loss_rpn_box_reg: 0.0628 (0.0612)  time: 1.7282  data: 0.7032  max mem: 20046\n",
      "Epoch: [3]  [ 570/2444]  eta: 0:54:04  lr: 0.002000  loss: 0.6466 (0.6566)  loss_classifier: 0.2225 (0.2286)  loss_box_reg: 0.2969 (0.2941)  loss_objectness: 0.0630 (0.0728)  loss_rpn_box_reg: 0.0603 (0.0611)  time: 1.7199  data: 0.6954  max mem: 20046\n",
      "Epoch: [3]  [ 580/2444]  eta: 0:53:47  lr: 0.002000  loss: 0.6470 (0.6566)  loss_classifier: 0.2299 (0.2286)  loss_box_reg: 0.2969 (0.2940)  loss_objectness: 0.0677 (0.0728)  loss_rpn_box_reg: 0.0603 (0.0611)  time: 1.7257  data: 0.7016  max mem: 20046\n",
      "Epoch: [3]  [ 590/2444]  eta: 0:53:30  lr: 0.002000  loss: 0.6488 (0.6567)  loss_classifier: 0.2280 (0.2287)  loss_box_reg: 0.2960 (0.2941)  loss_objectness: 0.0677 (0.0728)  loss_rpn_box_reg: 0.0596 (0.0611)  time: 1.7449  data: 0.7212  max mem: 20046\n",
      "Epoch: [3]  [ 600/2444]  eta: 0:53:13  lr: 0.002000  loss: 0.6700 (0.6572)  loss_classifier: 0.2292 (0.2288)  loss_box_reg: 0.2975 (0.2944)  loss_objectness: 0.0652 (0.0729)  loss_rpn_box_reg: 0.0616 (0.0612)  time: 1.7388  data: 0.7146  max mem: 20046\n",
      "Epoch: [3]  [ 610/2444]  eta: 0:52:55  lr: 0.002000  loss: 0.6697 (0.6569)  loss_classifier: 0.2351 (0.2287)  loss_box_reg: 0.2916 (0.2943)  loss_objectness: 0.0681 (0.0727)  loss_rpn_box_reg: 0.0624 (0.0612)  time: 1.7207  data: 0.6955  max mem: 20046\n",
      "Epoch: [3]  [ 620/2444]  eta: 0:52:38  lr: 0.002000  loss: 0.6357 (0.6570)  loss_classifier: 0.2164 (0.2287)  loss_box_reg: 0.2892 (0.2943)  loss_objectness: 0.0688 (0.0728)  loss_rpn_box_reg: 0.0624 (0.0613)  time: 1.7280  data: 0.7021  max mem: 20046\n",
      "Epoch: [3]  [ 630/2444]  eta: 0:52:21  lr: 0.002000  loss: 0.6499 (0.6571)  loss_classifier: 0.2274 (0.2287)  loss_box_reg: 0.2943 (0.2944)  loss_objectness: 0.0725 (0.0727)  loss_rpn_box_reg: 0.0595 (0.0612)  time: 1.7363  data: 0.7109  max mem: 20046\n",
      "Epoch: [3]  [ 640/2444]  eta: 0:52:04  lr: 0.002000  loss: 0.6652 (0.6574)  loss_classifier: 0.2267 (0.2287)  loss_box_reg: 0.2947 (0.2944)  loss_objectness: 0.0741 (0.0729)  loss_rpn_box_reg: 0.0595 (0.0613)  time: 1.7379  data: 0.7116  max mem: 20046\n",
      "Epoch: [3]  [ 650/2444]  eta: 0:51:46  lr: 0.002000  loss: 0.6567 (0.6572)  loss_classifier: 0.2265 (0.2286)  loss_box_reg: 0.2938 (0.2943)  loss_objectness: 0.0754 (0.0730)  loss_rpn_box_reg: 0.0593 (0.0613)  time: 1.7335  data: 0.7071  max mem: 20046\n",
      "Epoch: [3]  [ 660/2444]  eta: 0:51:29  lr: 0.002000  loss: 0.6573 (0.6579)  loss_classifier: 0.2318 (0.2288)  loss_box_reg: 0.2952 (0.2945)  loss_objectness: 0.0680 (0.0733)  loss_rpn_box_reg: 0.0593 (0.0613)  time: 1.7333  data: 0.7077  max mem: 20046\n",
      "Epoch: [3]  [ 670/2444]  eta: 0:51:12  lr: 0.002000  loss: 0.6759 (0.6583)  loss_classifier: 0.2282 (0.2289)  loss_box_reg: 0.3086 (0.2946)  loss_objectness: 0.0667 (0.0734)  loss_rpn_box_reg: 0.0664 (0.0614)  time: 1.7362  data: 0.7108  max mem: 20046\n",
      "Epoch: [3]  [ 680/2444]  eta: 0:50:54  lr: 0.002000  loss: 0.6704 (0.6584)  loss_classifier: 0.2282 (0.2289)  loss_box_reg: 0.2928 (0.2947)  loss_objectness: 0.0674 (0.0733)  loss_rpn_box_reg: 0.0623 (0.0614)  time: 1.7320  data: 0.7056  max mem: 20046\n",
      "Epoch: [3]  [ 690/2444]  eta: 0:50:38  lr: 0.002000  loss: 0.6704 (0.6584)  loss_classifier: 0.2375 (0.2290)  loss_box_reg: 0.2950 (0.2948)  loss_objectness: 0.0674 (0.0733)  loss_rpn_box_reg: 0.0607 (0.0614)  time: 1.7445  data: 0.7103  max mem: 20046\n",
      "Epoch: [3]  [ 700/2444]  eta: 0:50:20  lr: 0.002000  loss: 0.6395 (0.6581)  loss_classifier: 0.2325 (0.2289)  loss_box_reg: 0.2904 (0.2946)  loss_objectness: 0.0687 (0.0733)  loss_rpn_box_reg: 0.0593 (0.0614)  time: 1.7427  data: 0.7096  max mem: 20046\n",
      "Epoch: [3]  [ 710/2444]  eta: 0:50:03  lr: 0.002000  loss: 0.6112 (0.6581)  loss_classifier: 0.2174 (0.2289)  loss_box_reg: 0.2823 (0.2945)  loss_objectness: 0.0628 (0.0733)  loss_rpn_box_reg: 0.0618 (0.0614)  time: 1.7295  data: 0.7048  max mem: 20046\n",
      "Epoch: [3]  [ 720/2444]  eta: 0:49:46  lr: 0.002000  loss: 0.6502 (0.6579)  loss_classifier: 0.2174 (0.2288)  loss_box_reg: 0.2884 (0.2944)  loss_objectness: 0.0715 (0.0733)  loss_rpn_box_reg: 0.0623 (0.0614)  time: 1.7301  data: 0.7062  max mem: 20046\n",
      "Epoch: [3]  [ 730/2444]  eta: 0:49:29  lr: 0.002000  loss: 0.6590 (0.6588)  loss_classifier: 0.2265 (0.2291)  loss_box_reg: 0.2967 (0.2948)  loss_objectness: 0.0708 (0.0733)  loss_rpn_box_reg: 0.0661 (0.0615)  time: 1.7410  data: 0.7166  max mem: 20046\n",
      "Epoch: [3]  [ 740/2444]  eta: 0:49:12  lr: 0.002000  loss: 0.6890 (0.6591)  loss_classifier: 0.2396 (0.2291)  loss_box_reg: 0.3053 (0.2950)  loss_objectness: 0.0692 (0.0734)  loss_rpn_box_reg: 0.0661 (0.0615)  time: 1.7461  data: 0.7216  max mem: 20046\n",
      "Epoch: [3]  [ 750/2444]  eta: 0:48:55  lr: 0.002000  loss: 0.6273 (0.6589)  loss_classifier: 0.2220 (0.2291)  loss_box_reg: 0.2928 (0.2949)  loss_objectness: 0.0679 (0.0733)  loss_rpn_box_reg: 0.0560 (0.0615)  time: 1.7395  data: 0.7156  max mem: 20046\n",
      "Epoch: [3]  [ 760/2444]  eta: 0:48:37  lr: 0.002000  loss: 0.6242 (0.6587)  loss_classifier: 0.2223 (0.2291)  loss_box_reg: 0.2886 (0.2950)  loss_objectness: 0.0565 (0.0732)  loss_rpn_box_reg: 0.0545 (0.0614)  time: 1.7372  data: 0.7127  max mem: 20046\n",
      "Epoch: [3]  [ 770/2444]  eta: 0:48:20  lr: 0.002000  loss: 0.6550 (0.6592)  loss_classifier: 0.2324 (0.2292)  loss_box_reg: 0.3016 (0.2951)  loss_objectness: 0.0712 (0.0734)  loss_rpn_box_reg: 0.0624 (0.0615)  time: 1.7331  data: 0.7098  max mem: 20046\n",
      "Epoch: [3]  [ 780/2444]  eta: 0:48:03  lr: 0.002000  loss: 0.6802 (0.6595)  loss_classifier: 0.2335 (0.2293)  loss_box_reg: 0.3048 (0.2953)  loss_objectness: 0.0737 (0.0734)  loss_rpn_box_reg: 0.0628 (0.0615)  time: 1.7388  data: 0.7167  max mem: 20046\n",
      "Epoch: [3]  [ 790/2444]  eta: 0:47:45  lr: 0.002000  loss: 0.6670 (0.6595)  loss_classifier: 0.2323 (0.2293)  loss_box_reg: 0.2993 (0.2952)  loss_objectness: 0.0718 (0.0735)  loss_rpn_box_reg: 0.0628 (0.0616)  time: 1.7365  data: 0.7147  max mem: 20046\n",
      "Epoch: [3]  [ 800/2444]  eta: 0:47:28  lr: 0.002000  loss: 0.6630 (0.6595)  loss_classifier: 0.2303 (0.2292)  loss_box_reg: 0.2935 (0.2951)  loss_objectness: 0.0745 (0.0736)  loss_rpn_box_reg: 0.0606 (0.0615)  time: 1.7247  data: 0.7020  max mem: 20046\n",
      "Epoch: [3]  [ 810/2444]  eta: 0:47:10  lr: 0.002000  loss: 0.6670 (0.6597)  loss_classifier: 0.2297 (0.2292)  loss_box_reg: 0.2934 (0.2952)  loss_objectness: 0.0745 (0.0736)  loss_rpn_box_reg: 0.0558 (0.0616)  time: 1.7242  data: 0.7007  max mem: 20046\n",
      "Epoch: [3]  [ 820/2444]  eta: 0:46:53  lr: 0.002000  loss: 0.6474 (0.6594)  loss_classifier: 0.2159 (0.2292)  loss_box_reg: 0.2854 (0.2952)  loss_objectness: 0.0668 (0.0735)  loss_rpn_box_reg: 0.0558 (0.0616)  time: 1.7292  data: 0.7056  max mem: 20046\n",
      "Epoch: [3]  [ 830/2444]  eta: 0:46:36  lr: 0.002000  loss: 0.6204 (0.6591)  loss_classifier: 0.2148 (0.2290)  loss_box_reg: 0.2854 (0.2950)  loss_objectness: 0.0668 (0.0735)  loss_rpn_box_reg: 0.0555 (0.0615)  time: 1.7341  data: 0.7089  max mem: 20046\n",
      "Epoch: [3]  [ 840/2444]  eta: 0:46:18  lr: 0.002000  loss: 0.6524 (0.6592)  loss_classifier: 0.2214 (0.2290)  loss_box_reg: 0.2869 (0.2951)  loss_objectness: 0.0698 (0.0735)  loss_rpn_box_reg: 0.0580 (0.0615)  time: 1.7347  data: 0.7084  max mem: 20046\n",
      "Epoch: [3]  [ 850/2444]  eta: 0:46:01  lr: 0.002000  loss: 0.6756 (0.6593)  loss_classifier: 0.2329 (0.2290)  loss_box_reg: 0.3010 (0.2952)  loss_objectness: 0.0701 (0.0735)  loss_rpn_box_reg: 0.0590 (0.0615)  time: 1.7342  data: 0.7074  max mem: 20046\n",
      "Epoch: [3]  [ 860/2444]  eta: 0:45:44  lr: 0.002000  loss: 0.6594 (0.6590)  loss_classifier: 0.2221 (0.2289)  loss_box_reg: 0.2916 (0.2952)  loss_objectness: 0.0746 (0.0734)  loss_rpn_box_reg: 0.0590 (0.0615)  time: 1.7274  data: 0.7012  max mem: 20046\n",
      "Epoch: [3]  [ 870/2444]  eta: 0:45:26  lr: 0.002000  loss: 0.6443 (0.6592)  loss_classifier: 0.2150 (0.2289)  loss_box_reg: 0.2836 (0.2951)  loss_objectness: 0.0746 (0.0736)  loss_rpn_box_reg: 0.0630 (0.0616)  time: 1.7190  data: 0.6938  max mem: 20046\n",
      "Epoch: [3]  [ 880/2444]  eta: 0:45:09  lr: 0.002000  loss: 0.6331 (0.6590)  loss_classifier: 0.2150 (0.2288)  loss_box_reg: 0.2822 (0.2950)  loss_objectness: 0.0719 (0.0737)  loss_rpn_box_reg: 0.0575 (0.0615)  time: 1.7313  data: 0.7053  max mem: 20046\n",
      "Epoch: [3]  [ 890/2444]  eta: 0:44:52  lr: 0.002000  loss: 0.6201 (0.6591)  loss_classifier: 0.2140 (0.2288)  loss_box_reg: 0.2829 (0.2950)  loss_objectness: 0.0706 (0.0737)  loss_rpn_box_reg: 0.0571 (0.0615)  time: 1.7386  data: 0.7118  max mem: 20046\n",
      "Epoch: [3]  [ 900/2444]  eta: 0:44:34  lr: 0.002000  loss: 0.6605 (0.6591)  loss_classifier: 0.2261 (0.2288)  loss_box_reg: 0.2931 (0.2950)  loss_objectness: 0.0703 (0.0737)  loss_rpn_box_reg: 0.0609 (0.0615)  time: 1.7336  data: 0.7069  max mem: 20046\n",
      "Epoch: [3]  [ 910/2444]  eta: 0:44:17  lr: 0.002000  loss: 0.6603 (0.6587)  loss_classifier: 0.2227 (0.2286)  loss_box_reg: 0.2931 (0.2949)  loss_objectness: 0.0695 (0.0737)  loss_rpn_box_reg: 0.0588 (0.0615)  time: 1.7262  data: 0.6979  max mem: 20046\n",
      "Epoch: [3]  [ 920/2444]  eta: 0:43:59  lr: 0.002000  loss: 0.6403 (0.6585)  loss_classifier: 0.2153 (0.2286)  loss_box_reg: 0.2856 (0.2948)  loss_objectness: 0.0626 (0.0736)  loss_rpn_box_reg: 0.0540 (0.0614)  time: 1.7191  data: 0.6930  max mem: 20046\n",
      "Epoch: [3]  [ 930/2444]  eta: 0:43:42  lr: 0.002000  loss: 0.6075 (0.6581)  loss_classifier: 0.2117 (0.2284)  loss_box_reg: 0.2856 (0.2947)  loss_objectness: 0.0603 (0.0735)  loss_rpn_box_reg: 0.0546 (0.0614)  time: 1.7226  data: 0.6991  max mem: 20046\n",
      "Epoch: [3]  [ 940/2444]  eta: 0:43:25  lr: 0.002000  loss: 0.6358 (0.6582)  loss_classifier: 0.2187 (0.2284)  loss_box_reg: 0.2746 (0.2947)  loss_objectness: 0.0715 (0.0736)  loss_rpn_box_reg: 0.0549 (0.0614)  time: 1.7381  data: 0.7132  max mem: 20046\n",
      "Epoch: [3]  [ 950/2444]  eta: 0:43:08  lr: 0.002000  loss: 0.6667 (0.6583)  loss_classifier: 0.2209 (0.2284)  loss_box_reg: 0.2991 (0.2948)  loss_objectness: 0.0749 (0.0736)  loss_rpn_box_reg: 0.0624 (0.0615)  time: 1.7444  data: 0.7191  max mem: 20046\n",
      "Epoch: [3]  [ 960/2444]  eta: 0:42:50  lr: 0.002000  loss: 0.6359 (0.6581)  loss_classifier: 0.2209 (0.2283)  loss_box_reg: 0.2927 (0.2947)  loss_objectness: 0.0672 (0.0736)  loss_rpn_box_reg: 0.0619 (0.0614)  time: 1.7392  data: 0.7144  max mem: 20046\n",
      "Epoch: [3]  [ 970/2444]  eta: 0:42:33  lr: 0.002000  loss: 0.6295 (0.6577)  loss_classifier: 0.2096 (0.2282)  loss_box_reg: 0.2893 (0.2946)  loss_objectness: 0.0660 (0.0735)  loss_rpn_box_reg: 0.0582 (0.0614)  time: 1.7311  data: 0.7077  max mem: 20046\n",
      "Epoch: [3]  [ 980/2444]  eta: 0:42:16  lr: 0.002000  loss: 0.5966 (0.6574)  loss_classifier: 0.2095 (0.2281)  loss_box_reg: 0.2842 (0.2946)  loss_objectness: 0.0660 (0.0734)  loss_rpn_box_reg: 0.0513 (0.0613)  time: 1.7269  data: 0.7030  max mem: 20046\n",
      "Epoch: [3]  [ 990/2444]  eta: 0:41:58  lr: 0.002000  loss: 0.6081 (0.6574)  loss_classifier: 0.2201 (0.2281)  loss_box_reg: 0.2891 (0.2946)  loss_objectness: 0.0603 (0.0734)  loss_rpn_box_reg: 0.0521 (0.0613)  time: 1.7323  data: 0.7063  max mem: 20046\n",
      "Epoch: [3]  [1000/2444]  eta: 0:41:41  lr: 0.002000  loss: 0.6441 (0.6573)  loss_classifier: 0.2233 (0.2280)  loss_box_reg: 0.2979 (0.2945)  loss_objectness: 0.0644 (0.0735)  loss_rpn_box_reg: 0.0571 (0.0613)  time: 1.7299  data: 0.7034  max mem: 20046\n",
      "Epoch: [3]  [1010/2444]  eta: 0:41:23  lr: 0.002000  loss: 0.6632 (0.6574)  loss_classifier: 0.2376 (0.2281)  loss_box_reg: 0.2940 (0.2946)  loss_objectness: 0.0610 (0.0735)  loss_rpn_box_reg: 0.0602 (0.0613)  time: 1.7278  data: 0.7006  max mem: 20046\n",
      "Epoch: [3]  [1020/2444]  eta: 0:41:06  lr: 0.002000  loss: 0.6558 (0.6571)  loss_classifier: 0.2176 (0.2279)  loss_box_reg: 0.2872 (0.2944)  loss_objectness: 0.0612 (0.0735)  loss_rpn_box_reg: 0.0640 (0.0613)  time: 1.7197  data: 0.6931  max mem: 20046\n",
      "Epoch: [3]  [1030/2444]  eta: 0:40:48  lr: 0.002000  loss: 0.6450 (0.6573)  loss_classifier: 0.2244 (0.2279)  loss_box_reg: 0.2891 (0.2944)  loss_objectness: 0.0699 (0.0736)  loss_rpn_box_reg: 0.0640 (0.0614)  time: 1.7194  data: 0.6941  max mem: 20046\n",
      "Epoch: [3]  [1040/2444]  eta: 0:40:31  lr: 0.002000  loss: 0.6729 (0.6572)  loss_classifier: 0.2280 (0.2279)  loss_box_reg: 0.2945 (0.2944)  loss_objectness: 0.0769 (0.0736)  loss_rpn_box_reg: 0.0620 (0.0614)  time: 1.7308  data: 0.7058  max mem: 20046\n",
      "Epoch: [3]  [1050/2444]  eta: 0:40:14  lr: 0.002000  loss: 0.6749 (0.6573)  loss_classifier: 0.2280 (0.2280)  loss_box_reg: 0.2834 (0.2944)  loss_objectness: 0.0751 (0.0736)  loss_rpn_box_reg: 0.0616 (0.0614)  time: 1.7329  data: 0.7081  max mem: 20046\n",
      "Epoch: [3]  [1060/2444]  eta: 0:39:56  lr: 0.002000  loss: 0.6680 (0.6574)  loss_classifier: 0.2131 (0.2280)  loss_box_reg: 0.2790 (0.2944)  loss_objectness: 0.0701 (0.0736)  loss_rpn_box_reg: 0.0611 (0.0614)  time: 1.7266  data: 0.7020  max mem: 20046\n",
      "Epoch: [3]  [1070/2444]  eta: 0:39:39  lr: 0.002000  loss: 0.6518 (0.6573)  loss_classifier: 0.2221 (0.2279)  loss_box_reg: 0.2910 (0.2944)  loss_objectness: 0.0593 (0.0736)  loss_rpn_box_reg: 0.0611 (0.0614)  time: 1.7279  data: 0.7042  max mem: 20046\n",
      "Epoch: [3]  [1080/2444]  eta: 0:39:22  lr: 0.002000  loss: 0.6753 (0.6576)  loss_classifier: 0.2323 (0.2281)  loss_box_reg: 0.3079 (0.2945)  loss_objectness: 0.0634 (0.0736)  loss_rpn_box_reg: 0.0645 (0.0614)  time: 1.7485  data: 0.7252  max mem: 20046\n",
      "Epoch: [3]  [1090/2444]  eta: 0:39:05  lr: 0.002000  loss: 0.6689 (0.6576)  loss_classifier: 0.2323 (0.2281)  loss_box_reg: 0.3022 (0.2946)  loss_objectness: 0.0706 (0.0736)  loss_rpn_box_reg: 0.0614 (0.0615)  time: 1.7439  data: 0.7199  max mem: 20046\n",
      "Epoch: [3]  [1100/2444]  eta: 0:38:48  lr: 0.002000  loss: 0.6618 (0.6580)  loss_classifier: 0.2306 (0.2281)  loss_box_reg: 0.2971 (0.2946)  loss_objectness: 0.0714 (0.0737)  loss_rpn_box_reg: 0.0646 (0.0615)  time: 1.7336  data: 0.7083  max mem: 20046\n",
      "Epoch: [3]  [1110/2444]  eta: 0:38:30  lr: 0.002000  loss: 0.6647 (0.6581)  loss_classifier: 0.2275 (0.2281)  loss_box_reg: 0.2925 (0.2946)  loss_objectness: 0.0805 (0.0738)  loss_rpn_box_reg: 0.0646 (0.0615)  time: 1.7366  data: 0.7106  max mem: 20046\n",
      "Epoch: [3]  [1120/2444]  eta: 0:38:13  lr: 0.002000  loss: 0.6570 (0.6580)  loss_classifier: 0.2253 (0.2280)  loss_box_reg: 0.2932 (0.2945)  loss_objectness: 0.0782 (0.0738)  loss_rpn_box_reg: 0.0616 (0.0616)  time: 1.7272  data: 0.7013  max mem: 20046\n",
      "Epoch: [3]  [1130/2444]  eta: 0:37:55  lr: 0.002000  loss: 0.6064 (0.6576)  loss_classifier: 0.2120 (0.2279)  loss_box_reg: 0.2818 (0.2944)  loss_objectness: 0.0607 (0.0738)  loss_rpn_box_reg: 0.0588 (0.0615)  time: 1.7259  data: 0.7010  max mem: 20046\n",
      "Epoch: [3]  [1140/2444]  eta: 0:37:38  lr: 0.002000  loss: 0.6343 (0.6578)  loss_classifier: 0.2204 (0.2279)  loss_box_reg: 0.2928 (0.2945)  loss_objectness: 0.0643 (0.0738)  loss_rpn_box_reg: 0.0591 (0.0615)  time: 1.7370  data: 0.7115  max mem: 20046\n",
      "Epoch: [3]  [1150/2444]  eta: 0:37:21  lr: 0.002000  loss: 0.6618 (0.6579)  loss_classifier: 0.2324 (0.2280)  loss_box_reg: 0.3011 (0.2946)  loss_objectness: 0.0667 (0.0738)  loss_rpn_box_reg: 0.0603 (0.0615)  time: 1.7474  data: 0.7221  max mem: 20046\n",
      "Epoch: [3]  [1160/2444]  eta: 0:37:04  lr: 0.002000  loss: 0.6449 (0.6577)  loss_classifier: 0.2292 (0.2280)  loss_box_reg: 0.2994 (0.2946)  loss_objectness: 0.0552 (0.0736)  loss_rpn_box_reg: 0.0573 (0.0615)  time: 1.7397  data: 0.7149  max mem: 20046\n",
      "Epoch: [3]  [1170/2444]  eta: 0:36:46  lr: 0.002000  loss: 0.6203 (0.6576)  loss_classifier: 0.2254 (0.2279)  loss_box_reg: 0.2877 (0.2946)  loss_objectness: 0.0627 (0.0736)  loss_rpn_box_reg: 0.0561 (0.0615)  time: 1.7309  data: 0.7061  max mem: 20046\n",
      "Epoch: [3]  [1180/2444]  eta: 0:36:29  lr: 0.002000  loss: 0.6522 (0.6576)  loss_classifier: 0.2295 (0.2279)  loss_box_reg: 0.2856 (0.2945)  loss_objectness: 0.0757 (0.0737)  loss_rpn_box_reg: 0.0604 (0.0615)  time: 1.7305  data: 0.7067  max mem: 20046\n",
      "Epoch: [3]  [1190/2444]  eta: 0:36:12  lr: 0.002000  loss: 0.6632 (0.6577)  loss_classifier: 0.2284 (0.2279)  loss_box_reg: 0.2843 (0.2946)  loss_objectness: 0.0757 (0.0737)  loss_rpn_box_reg: 0.0640 (0.0615)  time: 1.7340  data: 0.7100  max mem: 20046\n",
      "Epoch: [3]  [1200/2444]  eta: 0:35:55  lr: 0.002000  loss: 0.6719 (0.6581)  loss_classifier: 0.2313 (0.2281)  loss_box_reg: 0.2912 (0.2946)  loss_objectness: 0.0807 (0.0740)  loss_rpn_box_reg: 0.0628 (0.0615)  time: 1.7392  data: 0.7147  max mem: 20046\n",
      "Epoch: [3]  [1210/2444]  eta: 0:35:37  lr: 0.002000  loss: 0.6457 (0.6581)  loss_classifier: 0.2280 (0.2280)  loss_box_reg: 0.3010 (0.2946)  loss_objectness: 0.0688 (0.0739)  loss_rpn_box_reg: 0.0580 (0.0615)  time: 1.7340  data: 0.7095  max mem: 20046\n",
      "Epoch: [3]  [1220/2444]  eta: 0:35:20  lr: 0.002000  loss: 0.6399 (0.6581)  loss_classifier: 0.2280 (0.2281)  loss_box_reg: 0.2909 (0.2946)  loss_objectness: 0.0644 (0.0739)  loss_rpn_box_reg: 0.0597 (0.0615)  time: 1.7372  data: 0.7140  max mem: 20046\n",
      "Epoch: [3]  [1230/2444]  eta: 0:35:03  lr: 0.002000  loss: 0.6507 (0.6581)  loss_classifier: 0.2261 (0.2280)  loss_box_reg: 0.2892 (0.2945)  loss_objectness: 0.0638 (0.0739)  loss_rpn_box_reg: 0.0635 (0.0615)  time: 1.7347  data: 0.7104  max mem: 20046\n",
      "Epoch: [3]  [1240/2444]  eta: 0:34:45  lr: 0.002000  loss: 0.6609 (0.6581)  loss_classifier: 0.2261 (0.2281)  loss_box_reg: 0.2913 (0.2945)  loss_objectness: 0.0625 (0.0739)  loss_rpn_box_reg: 0.0629 (0.0616)  time: 1.7274  data: 0.7002  max mem: 20046\n",
      "Epoch: [3]  [1250/2444]  eta: 0:34:28  lr: 0.002000  loss: 0.6609 (0.6581)  loss_classifier: 0.2276 (0.2280)  loss_box_reg: 0.2913 (0.2945)  loss_objectness: 0.0708 (0.0740)  loss_rpn_box_reg: 0.0607 (0.0616)  time: 1.7329  data: 0.7054  max mem: 20046\n",
      "Epoch: [3]  [1260/2444]  eta: 0:34:11  lr: 0.002000  loss: 0.6661 (0.6583)  loss_classifier: 0.2344 (0.2281)  loss_box_reg: 0.3085 (0.2946)  loss_objectness: 0.0673 (0.0740)  loss_rpn_box_reg: 0.0607 (0.0616)  time: 1.7379  data: 0.7115  max mem: 20046\n",
      "Epoch: [3]  [1270/2444]  eta: 0:33:54  lr: 0.002000  loss: 0.6661 (0.6585)  loss_classifier: 0.2380 (0.2282)  loss_box_reg: 0.3093 (0.2946)  loss_objectness: 0.0682 (0.0740)  loss_rpn_box_reg: 0.0621 (0.0617)  time: 1.7396  data: 0.7147  max mem: 20046\n",
      "Epoch: [3]  [1280/2444]  eta: 0:33:36  lr: 0.002000  loss: 0.6633 (0.6585)  loss_classifier: 0.2249 (0.2282)  loss_box_reg: 0.2904 (0.2947)  loss_objectness: 0.0730 (0.0739)  loss_rpn_box_reg: 0.0621 (0.0617)  time: 1.7422  data: 0.7112  max mem: 20046\n",
      "Epoch: [3]  [1290/2444]  eta: 0:33:19  lr: 0.002000  loss: 0.6647 (0.6585)  loss_classifier: 0.2244 (0.2282)  loss_box_reg: 0.3003 (0.2947)  loss_objectness: 0.0728 (0.0739)  loss_rpn_box_reg: 0.0603 (0.0617)  time: 1.7371  data: 0.7052  max mem: 20046\n",
      "Epoch: [3]  [1300/2444]  eta: 0:33:02  lr: 0.002000  loss: 0.6576 (0.6583)  loss_classifier: 0.2312 (0.2282)  loss_box_reg: 0.3003 (0.2947)  loss_objectness: 0.0681 (0.0738)  loss_rpn_box_reg: 0.0586 (0.0617)  time: 1.7314  data: 0.7069  max mem: 20046\n",
      "Epoch: [3]  [1310/2444]  eta: 0:32:44  lr: 0.002000  loss: 0.6263 (0.6579)  loss_classifier: 0.2089 (0.2280)  loss_box_reg: 0.2801 (0.2945)  loss_objectness: 0.0579 (0.0738)  loss_rpn_box_reg: 0.0586 (0.0617)  time: 1.7257  data: 0.7006  max mem: 20046\n",
      "Epoch: [3]  [1320/2444]  eta: 0:32:27  lr: 0.002000  loss: 0.6287 (0.6579)  loss_classifier: 0.2185 (0.2280)  loss_box_reg: 0.2808 (0.2945)  loss_objectness: 0.0670 (0.0737)  loss_rpn_box_reg: 0.0574 (0.0616)  time: 1.7242  data: 0.6988  max mem: 20046\n",
      "Epoch: [3]  [1330/2444]  eta: 0:32:10  lr: 0.002000  loss: 0.6558 (0.6579)  loss_classifier: 0.2260 (0.2280)  loss_box_reg: 0.2857 (0.2945)  loss_objectness: 0.0761 (0.0738)  loss_rpn_box_reg: 0.0587 (0.0617)  time: 1.7347  data: 0.7095  max mem: 20046\n",
      "Epoch: [3]  [1340/2444]  eta: 0:31:52  lr: 0.002000  loss: 0.6489 (0.6579)  loss_classifier: 0.2260 (0.2280)  loss_box_reg: 0.2866 (0.2945)  loss_objectness: 0.0727 (0.0738)  loss_rpn_box_reg: 0.0594 (0.0616)  time: 1.7300  data: 0.7033  max mem: 20046\n",
      "Epoch: [3]  [1350/2444]  eta: 0:31:35  lr: 0.002000  loss: 0.6320 (0.6577)  loss_classifier: 0.2242 (0.2280)  loss_box_reg: 0.2915 (0.2944)  loss_objectness: 0.0563 (0.0737)  loss_rpn_box_reg: 0.0592 (0.0616)  time: 1.7196  data: 0.6936  max mem: 20046\n",
      "Epoch: [3]  [1360/2444]  eta: 0:31:17  lr: 0.002000  loss: 0.6137 (0.6577)  loss_classifier: 0.2212 (0.2280)  loss_box_reg: 0.2874 (0.2944)  loss_objectness: 0.0594 (0.0737)  loss_rpn_box_reg: 0.0614 (0.0617)  time: 1.7256  data: 0.7005  max mem: 20046\n",
      "Epoch: [3]  [1370/2444]  eta: 0:31:00  lr: 0.002000  loss: 0.6142 (0.6575)  loss_classifier: 0.2135 (0.2279)  loss_box_reg: 0.2874 (0.2943)  loss_objectness: 0.0597 (0.0736)  loss_rpn_box_reg: 0.0586 (0.0617)  time: 1.7278  data: 0.7033  max mem: 20046\n",
      "Epoch: [3]  [1380/2444]  eta: 0:30:43  lr: 0.002000  loss: 0.6570 (0.6577)  loss_classifier: 0.2272 (0.2280)  loss_box_reg: 0.2994 (0.2944)  loss_objectness: 0.0617 (0.0736)  loss_rpn_box_reg: 0.0610 (0.0617)  time: 1.7316  data: 0.7072  max mem: 20046\n",
      "Epoch: [3]  [1390/2444]  eta: 0:30:25  lr: 0.002000  loss: 0.6663 (0.6576)  loss_classifier: 0.2279 (0.2280)  loss_box_reg: 0.2928 (0.2943)  loss_objectness: 0.0665 (0.0736)  loss_rpn_box_reg: 0.0617 (0.0617)  time: 1.7341  data: 0.7094  max mem: 20046\n",
      "Epoch: [3]  [1400/2444]  eta: 0:30:08  lr: 0.002000  loss: 0.6663 (0.6576)  loss_classifier: 0.2244 (0.2280)  loss_box_reg: 0.2830 (0.2943)  loss_objectness: 0.0744 (0.0736)  loss_rpn_box_reg: 0.0617 (0.0617)  time: 1.7333  data: 0.7088  max mem: 20046\n",
      "Epoch: [3]  [1410/2444]  eta: 0:29:51  lr: 0.002000  loss: 0.6359 (0.6574)  loss_classifier: 0.2188 (0.2279)  loss_box_reg: 0.2848 (0.2943)  loss_objectness: 0.0648 (0.0736)  loss_rpn_box_reg: 0.0605 (0.0616)  time: 1.7280  data: 0.7046  max mem: 20046\n",
      "Epoch: [3]  [1420/2444]  eta: 0:29:33  lr: 0.002000  loss: 0.6315 (0.6575)  loss_classifier: 0.2204 (0.2279)  loss_box_reg: 0.2848 (0.2943)  loss_objectness: 0.0638 (0.0737)  loss_rpn_box_reg: 0.0587 (0.0616)  time: 1.7253  data: 0.7014  max mem: 20046\n",
      "Epoch: [3]  [1430/2444]  eta: 0:29:16  lr: 0.002000  loss: 0.6815 (0.6577)  loss_classifier: 0.2249 (0.2280)  loss_box_reg: 0.2970 (0.2944)  loss_objectness: 0.0808 (0.0737)  loss_rpn_box_reg: 0.0600 (0.0616)  time: 1.7337  data: 0.7081  max mem: 20046\n",
      "Epoch: [3]  [1440/2444]  eta: 0:28:59  lr: 0.002000  loss: 0.6405 (0.6575)  loss_classifier: 0.2249 (0.2280)  loss_box_reg: 0.2917 (0.2943)  loss_objectness: 0.0650 (0.0736)  loss_rpn_box_reg: 0.0616 (0.0616)  time: 1.7296  data: 0.7042  max mem: 20046\n",
      "Epoch: [3]  [1450/2444]  eta: 0:28:41  lr: 0.002000  loss: 0.6342 (0.6576)  loss_classifier: 0.2238 (0.2280)  loss_box_reg: 0.2830 (0.2943)  loss_objectness: 0.0598 (0.0736)  loss_rpn_box_reg: 0.0610 (0.0616)  time: 1.7253  data: 0.6999  max mem: 20046\n",
      "Epoch: [3]  [1460/2444]  eta: 0:28:24  lr: 0.002000  loss: 0.6380 (0.6575)  loss_classifier: 0.2255 (0.2280)  loss_box_reg: 0.2852 (0.2942)  loss_objectness: 0.0564 (0.0736)  loss_rpn_box_reg: 0.0579 (0.0616)  time: 1.7293  data: 0.7045  max mem: 20046\n",
      "Epoch: [3]  [1470/2444]  eta: 0:28:07  lr: 0.002000  loss: 0.6529 (0.6575)  loss_classifier: 0.2255 (0.2280)  loss_box_reg: 0.2852 (0.2942)  loss_objectness: 0.0709 (0.0736)  loss_rpn_box_reg: 0.0563 (0.0616)  time: 1.7477  data: 0.7236  max mem: 20046\n",
      "Epoch: [3]  [1480/2444]  eta: 0:27:50  lr: 0.002000  loss: 0.6590 (0.6577)  loss_classifier: 0.2242 (0.2281)  loss_box_reg: 0.2963 (0.2943)  loss_objectness: 0.0709 (0.0736)  loss_rpn_box_reg: 0.0574 (0.0616)  time: 1.7576  data: 0.7250  max mem: 20046\n",
      "Epoch: [3]  [1490/2444]  eta: 0:27:32  lr: 0.002000  loss: 0.6780 (0.6578)  loss_classifier: 0.2242 (0.2281)  loss_box_reg: 0.2993 (0.2944)  loss_objectness: 0.0755 (0.0737)  loss_rpn_box_reg: 0.0594 (0.0616)  time: 1.7529  data: 0.7178  max mem: 20046\n",
      "Epoch: [3]  [1500/2444]  eta: 0:27:15  lr: 0.002000  loss: 0.6734 (0.6579)  loss_classifier: 0.2276 (0.2281)  loss_box_reg: 0.2828 (0.2944)  loss_objectness: 0.0755 (0.0737)  loss_rpn_box_reg: 0.0623 (0.0616)  time: 1.7457  data: 0.7192  max mem: 20046\n",
      "Epoch: [3]  [1510/2444]  eta: 0:26:58  lr: 0.002000  loss: 0.6382 (0.6579)  loss_classifier: 0.2276 (0.2281)  loss_box_reg: 0.2789 (0.2944)  loss_objectness: 0.0718 (0.0738)  loss_rpn_box_reg: 0.0623 (0.0616)  time: 1.7326  data: 0.7075  max mem: 20046\n",
      "Epoch: [3]  [1520/2444]  eta: 0:26:40  lr: 0.002000  loss: 0.6437 (0.6580)  loss_classifier: 0.2270 (0.2281)  loss_box_reg: 0.2893 (0.2943)  loss_objectness: 0.0874 (0.0738)  loss_rpn_box_reg: 0.0639 (0.0617)  time: 1.7275  data: 0.7017  max mem: 20046\n",
      "Epoch: [3]  [1530/2444]  eta: 0:26:23  lr: 0.002000  loss: 0.6270 (0.6577)  loss_classifier: 0.2227 (0.2280)  loss_box_reg: 0.2767 (0.2942)  loss_objectness: 0.0696 (0.0738)  loss_rpn_box_reg: 0.0579 (0.0616)  time: 1.7274  data: 0.7020  max mem: 20046\n",
      "Epoch: [3]  [1540/2444]  eta: 0:26:06  lr: 0.002000  loss: 0.5960 (0.6572)  loss_classifier: 0.2070 (0.2279)  loss_box_reg: 0.2671 (0.2940)  loss_objectness: 0.0605 (0.0737)  loss_rpn_box_reg: 0.0564 (0.0616)  time: 1.7236  data: 0.6980  max mem: 20046\n",
      "Epoch: [3]  [1550/2444]  eta: 0:25:48  lr: 0.002000  loss: 0.6311 (0.6573)  loss_classifier: 0.2146 (0.2279)  loss_box_reg: 0.2860 (0.2941)  loss_objectness: 0.0547 (0.0737)  loss_rpn_box_reg: 0.0577 (0.0616)  time: 1.7357  data: 0.7085  max mem: 20046\n",
      "Epoch: [3]  [1560/2444]  eta: 0:25:31  lr: 0.002000  loss: 0.6539 (0.6572)  loss_classifier: 0.2262 (0.2278)  loss_box_reg: 0.3007 (0.2941)  loss_objectness: 0.0665 (0.0737)  loss_rpn_box_reg: 0.0599 (0.0616)  time: 1.7333  data: 0.7053  max mem: 20046\n",
      "Epoch: [3]  [1570/2444]  eta: 0:25:14  lr: 0.002000  loss: 0.6549 (0.6574)  loss_classifier: 0.2262 (0.2279)  loss_box_reg: 0.2963 (0.2941)  loss_objectness: 0.0764 (0.0738)  loss_rpn_box_reg: 0.0609 (0.0616)  time: 1.7313  data: 0.7047  max mem: 20046\n",
      "Epoch: [3]  [1580/2444]  eta: 0:24:56  lr: 0.002000  loss: 0.6508 (0.6572)  loss_classifier: 0.2238 (0.2278)  loss_box_reg: 0.2889 (0.2940)  loss_objectness: 0.0817 (0.0738)  loss_rpn_box_reg: 0.0664 (0.0616)  time: 1.7358  data: 0.7090  max mem: 20046\n",
      "Epoch: [3]  [1590/2444]  eta: 0:24:39  lr: 0.002000  loss: 0.6137 (0.6571)  loss_classifier: 0.2109 (0.2277)  loss_box_reg: 0.2757 (0.2940)  loss_objectness: 0.0729 (0.0738)  loss_rpn_box_reg: 0.0587 (0.0616)  time: 1.7199  data: 0.6938  max mem: 20046\n",
      "Epoch: [3]  [1600/2444]  eta: 0:24:22  lr: 0.002000  loss: 0.6570 (0.6572)  loss_classifier: 0.2347 (0.2277)  loss_box_reg: 0.2886 (0.2940)  loss_objectness: 0.0738 (0.0738)  loss_rpn_box_reg: 0.0605 (0.0616)  time: 1.7305  data: 0.7058  max mem: 20046\n",
      "Epoch: [3]  [1610/2444]  eta: 0:24:04  lr: 0.002000  loss: 0.6725 (0.6571)  loss_classifier: 0.2352 (0.2277)  loss_box_reg: 0.3110 (0.2940)  loss_objectness: 0.0643 (0.0737)  loss_rpn_box_reg: 0.0605 (0.0616)  time: 1.7364  data: 0.7112  max mem: 20046\n",
      "Epoch: [3]  [1620/2444]  eta: 0:23:47  lr: 0.002000  loss: 0.6414 (0.6571)  loss_classifier: 0.2262 (0.2277)  loss_box_reg: 0.2967 (0.2941)  loss_objectness: 0.0591 (0.0737)  loss_rpn_box_reg: 0.0566 (0.0616)  time: 1.7249  data: 0.7002  max mem: 20046\n",
      "Epoch: [3]  [1630/2444]  eta: 0:23:30  lr: 0.002000  loss: 0.6509 (0.6571)  loss_classifier: 0.2262 (0.2277)  loss_box_reg: 0.3009 (0.2941)  loss_objectness: 0.0622 (0.0736)  loss_rpn_box_reg: 0.0620 (0.0616)  time: 1.7292  data: 0.7047  max mem: 20046\n",
      "Epoch: [3]  [1640/2444]  eta: 0:23:12  lr: 0.002000  loss: 0.6509 (0.6573)  loss_classifier: 0.2297 (0.2278)  loss_box_reg: 0.3029 (0.2942)  loss_objectness: 0.0603 (0.0737)  loss_rpn_box_reg: 0.0633 (0.0616)  time: 1.7327  data: 0.7083  max mem: 20046\n",
      "Epoch: [3]  [1650/2444]  eta: 0:22:55  lr: 0.002000  loss: 0.6392 (0.6572)  loss_classifier: 0.2279 (0.2278)  loss_box_reg: 0.2899 (0.2941)  loss_objectness: 0.0603 (0.0737)  loss_rpn_box_reg: 0.0548 (0.0616)  time: 1.7288  data: 0.7036  max mem: 20046\n",
      "Epoch: [3]  [1660/2444]  eta: 0:22:38  lr: 0.002000  loss: 0.6290 (0.6573)  loss_classifier: 0.2279 (0.2278)  loss_box_reg: 0.2893 (0.2942)  loss_objectness: 0.0603 (0.0737)  loss_rpn_box_reg: 0.0583 (0.0616)  time: 1.7295  data: 0.7037  max mem: 20046\n",
      "Epoch: [3]  [1670/2444]  eta: 0:22:20  lr: 0.002000  loss: 0.6358 (0.6572)  loss_classifier: 0.2225 (0.2278)  loss_box_reg: 0.2873 (0.2941)  loss_objectness: 0.0723 (0.0737)  loss_rpn_box_reg: 0.0616 (0.0616)  time: 1.7301  data: 0.7047  max mem: 20046\n",
      "Epoch: [3]  [1680/2444]  eta: 0:22:03  lr: 0.002000  loss: 0.6389 (0.6573)  loss_classifier: 0.2231 (0.2278)  loss_box_reg: 0.2891 (0.2942)  loss_objectness: 0.0731 (0.0737)  loss_rpn_box_reg: 0.0617 (0.0616)  time: 1.7407  data: 0.7079  max mem: 20046\n",
      "Epoch: [3]  [1690/2444]  eta: 0:21:46  lr: 0.002000  loss: 0.6706 (0.6574)  loss_classifier: 0.2407 (0.2279)  loss_box_reg: 0.2971 (0.2942)  loss_objectness: 0.0728 (0.0737)  loss_rpn_box_reg: 0.0588 (0.0616)  time: 1.7457  data: 0.7140  max mem: 20046\n",
      "Epoch: [3]  [1700/2444]  eta: 0:21:28  lr: 0.002000  loss: 0.6447 (0.6572)  loss_classifier: 0.2252 (0.2278)  loss_box_reg: 0.2888 (0.2941)  loss_objectness: 0.0694 (0.0736)  loss_rpn_box_reg: 0.0546 (0.0616)  time: 1.7333  data: 0.7079  max mem: 20046\n",
      "Epoch: [3]  [1710/2444]  eta: 0:21:11  lr: 0.002000  loss: 0.6301 (0.6571)  loss_classifier: 0.2184 (0.2278)  loss_box_reg: 0.2874 (0.2942)  loss_objectness: 0.0638 (0.0736)  loss_rpn_box_reg: 0.0537 (0.0615)  time: 1.7372  data: 0.7114  max mem: 20046\n",
      "Epoch: [3]  [1720/2444]  eta: 0:20:54  lr: 0.002000  loss: 0.6358 (0.6571)  loss_classifier: 0.2247 (0.2278)  loss_box_reg: 0.2854 (0.2941)  loss_objectness: 0.0735 (0.0737)  loss_rpn_box_reg: 0.0561 (0.0615)  time: 1.7356  data: 0.7108  max mem: 20046\n",
      "Epoch: [3]  [1730/2444]  eta: 0:20:36  lr: 0.002000  loss: 0.6733 (0.6573)  loss_classifier: 0.2250 (0.2278)  loss_box_reg: 0.2904 (0.2941)  loss_objectness: 0.0785 (0.0738)  loss_rpn_box_reg: 0.0640 (0.0616)  time: 1.7218  data: 0.6979  max mem: 20046\n",
      "Epoch: [3]  [1740/2444]  eta: 0:20:19  lr: 0.002000  loss: 0.6779 (0.6574)  loss_classifier: 0.2268 (0.2278)  loss_box_reg: 0.2997 (0.2941)  loss_objectness: 0.0718 (0.0738)  loss_rpn_box_reg: 0.0621 (0.0616)  time: 1.7233  data: 0.6985  max mem: 20046\n",
      "Epoch: [3]  [1750/2444]  eta: 0:20:02  lr: 0.002000  loss: 0.6590 (0.6575)  loss_classifier: 0.2283 (0.2279)  loss_box_reg: 0.2997 (0.2942)  loss_objectness: 0.0670 (0.0738)  loss_rpn_box_reg: 0.0626 (0.0616)  time: 1.7383  data: 0.7129  max mem: 20046\n",
      "Epoch: [3]  [1760/2444]  eta: 0:19:45  lr: 0.002000  loss: 0.6426 (0.6573)  loss_classifier: 0.2334 (0.2279)  loss_box_reg: 0.2923 (0.2942)  loss_objectness: 0.0576 (0.0737)  loss_rpn_box_reg: 0.0582 (0.0616)  time: 1.7395  data: 0.7147  max mem: 20046\n",
      "Epoch: [3]  [1770/2444]  eta: 0:19:27  lr: 0.002000  loss: 0.6346 (0.6573)  loss_classifier: 0.2203 (0.2278)  loss_box_reg: 0.2930 (0.2941)  loss_objectness: 0.0671 (0.0737)  loss_rpn_box_reg: 0.0546 (0.0616)  time: 1.7247  data: 0.7004  max mem: 20046\n",
      "Epoch: [3]  [1780/2444]  eta: 0:19:10  lr: 0.002000  loss: 0.6570 (0.6574)  loss_classifier: 0.2237 (0.2279)  loss_box_reg: 0.2931 (0.2942)  loss_objectness: 0.0707 (0.0737)  loss_rpn_box_reg: 0.0623 (0.0616)  time: 1.7350  data: 0.7107  max mem: 20046\n",
      "Epoch: [3]  [1790/2444]  eta: 0:18:53  lr: 0.002000  loss: 0.6552 (0.6575)  loss_classifier: 0.2324 (0.2279)  loss_box_reg: 0.2931 (0.2942)  loss_objectness: 0.0676 (0.0737)  loss_rpn_box_reg: 0.0602 (0.0616)  time: 1.7412  data: 0.7170  max mem: 20046\n",
      "Epoch: [3]  [1800/2444]  eta: 0:18:35  lr: 0.002000  loss: 0.6552 (0.6575)  loss_classifier: 0.2331 (0.2279)  loss_box_reg: 0.2850 (0.2942)  loss_objectness: 0.0610 (0.0738)  loss_rpn_box_reg: 0.0602 (0.0616)  time: 1.7316  data: 0.7058  max mem: 20046\n",
      "Epoch: [3]  [1810/2444]  eta: 0:18:18  lr: 0.002000  loss: 0.6708 (0.6576)  loss_classifier: 0.2330 (0.2280)  loss_box_reg: 0.2853 (0.2942)  loss_objectness: 0.0606 (0.0738)  loss_rpn_box_reg: 0.0584 (0.0616)  time: 1.7333  data: 0.7076  max mem: 20046\n",
      "Epoch: [3]  [1820/2444]  eta: 0:18:01  lr: 0.002000  loss: 0.6822 (0.6578)  loss_classifier: 0.2353 (0.2281)  loss_box_reg: 0.3050 (0.2943)  loss_objectness: 0.0688 (0.0738)  loss_rpn_box_reg: 0.0611 (0.0616)  time: 1.7365  data: 0.7118  max mem: 20046\n",
      "Epoch: [3]  [1830/2444]  eta: 0:17:43  lr: 0.002000  loss: 0.6860 (0.6578)  loss_classifier: 0.2353 (0.2280)  loss_box_reg: 0.3054 (0.2943)  loss_objectness: 0.0696 (0.0738)  loss_rpn_box_reg: 0.0645 (0.0616)  time: 1.7296  data: 0.7044  max mem: 20046\n",
      "Epoch: [3]  [1840/2444]  eta: 0:17:26  lr: 0.002000  loss: 0.6548 (0.6578)  loss_classifier: 0.2296 (0.2281)  loss_box_reg: 0.2877 (0.2943)  loss_objectness: 0.0696 (0.0738)  loss_rpn_box_reg: 0.0627 (0.0617)  time: 1.7348  data: 0.7095  max mem: 20046\n",
      "Epoch: [3]  [1850/2444]  eta: 0:17:09  lr: 0.002000  loss: 0.6406 (0.6578)  loss_classifier: 0.2287 (0.2281)  loss_box_reg: 0.2926 (0.2943)  loss_objectness: 0.0684 (0.0738)  loss_rpn_box_reg: 0.0548 (0.0616)  time: 1.7382  data: 0.7131  max mem: 20046\n",
      "Epoch: [3]  [1860/2444]  eta: 0:16:51  lr: 0.002000  loss: 0.6253 (0.6577)  loss_classifier: 0.2225 (0.2281)  loss_box_reg: 0.2936 (0.2943)  loss_objectness: 0.0611 (0.0737)  loss_rpn_box_reg: 0.0541 (0.0616)  time: 1.7275  data: 0.7023  max mem: 20046\n",
      "Epoch: [3]  [1870/2444]  eta: 0:16:34  lr: 0.002000  loss: 0.6120 (0.6575)  loss_classifier: 0.2168 (0.2280)  loss_box_reg: 0.2754 (0.2942)  loss_objectness: 0.0645 (0.0737)  loss_rpn_box_reg: 0.0526 (0.0616)  time: 1.7270  data: 0.7023  max mem: 20046\n",
      "Epoch: [3]  [1880/2444]  eta: 0:16:17  lr: 0.002000  loss: 0.6438 (0.6577)  loss_classifier: 0.2291 (0.2280)  loss_box_reg: 0.2883 (0.2943)  loss_objectness: 0.0697 (0.0738)  loss_rpn_box_reg: 0.0540 (0.0616)  time: 1.7317  data: 0.7073  max mem: 20046\n",
      "Epoch: [3]  [1890/2444]  eta: 0:15:59  lr: 0.002000  loss: 0.6815 (0.6578)  loss_classifier: 0.2358 (0.2281)  loss_box_reg: 0.3011 (0.2943)  loss_objectness: 0.0833 (0.0738)  loss_rpn_box_reg: 0.0611 (0.0616)  time: 1.7345  data: 0.7099  max mem: 20046\n",
      "Epoch: [3]  [1900/2444]  eta: 0:15:42  lr: 0.002000  loss: 0.6542 (0.6579)  loss_classifier: 0.2296 (0.2281)  loss_box_reg: 0.3011 (0.2943)  loss_objectness: 0.0736 (0.0739)  loss_rpn_box_reg: 0.0625 (0.0616)  time: 1.7354  data: 0.7112  max mem: 20046\n",
      "Epoch: [3]  [1910/2444]  eta: 0:15:25  lr: 0.002000  loss: 0.6465 (0.6577)  loss_classifier: 0.2181 (0.2280)  loss_box_reg: 0.2801 (0.2943)  loss_objectness: 0.0659 (0.0738)  loss_rpn_box_reg: 0.0594 (0.0616)  time: 1.7270  data: 0.7022  max mem: 20046\n",
      "Epoch: [3]  [1920/2444]  eta: 0:15:07  lr: 0.002000  loss: 0.6205 (0.6575)  loss_classifier: 0.2179 (0.2280)  loss_box_reg: 0.2891 (0.2942)  loss_objectness: 0.0602 (0.0737)  loss_rpn_box_reg: 0.0588 (0.0616)  time: 1.7198  data: 0.6949  max mem: 20046\n",
      "Epoch: [3]  [1930/2444]  eta: 0:14:50  lr: 0.002000  loss: 0.6541 (0.6576)  loss_classifier: 0.2274 (0.2280)  loss_box_reg: 0.2929 (0.2943)  loss_objectness: 0.0636 (0.0737)  loss_rpn_box_reg: 0.0621 (0.0616)  time: 1.7332  data: 0.7074  max mem: 20046\n",
      "Epoch: [3]  [1940/2444]  eta: 0:14:33  lr: 0.002000  loss: 0.6500 (0.6574)  loss_classifier: 0.2204 (0.2279)  loss_box_reg: 0.2929 (0.2942)  loss_objectness: 0.0596 (0.0736)  loss_rpn_box_reg: 0.0602 (0.0616)  time: 1.7374  data: 0.7114  max mem: 20046\n",
      "Epoch: [3]  [1950/2444]  eta: 0:14:15  lr: 0.002000  loss: 0.6222 (0.6575)  loss_classifier: 0.2187 (0.2280)  loss_box_reg: 0.2812 (0.2942)  loss_objectness: 0.0578 (0.0737)  loss_rpn_box_reg: 0.0567 (0.0616)  time: 1.7407  data: 0.7150  max mem: 20046\n",
      "Epoch: [3]  [1960/2444]  eta: 0:13:58  lr: 0.002000  loss: 0.6577 (0.6576)  loss_classifier: 0.2256 (0.2280)  loss_box_reg: 0.2812 (0.2942)  loss_objectness: 0.0748 (0.0738)  loss_rpn_box_reg: 0.0574 (0.0616)  time: 1.7329  data: 0.7077  max mem: 20046\n",
      "Epoch: [3]  [1970/2444]  eta: 0:13:41  lr: 0.002000  loss: 0.6473 (0.6576)  loss_classifier: 0.2281 (0.2280)  loss_box_reg: 0.2869 (0.2942)  loss_objectness: 0.0600 (0.0738)  loss_rpn_box_reg: 0.0587 (0.0616)  time: 1.7250  data: 0.7001  max mem: 20046\n",
      "Epoch: [3]  [1980/2444]  eta: 0:13:23  lr: 0.002000  loss: 0.6665 (0.6577)  loss_classifier: 0.2377 (0.2280)  loss_box_reg: 0.2946 (0.2943)  loss_objectness: 0.0629 (0.0738)  loss_rpn_box_reg: 0.0625 (0.0616)  time: 1.7369  data: 0.7109  max mem: 20046\n",
      "Epoch: [3]  [1990/2444]  eta: 0:13:06  lr: 0.002000  loss: 0.6866 (0.6578)  loss_classifier: 0.2363 (0.2281)  loss_box_reg: 0.2972 (0.2943)  loss_objectness: 0.0696 (0.0738)  loss_rpn_box_reg: 0.0648 (0.0616)  time: 1.7423  data: 0.7168  max mem: 20046\n",
      "Epoch: [3]  [2000/2444]  eta: 0:12:49  lr: 0.002000  loss: 0.6561 (0.6577)  loss_classifier: 0.2265 (0.2280)  loss_box_reg: 0.2909 (0.2943)  loss_objectness: 0.0791 (0.0738)  loss_rpn_box_reg: 0.0614 (0.0616)  time: 1.7308  data: 0.7048  max mem: 20046\n",
      "Epoch: [3]  [2010/2444]  eta: 0:12:31  lr: 0.002000  loss: 0.6168 (0.6575)  loss_classifier: 0.2135 (0.2280)  loss_box_reg: 0.2870 (0.2942)  loss_objectness: 0.0699 (0.0738)  loss_rpn_box_reg: 0.0597 (0.0616)  time: 1.7255  data: 0.6985  max mem: 20046\n",
      "Epoch: [3]  [2020/2444]  eta: 0:12:14  lr: 0.002000  loss: 0.6332 (0.6575)  loss_classifier: 0.2135 (0.2279)  loss_box_reg: 0.2861 (0.2942)  loss_objectness: 0.0699 (0.0738)  loss_rpn_box_reg: 0.0603 (0.0616)  time: 1.7212  data: 0.6954  max mem: 20046\n",
      "Epoch: [3]  [2030/2444]  eta: 0:11:57  lr: 0.002000  loss: 0.6309 (0.6573)  loss_classifier: 0.2139 (0.2279)  loss_box_reg: 0.2808 (0.2941)  loss_objectness: 0.0663 (0.0737)  loss_rpn_box_reg: 0.0596 (0.0616)  time: 1.7229  data: 0.6986  max mem: 20046\n",
      "Epoch: [3]  [2040/2444]  eta: 0:11:39  lr: 0.002000  loss: 0.6267 (0.6572)  loss_classifier: 0.2160 (0.2279)  loss_box_reg: 0.2868 (0.2941)  loss_objectness: 0.0641 (0.0737)  loss_rpn_box_reg: 0.0536 (0.0616)  time: 1.7380  data: 0.7128  max mem: 20046\n",
      "Epoch: [3]  [2050/2444]  eta: 0:11:22  lr: 0.002000  loss: 0.6534 (0.6571)  loss_classifier: 0.2270 (0.2279)  loss_box_reg: 0.2904 (0.2941)  loss_objectness: 0.0665 (0.0737)  loss_rpn_box_reg: 0.0555 (0.0616)  time: 1.7368  data: 0.7115  max mem: 20046\n",
      "Epoch: [3]  [2060/2444]  eta: 0:11:05  lr: 0.002000  loss: 0.6775 (0.6572)  loss_classifier: 0.2296 (0.2279)  loss_box_reg: 0.2923 (0.2940)  loss_objectness: 0.0718 (0.0737)  loss_rpn_box_reg: 0.0582 (0.0616)  time: 1.7264  data: 0.7013  max mem: 20046\n",
      "Epoch: [3]  [2070/2444]  eta: 0:10:47  lr: 0.002000  loss: 0.6686 (0.6572)  loss_classifier: 0.2272 (0.2278)  loss_box_reg: 0.2923 (0.2940)  loss_objectness: 0.0725 (0.0737)  loss_rpn_box_reg: 0.0592 (0.0615)  time: 1.7254  data: 0.7000  max mem: 20046\n",
      "Epoch: [3]  [2080/2444]  eta: 0:10:30  lr: 0.002000  loss: 0.6490 (0.6571)  loss_classifier: 0.2259 (0.2278)  loss_box_reg: 0.2859 (0.2940)  loss_objectness: 0.0661 (0.0737)  loss_rpn_box_reg: 0.0576 (0.0615)  time: 1.7397  data: 0.7145  max mem: 20046\n",
      "Epoch: [3]  [2090/2444]  eta: 0:10:13  lr: 0.002000  loss: 0.6404 (0.6571)  loss_classifier: 0.2182 (0.2278)  loss_box_reg: 0.2860 (0.2940)  loss_objectness: 0.0594 (0.0737)  loss_rpn_box_reg: 0.0567 (0.0615)  time: 1.7433  data: 0.7176  max mem: 20046\n",
      "Epoch: [3]  [2100/2444]  eta: 0:09:55  lr: 0.002000  loss: 0.6093 (0.6568)  loss_classifier: 0.2136 (0.2277)  loss_box_reg: 0.2831 (0.2939)  loss_objectness: 0.0594 (0.0736)  loss_rpn_box_reg: 0.0550 (0.0615)  time: 1.7286  data: 0.7027  max mem: 20046\n",
      "Epoch: [3]  [2110/2444]  eta: 0:09:38  lr: 0.002000  loss: 0.6117 (0.6567)  loss_classifier: 0.2113 (0.2277)  loss_box_reg: 0.2785 (0.2939)  loss_objectness: 0.0535 (0.0736)  loss_rpn_box_reg: 0.0538 (0.0615)  time: 1.7219  data: 0.6967  max mem: 20046\n",
      "Epoch: [3]  [2120/2444]  eta: 0:09:21  lr: 0.002000  loss: 0.6386 (0.6568)  loss_classifier: 0.2181 (0.2277)  loss_box_reg: 0.2840 (0.2939)  loss_objectness: 0.0734 (0.0737)  loss_rpn_box_reg: 0.0563 (0.0615)  time: 1.7300  data: 0.7047  max mem: 20046\n",
      "Epoch: [3]  [2130/2444]  eta: 0:09:03  lr: 0.002000  loss: 0.6670 (0.6567)  loss_classifier: 0.2308 (0.2277)  loss_box_reg: 0.2905 (0.2939)  loss_objectness: 0.0731 (0.0737)  loss_rpn_box_reg: 0.0586 (0.0615)  time: 1.7370  data: 0.7123  max mem: 20046\n",
      "Epoch: [3]  [2140/2444]  eta: 0:08:46  lr: 0.002000  loss: 0.6249 (0.6566)  loss_classifier: 0.2209 (0.2276)  loss_box_reg: 0.2886 (0.2938)  loss_objectness: 0.0667 (0.0737)  loss_rpn_box_reg: 0.0595 (0.0615)  time: 1.7378  data: 0.7121  max mem: 20046\n",
      "Epoch: [3]  [2150/2444]  eta: 0:08:29  lr: 0.002000  loss: 0.6340 (0.6566)  loss_classifier: 0.2190 (0.2276)  loss_box_reg: 0.2905 (0.2938)  loss_objectness: 0.0690 (0.0737)  loss_rpn_box_reg: 0.0620 (0.0615)  time: 1.7327  data: 0.7062  max mem: 20046\n",
      "Epoch: [3]  [2160/2444]  eta: 0:08:12  lr: 0.002000  loss: 0.6485 (0.6566)  loss_classifier: 0.2283 (0.2276)  loss_box_reg: 0.2945 (0.2939)  loss_objectness: 0.0682 (0.0736)  loss_rpn_box_reg: 0.0617 (0.0615)  time: 1.7377  data: 0.7128  max mem: 20046\n",
      "Epoch: [3]  [2170/2444]  eta: 0:07:54  lr: 0.002000  loss: 0.6485 (0.6567)  loss_classifier: 0.2295 (0.2276)  loss_box_reg: 0.3037 (0.2939)  loss_objectness: 0.0655 (0.0736)  loss_rpn_box_reg: 0.0617 (0.0615)  time: 1.7367  data: 0.7127  max mem: 20046\n",
      "Epoch: [3]  [2180/2444]  eta: 0:07:37  lr: 0.002000  loss: 0.6796 (0.6567)  loss_classifier: 0.2249 (0.2276)  loss_box_reg: 0.2982 (0.2939)  loss_objectness: 0.0770 (0.0737)  loss_rpn_box_reg: 0.0634 (0.0616)  time: 1.7196  data: 0.6971  max mem: 20046\n",
      "Epoch: [3]  [2190/2444]  eta: 0:07:20  lr: 0.002000  loss: 0.6125 (0.6566)  loss_classifier: 0.2191 (0.2276)  loss_box_reg: 0.2864 (0.2938)  loss_objectness: 0.0770 (0.0737)  loss_rpn_box_reg: 0.0539 (0.0615)  time: 1.7153  data: 0.6920  max mem: 20046\n",
      "Epoch: [3]  [2200/2444]  eta: 0:07:02  lr: 0.002000  loss: 0.6365 (0.6566)  loss_classifier: 0.2227 (0.2276)  loss_box_reg: 0.2878 (0.2938)  loss_objectness: 0.0756 (0.0737)  loss_rpn_box_reg: 0.0525 (0.0615)  time: 1.7268  data: 0.7020  max mem: 20046\n",
      "Epoch: [3]  [2210/2444]  eta: 0:06:45  lr: 0.002000  loss: 0.6708 (0.6568)  loss_classifier: 0.2377 (0.2277)  loss_box_reg: 0.2993 (0.2939)  loss_objectness: 0.0756 (0.0738)  loss_rpn_box_reg: 0.0630 (0.0615)  time: 1.7362  data: 0.7125  max mem: 20046\n",
      "Epoch: [3]  [2220/2444]  eta: 0:06:28  lr: 0.002000  loss: 0.6807 (0.6568)  loss_classifier: 0.2401 (0.2277)  loss_box_reg: 0.3008 (0.2938)  loss_objectness: 0.0763 (0.0738)  loss_rpn_box_reg: 0.0563 (0.0615)  time: 1.7311  data: 0.7070  max mem: 20046\n",
      "Epoch: [3]  [2230/2444]  eta: 0:06:10  lr: 0.002000  loss: 0.6519 (0.6568)  loss_classifier: 0.2310 (0.2277)  loss_box_reg: 0.2870 (0.2938)  loss_objectness: 0.0725 (0.0738)  loss_rpn_box_reg: 0.0563 (0.0615)  time: 1.7414  data: 0.7151  max mem: 20046\n",
      "Epoch: [3]  [2240/2444]  eta: 0:05:53  lr: 0.002000  loss: 0.6358 (0.6567)  loss_classifier: 0.2310 (0.2277)  loss_box_reg: 0.2836 (0.2938)  loss_objectness: 0.0654 (0.0737)  loss_rpn_box_reg: 0.0579 (0.0615)  time: 1.7463  data: 0.7206  max mem: 20046\n",
      "Epoch: [3]  [2250/2444]  eta: 0:05:36  lr: 0.002000  loss: 0.6358 (0.6568)  loss_classifier: 0.2361 (0.2277)  loss_box_reg: 0.2836 (0.2938)  loss_objectness: 0.0698 (0.0737)  loss_rpn_box_reg: 0.0599 (0.0615)  time: 1.7388  data: 0.7140  max mem: 20046\n",
      "Epoch: [3]  [2260/2444]  eta: 0:05:18  lr: 0.002000  loss: 0.6313 (0.6566)  loss_classifier: 0.2184 (0.2277)  loss_box_reg: 0.2836 (0.2938)  loss_objectness: 0.0715 (0.0737)  loss_rpn_box_reg: 0.0627 (0.0615)  time: 1.7323  data: 0.7080  max mem: 20046\n",
      "Epoch: [3]  [2270/2444]  eta: 0:05:01  lr: 0.002000  loss: 0.6309 (0.6566)  loss_classifier: 0.2177 (0.2277)  loss_box_reg: 0.2888 (0.2938)  loss_objectness: 0.0642 (0.0736)  loss_rpn_box_reg: 0.0620 (0.0615)  time: 1.7293  data: 0.7059  max mem: 20046\n",
      "Epoch: [3]  [2280/2444]  eta: 0:04:44  lr: 0.002000  loss: 0.6383 (0.6565)  loss_classifier: 0.2270 (0.2277)  loss_box_reg: 0.2937 (0.2938)  loss_objectness: 0.0643 (0.0736)  loss_rpn_box_reg: 0.0599 (0.0615)  time: 1.7419  data: 0.7101  max mem: 20046\n",
      "Epoch: [3]  [2290/2444]  eta: 0:04:26  lr: 0.002000  loss: 0.6383 (0.6565)  loss_classifier: 0.2204 (0.2276)  loss_box_reg: 0.2868 (0.2937)  loss_objectness: 0.0643 (0.0736)  loss_rpn_box_reg: 0.0592 (0.0615)  time: 1.7377  data: 0.7056  max mem: 20046\n",
      "Epoch: [3]  [2300/2444]  eta: 0:04:09  lr: 0.002000  loss: 0.6259 (0.6564)  loss_classifier: 0.2155 (0.2276)  loss_box_reg: 0.2815 (0.2937)  loss_objectness: 0.0712 (0.0736)  loss_rpn_box_reg: 0.0588 (0.0615)  time: 1.7314  data: 0.7074  max mem: 20046\n",
      "Epoch: [3]  [2310/2444]  eta: 0:03:52  lr: 0.002000  loss: 0.6357 (0.6563)  loss_classifier: 0.2242 (0.2276)  loss_box_reg: 0.2859 (0.2937)  loss_objectness: 0.0646 (0.0736)  loss_rpn_box_reg: 0.0535 (0.0615)  time: 1.7321  data: 0.7088  max mem: 20046\n",
      "Epoch: [3]  [2320/2444]  eta: 0:03:34  lr: 0.002000  loss: 0.6406 (0.6564)  loss_classifier: 0.2320 (0.2276)  loss_box_reg: 0.3055 (0.2938)  loss_objectness: 0.0646 (0.0736)  loss_rpn_box_reg: 0.0561 (0.0615)  time: 1.7351  data: 0.7121  max mem: 20046\n",
      "Epoch: [3]  [2330/2444]  eta: 0:03:17  lr: 0.002000  loss: 0.6816 (0.6564)  loss_classifier: 0.2384 (0.2276)  loss_box_reg: 0.3070 (0.2938)  loss_objectness: 0.0719 (0.0736)  loss_rpn_box_reg: 0.0602 (0.0614)  time: 1.7372  data: 0.7141  max mem: 20046\n",
      "Epoch: [3]  [2340/2444]  eta: 0:03:00  lr: 0.002000  loss: 0.6432 (0.6563)  loss_classifier: 0.2291 (0.2276)  loss_box_reg: 0.2864 (0.2937)  loss_objectness: 0.0706 (0.0736)  loss_rpn_box_reg: 0.0602 (0.0614)  time: 1.7260  data: 0.7013  max mem: 20046\n",
      "Epoch: [3]  [2350/2444]  eta: 0:02:42  lr: 0.002000  loss: 0.6458 (0.6564)  loss_classifier: 0.2352 (0.2276)  loss_box_reg: 0.2874 (0.2937)  loss_objectness: 0.0719 (0.0737)  loss_rpn_box_reg: 0.0610 (0.0614)  time: 1.7287  data: 0.7035  max mem: 20046\n",
      "Epoch: [3]  [2360/2444]  eta: 0:02:25  lr: 0.002000  loss: 0.6458 (0.6564)  loss_classifier: 0.2385 (0.2276)  loss_box_reg: 0.2889 (0.2937)  loss_objectness: 0.0719 (0.0736)  loss_rpn_box_reg: 0.0554 (0.0614)  time: 1.7300  data: 0.7066  max mem: 20046\n",
      "Epoch: [3]  [2370/2444]  eta: 0:02:08  lr: 0.002000  loss: 0.6439 (0.6566)  loss_classifier: 0.2278 (0.2277)  loss_box_reg: 0.2889 (0.2937)  loss_objectness: 0.0741 (0.0737)  loss_rpn_box_reg: 0.0585 (0.0615)  time: 1.7274  data: 0.7045  max mem: 20046\n",
      "Epoch: [3]  [2380/2444]  eta: 0:01:50  lr: 0.002000  loss: 0.6531 (0.6565)  loss_classifier: 0.2312 (0.2277)  loss_box_reg: 0.3022 (0.2938)  loss_objectness: 0.0709 (0.0737)  loss_rpn_box_reg: 0.0617 (0.0614)  time: 1.7300  data: 0.7061  max mem: 20046\n",
      "Epoch: [3]  [2390/2444]  eta: 0:01:33  lr: 0.002000  loss: 0.6258 (0.6565)  loss_classifier: 0.2167 (0.2276)  loss_box_reg: 0.2898 (0.2937)  loss_objectness: 0.0635 (0.0736)  loss_rpn_box_reg: 0.0570 (0.0614)  time: 1.7269  data: 0.7019  max mem: 20046\n",
      "Epoch: [3]  [2400/2444]  eta: 0:01:16  lr: 0.002000  loss: 0.6284 (0.6564)  loss_classifier: 0.2142 (0.2276)  loss_box_reg: 0.2836 (0.2937)  loss_objectness: 0.0635 (0.0736)  loss_rpn_box_reg: 0.0565 (0.0614)  time: 1.7273  data: 0.7016  max mem: 20046\n",
      "Epoch: [3]  [2410/2444]  eta: 0:00:58  lr: 0.002000  loss: 0.6324 (0.6564)  loss_classifier: 0.2124 (0.2276)  loss_box_reg: 0.2836 (0.2937)  loss_objectness: 0.0712 (0.0736)  loss_rpn_box_reg: 0.0603 (0.0614)  time: 1.7311  data: 0.7057  max mem: 20046\n",
      "Epoch: [3]  [2420/2444]  eta: 0:00:41  lr: 0.002000  loss: 0.6205 (0.6563)  loss_classifier: 0.2124 (0.2276)  loss_box_reg: 0.2807 (0.2937)  loss_objectness: 0.0626 (0.0736)  loss_rpn_box_reg: 0.0667 (0.0614)  time: 1.7311  data: 0.7056  max mem: 20046\n",
      "Epoch: [3]  [2430/2444]  eta: 0:00:24  lr: 0.002000  loss: 0.6369 (0.6563)  loss_classifier: 0.2189 (0.2276)  loss_box_reg: 0.2807 (0.2937)  loss_objectness: 0.0561 (0.0736)  loss_rpn_box_reg: 0.0616 (0.0615)  time: 1.7246  data: 0.6999  max mem: 20046\n",
      "Epoch: [3]  [2440/2444]  eta: 0:00:06  lr: 0.002000  loss: 0.6369 (0.6562)  loss_classifier: 0.2211 (0.2276)  loss_box_reg: 0.2950 (0.2936)  loss_objectness: 0.0561 (0.0736)  loss_rpn_box_reg: 0.0592 (0.0615)  time: 1.7276  data: 0.7029  max mem: 20046\n",
      "Epoch: [3]  [2443/2444]  eta: 0:00:01  lr: 0.002000  loss: 0.6184 (0.6564)  loss_classifier: 0.2200 (0.2277)  loss_box_reg: 0.2842 (0.2937)  loss_objectness: 0.0578 (0.0736)  loss_rpn_box_reg: 0.0556 (0.0615)  time: 1.6433  data: 0.6664  max mem: 20046\n",
      "Epoch: [3] Total time: 1:10:32 (1.7318 s / it)\n",
      "creating index...\n",
      "index created!\n",
      "Test:  [   0/4952]  eta: 0:06:59  model_time: 0.0464 (0.0464)  evaluator_time: 0.0161 (0.0161)  time: 0.0847  data: 0.0210  max mem: 20046\n",
      "Test:  [ 100/4952]  eta: 0:04:59  model_time: 0.0345 (0.0348)  evaluator_time: 0.0092 (0.0107)  time: 0.0645  data: 0.0176  max mem: 20046\n",
      "Test:  [ 200/4952]  eta: 0:04:59  model_time: 0.0343 (0.0349)  evaluator_time: 0.0061 (0.0117)  time: 0.0627  data: 0.0158  max mem: 20046\n",
      "Test:  [ 300/4952]  eta: 0:04:50  model_time: 0.0339 (0.0346)  evaluator_time: 0.0083 (0.0113)  time: 0.0648  data: 0.0167  max mem: 20046\n",
      "Test:  [ 400/4952]  eta: 0:04:43  model_time: 0.0334 (0.0346)  evaluator_time: 0.0069 (0.0110)  time: 0.0583  data: 0.0148  max mem: 20046\n",
      "Test:  [ 500/4952]  eta: 0:04:35  model_time: 0.0342 (0.0346)  evaluator_time: 0.0077 (0.0108)  time: 0.0625  data: 0.0159  max mem: 20046\n",
      "Test:  [ 600/4952]  eta: 0:04:30  model_time: 0.0343 (0.0347)  evaluator_time: 0.0060 (0.0111)  time: 0.0595  data: 0.0150  max mem: 20046\n",
      "Test:  [ 700/4952]  eta: 0:04:25  model_time: 0.0338 (0.0347)  evaluator_time: 0.0066 (0.0111)  time: 0.0602  data: 0.0155  max mem: 20046\n",
      "Test:  [ 800/4952]  eta: 0:04:20  model_time: 0.0341 (0.0347)  evaluator_time: 0.0094 (0.0115)  time: 0.0638  data: 0.0165  max mem: 20046\n",
      "Test:  [ 900/4952]  eta: 0:04:13  model_time: 0.0345 (0.0347)  evaluator_time: 0.0058 (0.0114)  time: 0.0598  data: 0.0149  max mem: 20046\n",
      "Test:  [1000/4952]  eta: 0:04:06  model_time: 0.0340 (0.0346)  evaluator_time: 0.0073 (0.0112)  time: 0.0580  data: 0.0144  max mem: 20046\n",
      "Test:  [1100/4952]  eta: 0:04:00  model_time: 0.0343 (0.0347)  evaluator_time: 0.0068 (0.0113)  time: 0.0585  data: 0.0150  max mem: 20046\n",
      "Test:  [1200/4952]  eta: 0:03:54  model_time: 0.0341 (0.0347)  evaluator_time: 0.0118 (0.0113)  time: 0.0680  data: 0.0174  max mem: 20046\n",
      "Test:  [1300/4952]  eta: 0:03:48  model_time: 0.0344 (0.0347)  evaluator_time: 0.0064 (0.0113)  time: 0.0634  data: 0.0158  max mem: 20046\n",
      "Test:  [1400/4952]  eta: 0:03:42  model_time: 0.0344 (0.0347)  evaluator_time: 0.0063 (0.0113)  time: 0.0626  data: 0.0156  max mem: 20046\n",
      "Test:  [1500/4952]  eta: 0:03:36  model_time: 0.0335 (0.0347)  evaluator_time: 0.0078 (0.0113)  time: 0.0629  data: 0.0165  max mem: 20046\n",
      "Test:  [1600/4952]  eta: 0:03:30  model_time: 0.0343 (0.0347)  evaluator_time: 0.0078 (0.0113)  time: 0.0649  data: 0.0165  max mem: 20046\n",
      "Test:  [1700/4952]  eta: 0:03:24  model_time: 0.0339 (0.0347)  evaluator_time: 0.0058 (0.0113)  time: 0.0619  data: 0.0158  max mem: 20046\n",
      "Test:  [1800/4952]  eta: 0:03:17  model_time: 0.0348 (0.0347)  evaluator_time: 0.0063 (0.0112)  time: 0.0624  data: 0.0152  max mem: 20046\n",
      "Test:  [1900/4952]  eta: 0:03:11  model_time: 0.0337 (0.0347)  evaluator_time: 0.0077 (0.0112)  time: 0.0601  data: 0.0153  max mem: 20046\n",
      "Test:  [2000/4952]  eta: 0:03:04  model_time: 0.0350 (0.0347)  evaluator_time: 0.0109 (0.0112)  time: 0.0668  data: 0.0168  max mem: 20046\n",
      "Test:  [2100/4952]  eta: 0:02:58  model_time: 0.0342 (0.0347)  evaluator_time: 0.0076 (0.0112)  time: 0.0631  data: 0.0163  max mem: 20046\n",
      "Test:  [2200/4952]  eta: 0:02:51  model_time: 0.0337 (0.0347)  evaluator_time: 0.0068 (0.0111)  time: 0.0651  data: 0.0162  max mem: 20046\n",
      "Test:  [2300/4952]  eta: 0:02:45  model_time: 0.0341 (0.0347)  evaluator_time: 0.0075 (0.0111)  time: 0.0646  data: 0.0163  max mem: 20046\n",
      "Test:  [2400/4952]  eta: 0:02:39  model_time: 0.0351 (0.0347)  evaluator_time: 0.0107 (0.0111)  time: 0.0686  data: 0.0172  max mem: 20046\n",
      "Test:  [2500/4952]  eta: 0:02:33  model_time: 0.0343 (0.0347)  evaluator_time: 0.0080 (0.0112)  time: 0.0650  data: 0.0176  max mem: 20046\n",
      "Test:  [2600/4952]  eta: 0:02:27  model_time: 0.0347 (0.0347)  evaluator_time: 0.0078 (0.0112)  time: 0.0625  data: 0.0150  max mem: 20046\n",
      "Test:  [2700/4952]  eta: 0:02:20  model_time: 0.0342 (0.0347)  evaluator_time: 0.0086 (0.0112)  time: 0.0613  data: 0.0149  max mem: 20046\n",
      "Test:  [2800/4952]  eta: 0:02:14  model_time: 0.0341 (0.0347)  evaluator_time: 0.0084 (0.0112)  time: 0.0629  data: 0.0154  max mem: 20046\n",
      "Test:  [2900/4952]  eta: 0:02:08  model_time: 0.0340 (0.0346)  evaluator_time: 0.0066 (0.0112)  time: 0.0644  data: 0.0168  max mem: 20046\n",
      "Test:  [3000/4952]  eta: 0:02:01  model_time: 0.0346 (0.0347)  evaluator_time: 0.0084 (0.0112)  time: 0.0634  data: 0.0158  max mem: 20046\n",
      "Test:  [3100/4952]  eta: 0:01:55  model_time: 0.0338 (0.0346)  evaluator_time: 0.0074 (0.0111)  time: 0.0638  data: 0.0161  max mem: 20046\n",
      "Test:  [3200/4952]  eta: 0:01:49  model_time: 0.0348 (0.0347)  evaluator_time: 0.0085 (0.0111)  time: 0.0655  data: 0.0165  max mem: 20046\n",
      "Test:  [3300/4952]  eta: 0:01:43  model_time: 0.0346 (0.0347)  evaluator_time: 0.0075 (0.0112)  time: 0.0634  data: 0.0163  max mem: 20046\n",
      "Test:  [3400/4952]  eta: 0:01:36  model_time: 0.0340 (0.0347)  evaluator_time: 0.0057 (0.0112)  time: 0.0632  data: 0.0161  max mem: 20046\n",
      "Test:  [3500/4952]  eta: 0:01:30  model_time: 0.0331 (0.0347)  evaluator_time: 0.0072 (0.0112)  time: 0.0567  data: 0.0140  max mem: 20046\n",
      "Test:  [3600/4952]  eta: 0:01:24  model_time: 0.0337 (0.0346)  evaluator_time: 0.0077 (0.0112)  time: 0.0606  data: 0.0161  max mem: 20046\n",
      "Test:  [3700/4952]  eta: 0:01:18  model_time: 0.0335 (0.0346)  evaluator_time: 0.0065 (0.0112)  time: 0.0582  data: 0.0148  max mem: 20046\n",
      "Test:  [3800/4952]  eta: 0:01:12  model_time: 0.0342 (0.0346)  evaluator_time: 0.0062 (0.0112)  time: 0.0574  data: 0.0142  max mem: 20046\n",
      "Test:  [3900/4952]  eta: 0:01:05  model_time: 0.0336 (0.0346)  evaluator_time: 0.0076 (0.0111)  time: 0.0599  data: 0.0150  max mem: 20046\n",
      "Test:  [4000/4952]  eta: 0:00:59  model_time: 0.0341 (0.0346)  evaluator_time: 0.0093 (0.0111)  time: 0.0632  data: 0.0161  max mem: 20046\n",
      "Test:  [4100/4952]  eta: 0:00:53  model_time: 0.0344 (0.0346)  evaluator_time: 0.0072 (0.0111)  time: 0.0617  data: 0.0158  max mem: 20046\n",
      "Test:  [4200/4952]  eta: 0:00:46  model_time: 0.0347 (0.0346)  evaluator_time: 0.0068 (0.0111)  time: 0.0607  data: 0.0144  max mem: 20046\n",
      "Test:  [4300/4952]  eta: 0:00:40  model_time: 0.0336 (0.0346)  evaluator_time: 0.0065 (0.0111)  time: 0.0582  data: 0.0142  max mem: 20046\n",
      "Test:  [4400/4952]  eta: 0:00:34  model_time: 0.0340 (0.0346)  evaluator_time: 0.0095 (0.0111)  time: 0.0610  data: 0.0156  max mem: 20046\n",
      "Test:  [4500/4952]  eta: 0:00:28  model_time: 0.0342 (0.0346)  evaluator_time: 0.0074 (0.0111)  time: 0.0636  data: 0.0162  max mem: 20046\n",
      "Test:  [4600/4952]  eta: 0:00:21  model_time: 0.0337 (0.0346)  evaluator_time: 0.0072 (0.0111)  time: 0.0592  data: 0.0155  max mem: 20046\n",
      "Test:  [4700/4952]  eta: 0:00:15  model_time: 0.0337 (0.0346)  evaluator_time: 0.0060 (0.0111)  time: 0.0580  data: 0.0150  max mem: 20046\n",
      "Test:  [4800/4952]  eta: 0:00:09  model_time: 0.0345 (0.0346)  evaluator_time: 0.0085 (0.0110)  time: 0.0623  data: 0.0156  max mem: 20046\n",
      "Test:  [4900/4952]  eta: 0:00:03  model_time: 0.0335 (0.0346)  evaluator_time: 0.0055 (0.0112)  time: 0.0574  data: 0.0140  max mem: 20046\n",
      "Test:  [4951/4952]  eta: 0:00:00  model_time: 0.0336 (0.0346)  evaluator_time: 0.0064 (0.0111)  time: 0.0603  data: 0.0144  max mem: 20046\n",
      "Test: Total time: 0:05:09 (0.0624 s / it)\n",
      "Averaged stats: model_time: 0.0336 (0.0346)  evaluator_time: 0.0064 (0.0111)\n",
      "Accumulating evaluation results...\n",
      "DONE (t=8.50s).\n",
      "IoU metric: bbox\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.322\n",
      " Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.537\n",
      " Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.340\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.133\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.316\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.436\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.280\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.463\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.492\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.270\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.495\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.605\n"
     ]
    }
   ],
   "source": [
    "train(model, data_loader_train, data_loader_test, n_epochs=4, save_dir='models_v5_a100')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 311
    },
    "id": "bsPPUvhFHSlm",
    "outputId": "6cbef455-f506-4278-a25b-d56009788e7f"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<All keys matched successfully>"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.load_state_dict(torch.load('models_v5_a100/faster-rcnn-gator-epochs-0.pt'))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Unmodified FasterRCNN evaluation results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "09kVQ3UWKvql",
    "outputId": "fa4f43f9-cda9-4d2a-daca-a4a0f269537d"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "creating index...\n",
      "index created!\n",
      "Test:  [   0/4952]  eta: 0:12:03  model_time: 0.0512 (0.0512)  evaluator_time: 0.0148 (0.0148)  time: 0.1461  data: 0.0791  max mem: 4510\n",
      "Test:  [ 100/4952]  eta: 0:07:26  model_time: 0.0400 (0.0412)  evaluator_time: 0.0066 (0.0084)  time: 0.0968  data: 0.0474  max mem: 4510\n",
      "Test:  [ 200/4952]  eta: 0:07:20  model_time: 0.0415 (0.0412)  evaluator_time: 0.0051 (0.0085)  time: 0.0913  data: 0.0393  max mem: 4510\n",
      "Test:  [ 300/4952]  eta: 0:07:10  model_time: 0.0424 (0.0412)  evaluator_time: 0.0062 (0.0085)  time: 0.1003  data: 0.0471  max mem: 4510\n",
      "Test:  [ 400/4952]  eta: 0:07:03  model_time: 0.0416 (0.0413)  evaluator_time: 0.0055 (0.0084)  time: 0.0892  data: 0.0397  max mem: 4510\n",
      "Test:  [ 500/4952]  eta: 0:06:50  model_time: 0.0401 (0.0413)  evaluator_time: 0.0056 (0.0083)  time: 0.0911  data: 0.0417  max mem: 4510\n",
      "Test:  [ 600/4952]  eta: 0:06:45  model_time: 0.0412 (0.0413)  evaluator_time: 0.0047 (0.0085)  time: 0.0958  data: 0.0461  max mem: 4510\n",
      "Test:  [ 700/4952]  eta: 0:06:35  model_time: 0.0420 (0.0414)  evaluator_time: 0.0053 (0.0085)  time: 0.0876  data: 0.0390  max mem: 4510\n",
      "Test:  [ 800/4952]  eta: 0:06:26  model_time: 0.0402 (0.0414)  evaluator_time: 0.0076 (0.0086)  time: 0.0952  data: 0.0440  max mem: 4510\n",
      "Test:  [ 900/4952]  eta: 0:06:16  model_time: 0.0401 (0.0413)  evaluator_time: 0.0045 (0.0085)  time: 0.0856  data: 0.0374  max mem: 4510\n",
      "Test:  [1000/4952]  eta: 0:06:05  model_time: 0.0416 (0.0414)  evaluator_time: 0.0053 (0.0084)  time: 0.0841  data: 0.0353  max mem: 4510\n",
      "Test:  [1100/4952]  eta: 0:05:57  model_time: 0.0414 (0.0414)  evaluator_time: 0.0046 (0.0085)  time: 0.0978  data: 0.0492  max mem: 4510\n",
      "Test:  [1200/4952]  eta: 0:05:49  model_time: 0.0409 (0.0414)  evaluator_time: 0.0088 (0.0086)  time: 0.1056  data: 0.0510  max mem: 4510\n",
      "Test:  [1300/4952]  eta: 0:05:39  model_time: 0.0399 (0.0413)  evaluator_time: 0.0048 (0.0086)  time: 0.0914  data: 0.0416  max mem: 4510\n",
      "Test:  [1400/4952]  eta: 0:05:30  model_time: 0.0402 (0.0413)  evaluator_time: 0.0048 (0.0086)  time: 0.0884  data: 0.0381  max mem: 4510\n",
      "Test:  [1500/4952]  eta: 0:05:20  model_time: 0.0401 (0.0413)  evaluator_time: 0.0066 (0.0086)  time: 0.0936  data: 0.0438  max mem: 4510\n",
      "Test:  [1600/4952]  eta: 0:05:12  model_time: 0.0401 (0.0413)  evaluator_time: 0.0054 (0.0086)  time: 0.0969  data: 0.0442  max mem: 4510\n",
      "Test:  [1700/4952]  eta: 0:05:02  model_time: 0.0426 (0.0413)  evaluator_time: 0.0044 (0.0086)  time: 0.0907  data: 0.0389  max mem: 4510\n",
      "Test:  [1800/4952]  eta: 0:04:52  model_time: 0.0401 (0.0413)  evaluator_time: 0.0049 (0.0086)  time: 0.0888  data: 0.0382  max mem: 4510\n",
      "Test:  [1900/4952]  eta: 0:04:42  model_time: 0.0413 (0.0412)  evaluator_time: 0.0047 (0.0086)  time: 0.0878  data: 0.0391  max mem: 4510\n",
      "Test:  [2000/4952]  eta: 0:04:34  model_time: 0.0423 (0.0412)  evaluator_time: 0.0063 (0.0086)  time: 0.0971  data: 0.0435  max mem: 4510\n",
      "Test:  [2100/4952]  eta: 0:04:24  model_time: 0.0407 (0.0412)  evaluator_time: 0.0058 (0.0086)  time: 0.0966  data: 0.0454  max mem: 4510\n",
      "Test:  [2200/4952]  eta: 0:04:14  model_time: 0.0413 (0.0412)  evaluator_time: 0.0052 (0.0085)  time: 0.0958  data: 0.0425  max mem: 4510\n",
      "Test:  [2300/4952]  eta: 0:04:05  model_time: 0.0423 (0.0412)  evaluator_time: 0.0056 (0.0085)  time: 0.1112  data: 0.0582  max mem: 4510\n",
      "Test:  [2400/4952]  eta: 0:03:56  model_time: 0.0418 (0.0412)  evaluator_time: 0.0076 (0.0085)  time: 0.1060  data: 0.0527  max mem: 4510\n",
      "Test:  [2500/4952]  eta: 0:03:47  model_time: 0.0417 (0.0412)  evaluator_time: 0.0055 (0.0085)  time: 0.0969  data: 0.0459  max mem: 4510\n",
      "Test:  [2600/4952]  eta: 0:03:38  model_time: 0.0401 (0.0412)  evaluator_time: 0.0059 (0.0085)  time: 0.1111  data: 0.0605  max mem: 4510\n",
      "Test:  [2700/4952]  eta: 0:03:28  model_time: 0.0407 (0.0412)  evaluator_time: 0.0058 (0.0085)  time: 0.0944  data: 0.0434  max mem: 4510\n",
      "Test:  [2800/4952]  eta: 0:03:19  model_time: 0.0402 (0.0412)  evaluator_time: 0.0053 (0.0085)  time: 0.0943  data: 0.0434  max mem: 4510\n",
      "Test:  [2900/4952]  eta: 0:03:10  model_time: 0.0402 (0.0412)  evaluator_time: 0.0053 (0.0085)  time: 0.0946  data: 0.0425  max mem: 4510\n",
      "Test:  [3000/4952]  eta: 0:03:00  model_time: 0.0413 (0.0412)  evaluator_time: 0.0055 (0.0085)  time: 0.0927  data: 0.0416  max mem: 4510\n",
      "Test:  [3100/4952]  eta: 0:02:51  model_time: 0.0423 (0.0412)  evaluator_time: 0.0055 (0.0084)  time: 0.0932  data: 0.0423  max mem: 4510\n",
      "Test:  [3200/4952]  eta: 0:02:42  model_time: 0.0417 (0.0413)  evaluator_time: 0.0062 (0.0085)  time: 0.0989  data: 0.0458  max mem: 4510\n",
      "Test:  [3300/4952]  eta: 0:02:33  model_time: 0.0401 (0.0413)  evaluator_time: 0.0054 (0.0085)  time: 0.0901  data: 0.0400  max mem: 4510\n",
      "Test:  [3400/4952]  eta: 0:02:24  model_time: 0.0423 (0.0413)  evaluator_time: 0.0045 (0.0085)  time: 0.0900  data: 0.0377  max mem: 4510\n",
      "Test:  [3500/4952]  eta: 0:02:14  model_time: 0.0419 (0.0413)  evaluator_time: 0.0057 (0.0085)  time: 0.0856  data: 0.0374  max mem: 4510\n",
      "Test:  [3600/4952]  eta: 0:02:05  model_time: 0.0424 (0.0413)  evaluator_time: 0.0060 (0.0085)  time: 0.0882  data: 0.0384  max mem: 4510\n",
      "Test:  [3700/4952]  eta: 0:01:56  model_time: 0.0422 (0.0413)  evaluator_time: 0.0047 (0.0085)  time: 0.0859  data: 0.0360  max mem: 4510\n",
      "Test:  [3800/4952]  eta: 0:01:46  model_time: 0.0422 (0.0413)  evaluator_time: 0.0050 (0.0085)  time: 0.0833  data: 0.0346  max mem: 4510\n",
      "Test:  [3900/4952]  eta: 0:01:37  model_time: 0.0401 (0.0413)  evaluator_time: 0.0051 (0.0085)  time: 0.0892  data: 0.0403  max mem: 4510\n",
      "Test:  [4000/4952]  eta: 0:01:28  model_time: 0.0401 (0.0413)  evaluator_time: 0.0068 (0.0085)  time: 0.0951  data: 0.0445  max mem: 4510\n",
      "Test:  [4100/4952]  eta: 0:01:19  model_time: 0.0422 (0.0413)  evaluator_time: 0.0046 (0.0085)  time: 0.0880  data: 0.0382  max mem: 4510\n",
      "Test:  [4200/4952]  eta: 0:01:09  model_time: 0.0402 (0.0413)  evaluator_time: 0.0046 (0.0085)  time: 0.0877  data: 0.0390  max mem: 4510\n",
      "Test:  [4300/4952]  eta: 0:01:00  model_time: 0.0409 (0.0413)  evaluator_time: 0.0049 (0.0085)  time: 0.0865  data: 0.0376  max mem: 4510\n",
      "Test:  [4400/4952]  eta: 0:00:51  model_time: 0.0414 (0.0412)  evaluator_time: 0.0056 (0.0085)  time: 0.0920  data: 0.0434  max mem: 4510\n",
      "Test:  [4500/4952]  eta: 0:00:41  model_time: 0.0422 (0.0412)  evaluator_time: 0.0060 (0.0085)  time: 0.0909  data: 0.0402  max mem: 4510\n",
      "Test:  [4600/4952]  eta: 0:00:32  model_time: 0.0417 (0.0412)  evaluator_time: 0.0051 (0.0084)  time: 0.0869  data: 0.0385  max mem: 4510\n",
      "Test:  [4700/4952]  eta: 0:00:23  model_time: 0.0414 (0.0412)  evaluator_time: 0.0047 (0.0084)  time: 0.0836  data: 0.0353  max mem: 4510\n",
      "Test:  [4800/4952]  eta: 0:00:14  model_time: 0.0419 (0.0412)  evaluator_time: 0.0058 (0.0084)  time: 0.0909  data: 0.0405  max mem: 4510\n",
      "Test:  [4900/4952]  eta: 0:00:04  model_time: 0.0403 (0.0412)  evaluator_time: 0.0044 (0.0084)  time: 0.0820  data: 0.0332  max mem: 4510\n",
      "Test:  [4951/4952]  eta: 0:00:00  model_time: 0.0422 (0.0412)  evaluator_time: 0.0052 (0.0084)  time: 0.0906  data: 0.0386  max mem: 4510\n",
      "Test: Total time: 0:07:38 (0.0927 s / it)\n",
      "Averaged stats: model_time: 0.0422 (0.0412)  evaluator_time: 0.0052 (0.0084)\n",
      "Accumulating evaluation results...\n",
      "DONE (t=5.45s).\n",
      "IoU metric: bbox\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.370\n",
      " Average Precision  (AP) @[ IoU=0.50      | area=   all | maxDets=100 ] = 0.586\n",
      " Average Precision  (AP) @[ IoU=0.75      | area=   all | maxDets=100 ] = 0.396\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.163\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.357\n",
      " Average Precision  (AP) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.491\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=  1 ] = 0.307\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets= 10 ] = 0.485\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=   all | maxDets=100 ] = 0.509\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area= small | maxDets=100 ] = 0.289\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area=medium | maxDets=100 ] = 0.502\n",
      " Average Recall     (AR) @[ IoU=0.50:0.95 | area= large | maxDets=100 ] = 0.623\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<coco_eval.CocoEvaluator at 0x7fd90e47db90>"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "faster_rcnn = torchvision.models.detection.fasterrcnn_resnet50_fpn(pretrained=True)\n",
    "faster_rcnn.to(device)\n",
    "evaluate(faster_rcnn, data_loader_test, device=device)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "jMNmFOpuW5g3"
   },
   "source": [
    "### Evaluation using standard COCO pipeline\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Converting torch predictions to Fiftyone compatible objects"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "id": "PgGTVy4wW4J_"
   },
   "outputs": [],
   "source": [
    "def convert_torch_predictions(preds, det_id, s_id, w, h, classes):\n",
    "  # Convert the outputs of the torch model into a FiftyOne Detections object\n",
    "  dets = []\n",
    "  for bbox, label, score in zip(preds[\"boxes\"].cpu().detach().numpy(), preds[\"labels\"].cpu().detach().numpy(), preds[\"scores\"].cpu().detach().numpy()):\n",
    "    # Parse prediction into FiftyOne Detection object\n",
    "    x0,y0,x1,y1 = bbox\n",
    "    coco_obj = fouc.COCOObject(det_id, s_id, int(label), [x0, y0, x1-x0, y1-y0])\n",
    "    det = coco_obj.to_detection((w,h), classes)\n",
    "    det[\"confidence\"] = float(score)\n",
    "    dets.append(det)\n",
    "    det_id += 1\n",
    "        \n",
    "  detections = fo.Detections(detections=dets)\n",
    "      \n",
    "  return detections, det_id"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Adding detections to Fiftyone dataset so they can be inspected in the Fiftyone client"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "id": "Ly9V_8oyW4VF"
   },
   "outputs": [],
   "source": [
    "def add_detections(model, torch_dataset, view, field_name=\"predictions\"):\n",
    "  # Run inference on a dataset and add results to FiftyOne\n",
    "  device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')\n",
    "\n",
    "  model.eval()\n",
    "  model.to(device)\n",
    "  image_paths = torch_dataset.img_paths\n",
    "  classes = torch_dataset.classes\n",
    "  det_id = 0\n",
    "  \n",
    "  print(\"Adding detections to FiftyOne view...\")\n",
    "  with fo.ProgressBar() as pb:\n",
    "    for img, targets in pb(torch_dataset):\n",
    "      # Get FiftyOne sample indexed by unique image filepath\n",
    "      img_id = int(targets[\"image_id\"][0])\n",
    "      img_path = image_paths[img_id]\n",
    "      sample = view[img_path]\n",
    "      s_id = sample.id\n",
    "      w = sample.metadata[\"width\"]\n",
    "      h = sample.metadata[\"height\"]\n",
    "      \n",
    "      # Inference\n",
    "      preds = model(img.unsqueeze(0).to(device))[0]\n",
    "      \n",
    "      detections, det_id = convert_torch_predictions(preds, det_id, s_id, w, h, classes)\n",
    "      \n",
    "      sample[field_name] = detections\n",
    "      sample.save()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Running evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "p98hYoV0W43y",
    "outputId": "89aeb503-f100-41e3-e380-77ca6b5f9b65"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Adding detections to FiftyOne view...\n",
      " 100% |███████████████| 4952/4952 [13.3m elapsed, 0s remaining, 7.8 samples/s]      \n",
      "Evaluating detections...\n",
      " 100% |███████████████| 4952/4952 [4.9m elapsed, 0s remaining, 19.9 samples/s]      \n",
      "Performing IoU sweep...\n",
      " 100% |███████████████| 4952/4952 [7.6m elapsed, 0s remaining, 12.8 samples/s]      \n",
      "0.32474702284023343\n"
     ]
    }
   ],
   "source": [
    "add_detections(model, torch_dataset_test, test_view, field_name=\"v2_predictions_stochastic_edge_sampling\")\n",
    "\n",
    "results = fo.evaluate_detections(test_view, \"v2_predictions_stochastic_edge_sampling\", eval_key=\"eval_v2_stochastic_edge_sampling\", compute_mAP=True) \n",
    "print(results.mAP())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "2JGkH6w6W47I"
   },
   "outputs": [],
   "source": [
    "results.plot_confusion_matrix()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "DNq6Pip6XNa1"
   },
   "outputs": [],
   "source": [
    "results.plot_pr_curves()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "Tq2U_3DTW4-j"
   },
   "outputs": [],
   "source": [
    "session.view = test_view"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "XjqZ0AZrXS3_"
   },
   "source": [
    "## Different training approach"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "azbHCFYaWmA6"
   },
   "source": [
    "### Processing COCO dataset to Torch geometric data dataset to speed up training by saving graphs created from precomputed embedding for each image in the dataset (~1 TB)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "zaQRCNuUWnAW"
   },
   "outputs": [],
   "source": [
    "class TorchGemetricCocoRelationalGraphDataaset(pyg_data.Dataset):\n",
    "  def __init__(self, root, model, coco_loader, phase, transform=None, pre_transform=None, pre_filter=None):\n",
    "    if phase not in ['train', 'test']:\n",
    "      raise Exception(f\"Unexpected value for phase argument: '{phase}'. Expected 'train' or 'test'.\")\n",
    "    self._phase = phase\n",
    "    self._processed_file_names = []\n",
    "    self._model = model\n",
    "    self._coco_loader = coco_loader\n",
    "    super().__init__(root, transform, pre_transform, pre_filter)\n",
    "\n",
    "  @property\n",
    "  def processed_file_names(self):\n",
    "      return self._processed_file_names\n",
    "\n",
    "  def process(self):\n",
    "    batch_idx = 0\n",
    "\n",
    "    def build_saver(targets):\n",
    "      def saver(data, in_batch_idx):\n",
    "        data_idx = batch_idx + in_batch_idx\n",
    "        data.y = targets[in_batch_idx]\n",
    "        file_path = osp.join(self.processed_dir, f'data_{data_idx}.pt')\n",
    "        torch.save(data, file_path)\n",
    "        self._processed_file_names.append(file_path)\n",
    "      return saver\n",
    "\n",
    "    if self._phase == 'train':\n",
    "      self._model.train()\n",
    "    else:\n",
    "      self._model.eval()\n",
    "\n",
    "    self._model.to(device)\n",
    "\n",
    "    with fo.ProgressBar() as pb:\n",
    "      for images, targets in pb(self._coco_loader):\n",
    "        self._register_callback(build_saver(targets))\n",
    "        batch_size = len(images)\n",
    "        images = list(image.to(device) for image in images)\n",
    "        targets = [{k: v.to(device) for k, v in t.items()} for t in targets]\n",
    "        self._model(images, targets)\n",
    "        batch_idx += batch_size\n",
    "\n",
    "    self._register_callback()\n",
    "\n",
    "  def len(self):\n",
    "    return len(self._processed_file_names)\n",
    "\n",
    "  def get(self, idx):\n",
    "    data = torch.load(osp.join(self.processed_dir, f'data_{idx}.pt'))\n",
    "    return data\n",
    "\n",
    "  def _register_callback(self, callback=None):\n",
    "    self._model.roi_heads.box_head.gator_graph.on_data_created = callback"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "2BcmhpliWpsj"
   },
   "outputs": [],
   "source": [
    "graph_dataset_train = TorchGemetricCocoRelationalGraphDataaset('/root/graph-data/train', model, data_loader_train, 'train')\n",
    "graph_dataset_test = TorchGemetricCocoRelationalGraphDataaset('/root/graph-data/test', model, data_loader_test, 'test')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "kKXbpvODQ9W7"
   },
   "source": [
    "### Defining GATorHidra extension model for object detection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "lcsNCkO5e7lX"
   },
   "outputs": [],
   "source": [
    "class GATorPredictor(nn.Module):\n",
    "  def __init__(self, in_channels, num_classes):\n",
    "    super(GATorPredictor, self).__init__()\n",
    "    self.cls_score = nn.Linear(in_channels, num_classes)\n",
    "    self.bbox_pred = nn.Linear(in_channels, num_classes * 4)\n",
    "\n",
    "  def forward(self, x):\n",
    "    if x.dim() == 4:\n",
    "      assert list(x.shape[2:]) == [1, 1]\n",
    "    x = x.flatten(start_dim=1)\n",
    "    scores = self.cls_score(x)\n",
    "    bbox_deltas = self.bbox_pred(x)\n",
    "    return scores, bbox_deltas"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Self explanatory"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "QylQqKhENoUR"
   },
   "outputs": [],
   "source": [
    "class GATorHidra(torchvision.models.detection.roi_heads.RoIHeads):\n",
    "  def __init__(self,\n",
    "      # GATor\n",
    "      gator_input_dim=1024,\n",
    "      gator_hidden_dim=1024,\n",
    "      gator_output_dim=1024,\n",
    "      gator_num_layers=2,\n",
    "      gator_num_heads=8,\n",
    "      gator_dropout=0.5,\n",
    "      gator_edge_dim=None,\n",
    "      gator_num_classes=91,\n",
    "      # Faster R-CNN training\n",
    "      fg_iou_thresh=0.5,\n",
    "      bg_iou_thresh=0.5,\n",
    "      batch_size_per_image=512,\n",
    "      positive_fraction=0.25,\n",
    "      bbox_reg_weights=None,\n",
    "      # Faster R-CNN inference\n",
    "      score_thresh=0.05,\n",
    "      nms_thresh=0.5,\n",
    "      detections_per_img=100,\n",
    "    ):\n",
    "    super(GATorHidra, self).__init__(\n",
    "        box_roi_pool=None,\n",
    "        box_head=None,\n",
    "        box_predictor=None,\n",
    "        # Faster R-CNN training\n",
    "        fg_iou_thresh=fg_iou_thresh,\n",
    "        bg_iou_thresh=bg_iou_thresh,\n",
    "        batch_size_per_image=batch_size_per_image,\n",
    "        positive_fraction=positive_fraction,\n",
    "        bbox_reg_weights=bbox_reg_weights,\n",
    "        # Faster R-CNN inference\n",
    "        score_thresh=score_thresh,\n",
    "        nms_thresh=nms_thresh,\n",
    "        detections_per_img=detections_per_img,\n",
    "    )\n",
    "\n",
    "    self.gator = GATor(input_dim=gator_input_dim,\n",
    "                       hidden_dim=gator_hidden_dim,\n",
    "                       output_dim=gator_output_dim,\n",
    "                       layers=gator_num_layers,\n",
    "                       heads=gator_num_heads, \n",
    "                       dropout=gator_dropout, \n",
    "                       edge_dim=gator_edge_dim)\n",
    "    self.predictor = GATorPredictor(in_channels=gator_output_dim, \n",
    "                                    num_classes=gator_num_classes)\n",
    "\n",
    "  def forward(self, data):\n",
    "    proposals, targets, image_shapes = data.boxes, data.y, data.image_shape\n",
    "    box_features = self.gator(data)\n",
    "    class_logits, box_regression = self.predictor(box_features)\n",
    "\n",
    "    if self.training:\n",
    "      proposals, matched_idxs, labels, regression_targets = self.select_training_samples(proposals, targets)\n",
    "      loss_classifier, loss_box_reg = fastrcnn_loss(class_logits, box_regression, labels, regression_targets)\n",
    "      return {\"loss_classifier\": loss_classifier, \"loss_box_reg\": loss_box_reg}\n",
    "    else:\n",
    "      boxes, scores, labels = self.postprocess_detections(class_logits, box_regression, proposals, image_shapes)\n",
    "      num_images = len(boxes)\n",
    "      result = []\n",
    "      for i in range(num_images):\n",
    "        result.append(\n",
    "            {\n",
    "              \"boxes\": boxes[i],\n",
    "              \"labels\": labels[i],\n",
    "              \"scores\": scores[i],\n",
    "            }\n",
    "        )\n",
    "      return result"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "fNmDNFDmdnOk"
   },
   "outputs": [],
   "source": [
    "gator_hidra = GATorHidra()\n",
    "gator_hidra.to(device)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "GFaYy8dIVsOG"
   },
   "source": [
    "### Novel training approach using our torch geometric dataset (only for chads)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "iYbxyztWVeX5"
   },
   "outputs": [],
   "source": [
    "def graph_collate_function(data_list):\n",
    "  batch = pyg_data.Batch.from_data_list(data_list)\n",
    "  y = [{k: v for k, v in data.y.items()} for data in data_list]\n",
    "  proposals = [data.boxes for data in data_list]\n",
    "  batch.y = y\n",
    "  batch.boxes = proposals\n",
    "  return batch"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "sxudu3NKVe8C"
   },
   "outputs": [],
   "source": [
    "class BatchingDataLoader(torch.utils.data.DataLoader):\n",
    "  def __init__(self, dataset, batch_size=1, shuffle=False, **kwargs):\n",
    "    super(BatchingDataLoader, self).__init__(dataset, batch_size, shuffle, collate_fn=graph_collate_function, **kwargs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "CzBtW98XVhtD"
   },
   "outputs": [],
   "source": [
    "pyg_data_loader_train = BatchingDataLoader(graph_dataset_train, batch_size=4, shuffle=False)\n",
    "pyg_data_loader_test = BatchingDataLoader(graph_dataset_test, batch_size=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "ubA0UcW4V8mX"
   },
   "outputs": [],
   "source": [
    "# https://github.com/pytorch/vision/blob/afda28accbc79035384952c0359f0e4de8454cb3/torchvision/models/detection/roi_heads.py\n",
    "\n",
    "def fastrcnn_loss(class_logits, box_regression, labels, regression_targets):\n",
    "  # type: (Tensor, Tensor, List[Tensor], List[Tensor]) -> Tuple[Tensor, Tensor]\n",
    "  \"\"\"\n",
    "  Computes the loss for Faster R-CNN.\n",
    "\n",
    "  Args:\n",
    "      class_logits (Tensor)\n",
    "      box_regression (Tensor)\n",
    "      labels (list[BoxList])\n",
    "      regression_targets (Tensor)\n",
    "\n",
    "  Returns:\n",
    "      classification_loss (Tensor)\n",
    "      box_loss (Tensor)\n",
    "  \"\"\"\n",
    "\n",
    "  labels = torch.cat(labels, dim=0)\n",
    "  regression_targets = torch.cat(regression_targets, dim=0)\n",
    "\n",
    "  classification_loss = F.cross_entropy(class_logits, labels)\n",
    "\n",
    "  # get indices that correspond to the regression targets for\n",
    "  # the corresponding ground truth labels, to be used with\n",
    "  # advanced indexing\n",
    "  sampled_pos_inds_subset = torch.where(labels > 0)[0]\n",
    "  labels_pos = labels[sampled_pos_inds_subset]\n",
    "  N, num_classes = class_logits.shape\n",
    "  box_regression = box_regression.reshape(N, box_regression.size(-1) // 4, 4)\n",
    "\n",
    "  box_loss = F.smooth_l1_loss(\n",
    "    box_regression[sampled_pos_inds_subset, labels_pos],\n",
    "    regression_targets[sampled_pos_inds_subset],\n",
    "    beta=1 / 9,\n",
    "    reduction=\"sum\",\n",
    "  )\n",
    "  box_loss = box_loss / labels.numel()\n",
    "\n",
    "  return classification_loss, box_loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "wFePOeiwUBi0"
   },
   "outputs": [],
   "source": [
    "def train_one_epoch_geometric(model, optimizer, train_loader, device, epoch, print_freq=10):\n",
    "  model.train()\n",
    "  metric_logger = utils.MetricLogger(delimiter=\"  \")\n",
    "  metric_logger.add_meter('lr', utils.SmoothedValue(window_size=1, fmt='{value:.6f}'))\n",
    "  header = 'Epoch: [{}]'.format(epoch)\n",
    "\n",
    "  lr_scheduler = None\n",
    "  if epoch == 0:\n",
    "    warmup_factor = 1. / 1000\n",
    "    warmup_iters = min(1000, len(train_loader) - 1)\n",
    "\n",
    "    lr_scheduler = utils.warmup_lr_scheduler(optimizer, warmup_iters, warmup_factor)\n",
    "\n",
    "  for data_batch in metric_logger.log_every(train_loader, print_freq, header):\n",
    "    # Put batch to device\n",
    "    data_batch.y = [{k: v.to(device) for k, v in y.items()} for y in data_batch.y]\n",
    "    data_batch.x = data_batch.x.to(device)\n",
    "    data_batch.boxes = [box.to(device) for box in data_batch.boxes]\n",
    "    loss_dict = model(data_batch)\n",
    "\n",
    "    losses = sum(loss for loss in loss_dict.values())\n",
    "\n",
    "    # reduce losses over all GPUs for logging purposes\n",
    "    loss_dict_reduced = utils.reduce_dict(loss_dict)\n",
    "    losses_reduced = sum(loss for loss in loss_dict_reduced.values())\n",
    "\n",
    "    loss_value = losses_reduced.item()\n",
    "\n",
    "    if not math.isfinite(loss_value):\n",
    "      print(\"Loss is {}, stopping training\".format(loss_value))\n",
    "      print(loss_dict_reduced)\n",
    "      sys.exit(1)\n",
    "\n",
    "    optimizer.zero_grad()\n",
    "    losses.backward()\n",
    "    optimizer.step()\n",
    "\n",
    "    if lr_scheduler is not None:\n",
    "      lr_scheduler.step()\n",
    "\n",
    "    metric_logger.update(loss=losses_reduced, **loss_dict_reduced)\n",
    "    metric_logger.update(lr=optimizer.param_groups[0][\"lr\"])\n",
    "\n",
    "    return metric_logger"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "EPOppUdAREM_"
   },
   "outputs": [],
   "source": [
    "def train_geometric(model, train_loader, test_loader, n_epochs):\n",
    "  device = torch.device('cuda') if torch.cuda.is_available() else torch.device('cpu')\n",
    "  model.train()\n",
    "  model.to(device)\n",
    "  params = [p for p in model.parameters() if p.requires_grad]\n",
    "  optimizer = torch.optim.SGD(params, lr=0.005, momentum=0.9, weight_decay=0.0005)\n",
    "\n",
    "  lr_scheduler = torch.optim.lr_scheduler.StepLR(optimizer,\n",
    "                                                step_size=3,\n",
    "                                                gamma=0.1)\n",
    "\n",
    "  for epoch in range(n_epochs):\n",
    "    # train for one epoch, printing every 10 iterations\n",
    "    train_one_epoch_geometric(model, optimizer, train_loader, device, epoch, print_freq=10)\n",
    "    # update the learning rate\n",
    "    lr_scheduler.step()\n",
    "    # evaluate on the test dataset\n",
    "    # evaluate(model, test_loader, device=device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "7iEcGmcQReVl"
   },
   "outputs": [],
   "source": [
    "train_geometric(gator_hidra, pyg_data_loader_train, pyg_data_loader_test, n_epochs=10)"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "include_colab_link": true,
   "name": "gat-or.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.12"
  },
  "widgets": {
   "application/vnd.jupyter.widget-state+json": {
    "10dc5c906a68455e8ab9966d2ecc1bbe": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HBoxModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HBoxModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HBoxView",
      "box_style": "",
      "children": [
       "IPY_MODEL_314b12ed64c044208a8ff2a6e4fdc59c",
       "IPY_MODEL_d5ae0d02c3b844bdafdec479bdb159e7",
       "IPY_MODEL_452cc2936c62488aae19aab38eca67f5"
      ],
      "layout": "IPY_MODEL_38d76333d35b4823b6de75f6e30d159a"
     }
    },
    "314b12ed64c044208a8ff2a6e4fdc59c": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_cf47b4248ecd459280bbe9f9c313c035",
      "placeholder": "​",
      "style": "IPY_MODEL_ab2c3648905e4c769249c935ceaf5ec5",
      "value": "100%"
     }
    },
    "38d76333d35b4823b6de75f6e30d159a": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "452cc2936c62488aae19aab38eca67f5": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_e7cdbe0e4c024ca8a50aa6b5a8606280",
      "placeholder": "​",
      "style": "IPY_MODEL_69fc04dc526f4d7188b2e0ea62402d8c",
      "value": " 160M/160M [00:04&lt;00:00, 28.9MB/s]"
     }
    },
    "69fc04dc526f4d7188b2e0ea62402d8c": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "8ccf33413b0a41f5b6dd6765f7689fbc": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "ab2c3648905e4c769249c935ceaf5ec5": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "cf47b4248ecd459280bbe9f9c313c035": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "d5ae0d02c3b844bdafdec479bdb159e7": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "FloatProgressModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "FloatProgressModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "ProgressView",
      "bar_style": "success",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_8ccf33413b0a41f5b6dd6765f7689fbc",
      "max": 167502836,
      "min": 0,
      "orientation": "horizontal",
      "style": "IPY_MODEL_e013dd58a6274902a2df85eb8c00e31d",
      "value": 167502836
     }
    },
    "e013dd58a6274902a2df85eb8c00e31d": {
     "model_module": "@jupyter-widgets/controls",
     "model_module_version": "1.5.0",
     "model_name": "ProgressStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "ProgressStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "bar_color": null,
      "description_width": ""
     }
    },
    "e7cdbe0e4c024ca8a50aa6b5a8606280": {
     "model_module": "@jupyter-widgets/base",
     "model_module_version": "1.2.0",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    }
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}