self_supervised_train.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "EXP_NAME = 'LUCAS_Self560_ViT_Trans'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from train_SimCLR_utils import *\n",
    "import torch\n",
    "from torch.utils.data import DataLoader\n",
    "from dataset.dataset_loader import SNDataset,SNDatasetClimate, myNormalize, myToTensor, Augmentations\n",
    "from torchvision import transforms\n",
    "import random\n",
    "import numpy as np\n",
    "from dataset.utils.utils import TextColors as tc\n",
    "from plot_utils.plot import plot_train_test_losses\n",
    "from datetime import date, datetime\n",
    "import torch.nn.functional as F\n",
    "import cv2\n",
    "import json"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "DATASET = 'LUCAS' # 'LUCAS', 'RaCA'\n",
    "\n",
    "if DATASET == 'LUCAS':\n",
    "    from dataset.dataset_loader import SNDataset,SNDatasetClimate, myNormalize, myToTensor, Augmentations\n",
    "    OC_MAX = 560.2\n",
    "if DATASET == 'RaCA':\n",
    "    from dataset.dataset_loader import SNDataset,SNDatasetClimate, myNormalize, myToTensor, Augmentations\n",
    "    OC_MAX = 4115"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create a folder called 'results' in the current directory if it doesn't exist\n",
    "if not os.path.exists('results_simclr_lucas'):\n",
    "    os.mkdir('results_simclr_lucas')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Current Date and Time: 2024-03-04 08:48:39\n"
     ]
    }
   ],
   "source": [
    "# Format the date and time\n",
    "now = datetime.now()\n",
    "start_string = now.strftime(\"%Y-%m-%d %H:%M:%S\")\n",
    "print(\"Current Date and Time:\", start_string)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Nafiseh'"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import os\n",
    "os.getcwd()\n",
    "\n",
    "NAFISEH = \"Nafiseh\"\n",
    "MOIEN = \"Moien\"\n",
    "\n",
    "if \"d:\" in os.getcwd():\n",
    "    USER = MOIEN\n",
    "elif \"c:\" in os.getcwd():\n",
    "    USER = NAFISEH\n",
    "else:\n",
    "    raise Exception(\"Unknown user\")\n",
    "\n",
    "USER"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'cuda'"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Setup device-agnostic code\n",
    "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
    "device"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# CONFIG\n",
    "NUM_WORKERS = 6 if USER == NAFISEH else 2\n",
    "TRAIN_BATCH_SIZE = 32 if USER == NAFISEH else 4\n",
    "TEST_BATCH_SIZE = 32 if USER == NAFISEH else 4\n",
    "LEARNING_RATE = 1e-4 if USER == NAFISEH else 1e-4\n",
    "NUM_EPOCHS = 40 if USER == NAFISEH else 2\n",
    "LR_SCHEDULER = \"step\" # step, plateau or None\n",
    "\n",
    "USE_SRTM = True\n",
    "\n",
    "USE_SPATIAL_ATTENTION = False if USER == NAFISEH else False\n",
    "CNN_ARCHITECTURE = 'ViT' # vgg16 or resnet101 or \"ViT\"\n",
    "RNN_ARCHITECTURE = 'Transformer' # LSTM, GRU, RNN, Transformer\n",
    "REG_VERSION = 1 if USER == NAFISEH else 2\n",
    "USE_LSTM_BRANCH = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "if USE_SRTM:\n",
    "    mynorm = myNormalize(img_bands_min_max =[[(0,7),(0,1)], [(7,12),(-1,1)], [(12), (-4,2963)], [(13), (0, 90)]], oc_min = 0, oc_max = OC_MAX)\n",
    "else:\n",
    "    mynorm = myNormalize(img_bands_min_max =[[(0,7),(0,1)], [(7,12),(-1,1)]], oc_min = 0, oc_max = OC_MAX)\n",
    "    \n",
    "my_to_tensor = myToTensor()\n",
    "my_augmentation = Augmentations()\n",
    "train_transform = transforms.Compose([mynorm, my_to_tensor,my_augmentation])\n",
    "test_transform = transforms.Compose([mynorm, my_to_tensor])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "if USER == MOIEN:\n",
    "    train_l8_folder_path = 'D:\\python\\SoilNet\\dataset\\l8_images\\\\train\\\\'\n",
    "    test_l8_folder_path = 'D:\\python\\SoilNet\\dataset\\l8_images\\\\test\\\\'\n",
    "    val_l8_folder_path = 'D:\\python\\SoilNet\\dataset\\l8_images\\\\val\\\\'\n",
    "    testval_l8_folder_path = 'D:\\python\\SoilNet\\dataset\\l8_images\\\\val\\\\'\n",
    "    lucas_csv_path = 'D:\\python\\SoilNet\\dataset\\LUCAS_2015_all.csv'\n",
    "    climate_csv_folder_path = \"D:\\\\python\\\\SoilNet\\\\dataset\\\\Climate\\\\All\\\\filled\\\\\"\n",
    "\n",
    "\n",
    "elif USER == NAFISEH:\n",
    "\n",
    "    if DATASET == 'LUCAS':\n",
    "     train_l8_folder_path = 'C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-7\\\\SoilNet-PreRelease\\\\dataset\\\\l8_images\\\\train'\n",
    "     test_l8_folder_path = 'C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-7\\\\SoilNet-PreRelease\\\\dataset\\\\l8_images\\\\test'\n",
    "     val_l8_folder_path = 'C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-7\\\\SoilNet-PreRelease\\\\dataset\\\\l8_images\\\\val'\n",
    "     testval_l8_folder_path = 'C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-7\\\\SoilNet-PreRelease\\\\dataset\\\\l8_images\\\\val'\n",
    "     # lucas_csv_path = 'C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-3\\\\SoilNet\\\\dataset\\\\LUCAS_2015_all.csv'\n",
    "    #  lucas_csv_path = \"C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-7\\\\SoilNet-PreRelease\\\\dataset\\\\simclr\\\\simclr.csv\" # The LUCAS will be replaced by SimCLR file\n",
    "     lucas_csv_path = 'C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-7\\\\SoilNet-PreRelease\\\\dataset\\\\simclr\\\\LUCAS_SimCLR_1.csv'\n",
    "     # climate_csv_folder_path = 'C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-7\\\\SoilNet-PreRelease\\\\dataset\\\\Climate\\\\All\\\\filled'\n",
    "     climate_csv_folder_path = 'C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-7\\\\SoilNet-PreRelease\\\\dataset\\\\Climate\\\\LUCAS_SimCLR_Cleaned'\n",
    "\n",
    "\n",
    "    # if DATASET == 'RaCA':\n",
    "    #  train_l8_folder_path = 'C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-7\\\\SoilNet-PreRelease\\\\dataset\\\\l8_images_simclr_RaCA\\\\train'\n",
    "    #  test_l8_folder_path = 'C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-7\\\\SoilNet-PreRelease\\\\dataset\\\\l8_images_simclr_RaCA\\\\test'\n",
    "    #  val_l8_folder_path = 'C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-7\\\\SoilNet-PreRelease\\\\dataset\\\\l8_images_simclr_RaCA\\\\val'\n",
    "    #  testval_l8_folder_path = 'C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-7\\\\SoilNet-PreRelease\\\\dataset\\\\l8_images_simclr_RaCA\\\\test_val'\n",
    "    #  # lucas_csv_path = \"C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-7\\\\SoilNet-PreRelease\\\\dataset\\\\RapidCarbon_OC.csv\"\n",
    "    #  lucas_csv_path = \"C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-7\\\\SoilNet-PreRelease\\\\dataset\\\\simclr\\\\simclr_RaCA.csv\"\n",
    "    #  climate_csv_folder_path = 'C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-7\\\\SoilNet-PreRelease\\\\dataset\\\\Climate\\\\RaCA_SimCLR\\\\filled'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "bands = [0,1,2,3,4,5,6,7,8,9,10,11] if not USE_SRTM else [0,1,2,3,4,5,6,7,8,9,10,11,12,13]\n",
    "\n",
    "\n",
    "################################# IF Not USE_LSTM_BRANCH ###############################\n",
    "if not USE_LSTM_BRANCH: # NOT USING THE CLIMATE DATA\n",
    "\n",
    "    train_ds = SNDataset(train_l8_folder_path, lucas_csv_path,l8_bands=bands, transform=train_transform)\n",
    "    test_ds =  SNDataset(test_l8_folder_path, lucas_csv_path,l8_bands=bands, transform=test_transform)\n",
    "    val_ds = SNDataset(val_l8_folder_path, lucas_csv_path,l8_bands=bands, transform=test_transform)\n",
    "    test_ds_w_id =  SNDataset(test_l8_folder_path, lucas_csv_path,l8_bands=bands, transform=test_transform, return_point_id=True)\n",
    "    testval_ds_w_id =  SNDataset(testval_l8_folder_path, lucas_csv_path,l8_bands=bands, transform=test_transform, return_point_id=True)\n",
    "    \n",
    "################################### IF USE_LSTM_BRANCH #################################\n",
    "else: # USING THE CLIMATE DATA\n",
    "    train_ds = SNDatasetClimate(train_l8_folder_path,\n",
    "                                    lucas_csv_path,\n",
    "                                    climate_csv_folder_path,\n",
    "                                    l8_bands=bands, transform=train_transform)\n",
    "\n",
    "    test_ds = SNDatasetClimate(test_l8_folder_path,\n",
    "                                lucas_csv_path,\n",
    "                                climate_csv_folder_path,\n",
    "                                l8_bands=bands, transform=test_transform)\n",
    "    \n",
    "    val_ds = SNDatasetClimate(val_l8_folder_path,\n",
    "                                lucas_csv_path,\n",
    "                                climate_csv_folder_path,\n",
    "                                l8_bands=bands, transform=test_transform)\n",
    "    \n",
    "    test_ds_w_id = SNDatasetClimate(test_l8_folder_path,\n",
    "                                lucas_csv_path,\n",
    "                                climate_csv_folder_path,\n",
    "                                l8_bands=bands, transform=test_transform, return_point_id=True)\n",
    "\n",
    "    testval_ds_w_id = SNDatasetClimate(testval_l8_folder_path,\n",
    "                                lucas_csv_path,\n",
    "                                climate_csv_folder_path,\n",
    "                                l8_bands=bands, transform=test_transform, return_point_id=True)\n",
    "    \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "61"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "SEQ_LEN = test_ds_w_id[0][0][1].shape[0]\n",
    "SEQ_LEN # max sequence length for Transformer model set up"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "11"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# COUNTING the csv files in the csv folder\n",
    "CSV_FILES = [f for f in os.listdir(climate_csv_folder_path) if f.endswith('.csv')]\n",
    "NUM_CLIMATE_FEATURES = len(CSV_FILES)\n",
    "NUM_CLIMATE_FEATURES"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__name__: submodules.src.transformer.transformer\n"
     ]
    }
   ],
   "source": [
    "from soilnet.soil_net import SoilNet, SoilNetLSTM, SoilNetSimCLR"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# train_dl = DataLoader(train_ds, batch_size=TRAIN_BATCH_SIZE, shuffle=True, num_workers=NUM_WORKERS)\n",
    "# # Get image and label from custom DataLoader\n",
    "# img_custom, climate,label_custom = next(iter(train_dl))\n",
    "# print(f\"Image shape: {img_custom.shape} -> [batch_size, color_channels, height, width]\")\n",
    "# print(f\"Climate shape: {climate.shape} -> [batch_size, climate_channels]\")\n",
    "# print(f\"Label shape: {label_custom.shape}\")\n",
    "# img_gpu = img_custom.to(device)\n",
    "# print(img_gpu.device, img_gpu.shape)\n",
    "# print(img_gpu.shape)\n",
    "# y = model(img_gpu)\n",
    "# y.detach()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "cv_results = {\"train_loss\": [],\n",
    "                \"train_acc_top1\": [],\n",
    "                \"train_acc_top5\": [],\n",
    "                \"train_acc_mean_pos\": [],\n",
    "                \"val_loss\": [],\n",
    "                \"val_acc_top1\": [],\n",
    "                \"val_acc_top5\": [],\n",
    "                \"val_acc_mean_pos\": [],\n",
    "    }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Current Date and Time: D_2024_03_04_T_08_49\n"
     ]
    }
   ],
   "source": [
    "from datetime import date, datetime\n",
    "# Format the date and time\n",
    "now = datetime.now()\n",
    "run_name = now.strftime(\"D_%Y_%m_%d_T_%H_%M\")\n",
    "print(\"Current Date and Time:\", run_name)\n",
    "# create a folder called 'results' in the current directory if it doesn't exist\n",
    "if not os.path.exists('results'):\n",
    "    os.mkdir('results')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# SEEDS = [75] \n",
    "SEEDS = [1]  if USER == MOIEN else [1, 4]\n",
    "# SEEDS = [1] if USER == MOIEN else [1, 4, 69, 75, 79, 128, 474, 786, 2048, 3333]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1;45m CROSS VAL 1 \u001b[0m\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 1\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:44<00:00,  7.59it/s, Train_Loss=3.35]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 1 Results: |  train_loss: 3.3532735086942504 |  val_loss: 3.063500313132022  train_acc_top1: 0.07999254206956273 |  val_acc_top1: 0.12221537558405612  train_acc_top5: 0.3521768632629108 |  val_acc_top5: 0.5187309272412403  train_acc_mean_pos: 12.840580374702041 |  val_acc_mean_pos: 7.25860328629543  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 2\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:43<00:00,  7.63it/s, Train_Loss=2.98]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 2 Results: |  train_loss: 2.9827731402826982 |  val_loss: 2.9797160843728294  train_acc_top1: 0.18301606514084506 |  val_acc_top1: 0.19396273469980893  train_acc_top5: 0.6083663438967136 |  val_acc_top5: 0.6344806337860268  train_acc_mean_pos: 6.060384878740065 |  val_acc_mean_pos: 5.832972417974696  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 3\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:43<00:00,  7.61it/s, Train_Loss=2.92]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 3 Results: |  train_loss: 2.9158082591815733 |  val_loss: 2.948641383591952  train_acc_top1: 0.24532656006060297 |  val_acc_top1: 0.20536825116811225  train_acc_top5: 0.689294185262051 |  val_acc_top5: 0.6498591548680139  train_acc_mean_pos: 5.033618813669178 |  val_acc_mean_pos: 5.424922241851197  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 4\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:43<00:00,  7.62it/s, Train_Loss=2.87]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 4 Results: |  train_loss: 2.869849876548763 |  val_loss: 2.8449897391135703  train_acc_top1: 0.28185213713998525 |  val_acc_top1: 0.31070275821596244  train_acc_top5: 0.7326327758215962 |  val_acc_top5: 0.7437235914485555  train_acc_mean_pos: 4.518244449311579 |  val_acc_mean_pos: 4.274304577442402  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 5\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:42<00:00,  7.65it/s, Train_Loss=2.84]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 5 Results: |  train_loss: 2.841607510274005 |  val_loss: 2.824819307372044  train_acc_top1: 0.3082391921203461 |  val_acc_top1: 0.2996728285937242  train_acc_top5: 0.7567396566901409 |  val_acc_top5: 0.7674647887267977  train_acc_mean_pos: 4.210894708552271 |  val_acc_mean_pos: 4.0673195420296535  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 6\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:43<00:00,  7.64it/s, Train_Loss=2.82]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 6 Results: |  train_loss: 2.8175668855648084 |  val_loss: 2.815533054266737  train_acc_top1: 0.33303073650234744 |  val_acc_top1: 0.3150821595404629  train_acc_top5: 0.7791960093896714 |  val_acc_top5: 0.7764509976469854  train_acc_mean_pos: 3.9540694688967135 |  val_acc_mean_pos: 3.9381734155153443  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 7\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:42<00:00,  7.68it/s, Train_Loss=2.8]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 7 Results: |  train_loss: 2.7966199680253374 |  val_loss: 2.7738380448918947  train_acc_top1: 0.3517458920187793 |  val_acc_top1: 0.37018926056897694  train_acc_top5: 0.7980578785211268 |  val_acc_top5: 0.8197681924826662  train_acc_mean_pos: 3.727461120892019 |  val_acc_mean_pos: 3.4883729460093895  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 8\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:45<00:00,  7.56it/s, Train_Loss=2.78]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 8 Results: |  train_loss: 2.7811831615060707 |  val_loss: 2.789650045090438  train_acc_top1: 0.3715950215296566 |  val_acc_top1: 0.36030223000217493  train_acc_top5: 0.8130226672535211 |  val_acc_top5: 0.8138365609824937  train_acc_mean_pos: 3.5686466891860738 |  val_acc_mean_pos: 3.5592356220657275  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 9\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:44<00:00,  7.58it/s, Train_Loss=2.77]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 9 Results: |  train_loss: 2.767548678626477 |  val_loss: 2.7408553473826305  train_acc_top1: 0.387669331975666 |  val_acc_top1: 0.4328447769225483  train_acc_top5: 0.8271530223004695 |  val_acc_top5: 0.8522593896713615  train_acc_mean_pos: 3.4179106758374003 |  val_acc_mean_pos: 3.11504401399496  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 10\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:47<00:00,  7.47it/s, Train_Loss=2.75]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 10 Results: |  train_loss: 2.7530174590472325 |  val_loss: 2.7494244413196762  train_acc_top1: 0.40697745503114424 |  val_acc_top1: 0.4235563380337657  train_acc_top5: 0.8413108861502347 |  val_acc_top5: 0.8383582746478874  train_acc_mean_pos: 3.2601171263926467 |  val_acc_mean_pos: 3.2265375586742526  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 11\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [04:21<00:00,  6.52it/s, Train_Loss=2.7]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 11 Results: |  train_loss: 2.696559253712775 |  val_loss: 2.679952147421143  train_acc_top1: 0.48095779540673106 |  val_acc_top1: 0.5196948356471711  train_acc_top5: 0.8810244278169014 |  val_acc_top5: 0.8952362088250442  train_acc_mean_pos: 2.778196522887324 |  val_acc_mean_pos: 2.5813350940534208  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 12\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [04:20<00:00,  6.54it/s, Train_Loss=2.68]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 12 Results: |  train_loss: 2.6801941266781846 |  val_loss: 2.6754576242026027  train_acc_top1: 0.5004126320422535 |  val_acc_top1: 0.5244204811926739  train_acc_top5: 0.8921654929577465 |  val_acc_top5: 0.8981338027217578  train_acc_mean_pos: 2.6446596244131455 |  val_acc_mean_pos: 2.5607482393582663  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 13\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:52<00:00,  7.32it/s, Train_Loss=2.67]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 13 Results: |  train_loss: 2.67308314396462 |  val_loss: 2.6695622670258716  train_acc_top1: 0.5093529929577465 |  val_acc_top1: 0.5281514084674943  train_acc_top5: 0.8977131015258216 |  val_acc_top5: 0.9007746477921804  train_acc_mean_pos: 2.5897428843896715 |  val_acc_mean_pos: 2.5109184272412404  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 14\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.74it/s, Train_Loss=2.67]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 14 Results: |  train_loss: 2.6683763986182325 |  val_loss: 2.6804145703293347  train_acc_top1: 0.5178562695851349 |  val_acc_top1: 0.5064275233958249  train_acc_top5: 0.9000696889671361 |  val_acc_top5: 0.8931088614072038  train_acc_mean_pos: 2.5594342967993775 |  val_acc_mean_pos: 2.6185108568746718  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 15\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:39<00:00,  7.76it/s, Train_Loss=2.66]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 15 Results: |  train_loss: 2.658834403599372 |  val_loss: 2.6619592233442924  train_acc_top1: 0.5252928159466372 |  val_acc_top1: 0.5384316314274157  train_acc_top5: 0.9070660944835681 |  val_acc_top5: 0.9066431923931193  train_acc_mean_pos: 2.4866276652740202 |  val_acc_mean_pos: 2.459421948368001  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 16\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:48<00:00,  7.45it/s, Train_Loss=2.66]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 16 Results: |  train_loss: 2.6562924167499857 |  val_loss: 2.654599889903001  train_acc_top1: 0.5287436424216754 |  val_acc_top1: 0.5510123239156786  train_acc_top5: 0.9089183538732394 |  val_acc_top5: 0.9137954811255137  train_acc_mean_pos: 2.4647184321438202 |  val_acc_mean_pos: 2.3879518777551785  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 17\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [04:05<00:00,  6.93it/s, Train_Loss=2.65]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 17 Results: |  train_loss: 2.648138537723134 |  val_loss: 2.6589709503549925  train_acc_top1: 0.5359754010405339 |  val_acc_top1: 0.5470715962665181  train_acc_top5: 0.9133014231220657 |  val_acc_top5: 0.907219776944935  train_acc_mean_pos: 2.4142672877776232 |  val_acc_mean_pos: 2.439028755879738  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 18\n",
      "------------------------------- \u001b[0m\n"
     ]

    },

    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:43<00:00,  7.62it/s, Train_Loss=2.65]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 18 Results: |  train_loss: 2.645009145000731 |  val_loss: 2.644586482518156  train_acc_top1: 0.5442952856141636 |  val_acc_top1: 0.566657863872152  train_acc_top5: 0.9150894953051644 |  val_acc_top5: 0.9184903168062649  train_acc_mean_pos: 2.3921440971989028 |  val_acc_mean_pos: 2.331456866062863  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 19\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:44<00:00,  7.59it/s, Train_Loss=2.64]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 19 Results: |  train_loss: 2.6413297748341806 |  val_loss: 2.6490288194916056  train_acc_top1: 0.5481618006884212 |  val_acc_top1: 0.5596522887547811  train_acc_top5: 0.9177303403755869 |  val_acc_top5: 0.9147857980269222  train_acc_mean_pos: 2.3653382971672943 |  val_acc_mean_pos: 2.374417546769263  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 20\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:44<00:00,  7.59it/s, Train_Loss=2.64]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 20 Results: |  train_loss: 2.635787965056482 |  val_loss: 2.644822997106633  train_acc_top1: 0.5555280467643984 |  val_acc_top1: 0.5656514083835441  train_acc_top5: 0.9215540639671361 |  val_acc_top5: 0.9172432510785653  train_acc_mean_pos: 2.3213272691472597 |  val_acc_mean_pos: 2.3396684272748205  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 21\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:44<00:00,  7.60it/s, Train_Loss=2.62]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 21 Results: |  train_loss: 2.620221506192091 |  val_loss: 2.6255597698856405  train_acc_top1: 0.5779018730554782 |  val_acc_top1: 0.597954812184186  train_acc_top5: 0.9311454665492958 |  val_acc_top5: 0.9283568074165935  train_acc_mean_pos: 2.2106226770111093 |  val_acc_mean_pos: 2.1981572769057585  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 22\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:45<00:00,  7.57it/s, Train_Loss=2.62]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 22 Results: |  train_loss: 2.6170561712392617 |  val_loss: 2.623523702084179  train_acc_top1: 0.5819578931924883 |  val_acc_top1: 0.602723004672449  train_acc_top5: 0.9325300762910798 |  val_acc_top5: 0.92989730037434  train_acc_mean_pos: 2.18865536971831 |  val_acc_mean_pos: 2.1799603874134905  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 23\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:44<00:00,  7.58it/s, Train_Loss=2.62]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 23 Results: |  train_loss: 2.6166053926441033 |  val_loss: 2.6264426669044676  train_acc_top1: 0.5838590571438202 |  val_acc_top1: 0.5943705985243891  train_acc_top5: 0.9337679724178404 |  val_acc_top5: 0.9288703050691757  train_acc_mean_pos: 2.1842539612676055 |  val_acc_mean_pos: 2.2031983568075115  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 24\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:44<00:00,  7.60it/s, Train_Loss=2.61]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 24 Results: |  train_loss: 2.6139977158095355 |  val_loss: 2.620511684619205  train_acc_top1: 0.5865763155437411 |  val_acc_top1: 0.6123224765482084  train_acc_top5: 0.9334928843896714 |  val_acc_top5: 0.9318045773696452  train_acc_mean_pos: 2.1736600156261328 |  val_acc_mean_pos: 2.1552127347865575  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 25\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:43<00:00,  7.61it/s, Train_Loss=2.61]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 25 Results: |  train_loss: 2.612550166361209 |  val_loss: 2.622655277520838  train_acc_top1: 0.5859680653597827 |  val_acc_top1: 0.6048136736865335  train_acc_top5: 0.9343823356807511 |  val_acc_top5: 0.9296772299518048  train_acc_mean_pos: 2.172150088028169 |  val_acc_mean_pos: 2.179724178403756  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 26\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:44<00:00,  7.60it/s, Train_Loss=2.61]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 26 Results: |  train_loss: 2.6106677722063423 |  val_loss: 2.6273554501958856  train_acc_top1: 0.5891101819248826 |  val_acc_top1: 0.5890889083835441  train_acc_top5: 0.9353084653755869 |  val_acc_top5: 0.9277332745527438  train_acc_mean_pos: 2.15687353286385 |  val_acc_mean_pos: 2.2090610327295295  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 27\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:44<00:00,  7.58it/s, Train_Loss=2.61]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 27 Results: |  train_loss: 2.610927487483047 |  val_loss: 2.6222115587180768  train_acc_top1: 0.5928605487089202 |  val_acc_top1: 0.6012030515312589  train_acc_top5: 0.9366013791079812 |  val_acc_top5: 0.9297505867593165  train_acc_mean_pos: 2.1459097466978108 |  val_acc_mean_pos: 2.1691607981220655  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 28\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:44<00:00,  7.58it/s, Train_Loss=2.61]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 28 Results: |  train_loss: 2.6087754046972926 |  val_loss: 2.6225900627637695  train_acc_top1: 0.5932701242291871 |  val_acc_top1: 0.5987822768833716  train_acc_top5: 0.9370415199530516 |  val_acc_top5: 0.9295671947405372  train_acc_mean_pos: 2.1385679528327057 |  val_acc_mean_pos: 2.178959800603804  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 29\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:45<00:00,  7.56it/s, Train_Loss=2.61]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 29 Results: |  train_loss: 2.607718471000452 |  val_loss: 2.6173006497638327  train_acc_top1: 0.594859521713615 |  val_acc_top1: 0.6131763497428715  train_acc_top5: 0.9373899647887324 |  val_acc_top5: 0.9323180750222273  train_acc_mean_pos: 2.1273290786384975 |  val_acc_mean_pos: 2.1399750586406725  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 30\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:45<00:00,  7.56it/s, Train_Loss=2.61]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 30 Results: |  train_loss: 2.606740376059438 |  val_loss: 2.620220337675211  train_acc_top1: 0.5974942537400644 |  val_acc_top1: 0.605641138385719  train_acc_top5: 0.9384261296948356 |  val_acc_top5: 0.9318045773696452  train_acc_mean_pos: 2.122487529342723 |  val_acc_mean_pos: 2.15839348600504  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 31\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:45<00:00,  7.56it/s, Train_Loss=2.6] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 31 Results: |  train_loss: 2.6025394917303966 |  val_loss: 2.618545570283988  train_acc_top1: 0.6022410504694836 |  val_acc_top1: 0.6067678989938727  train_acc_top5: 0.9405351379107981 |  val_acc_top5: 0.9326481806560302  train_acc_mean_pos: 2.0950612529342725 |  val_acc_mean_pos: 2.149537852112676  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 32\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:45<00:00,  7.57it/s, Train_Loss=2.6]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 32 Results: |  train_loss: 2.6025788520283544 |  val_loss: 2.614026841982989  train_acc_top1: 0.6025161384976526 |  val_acc_top1: 0.616954225329726  train_acc_top5: 0.9401041666666666 |  val_acc_top5: 0.9347755280738705  train_acc_mean_pos: 2.1010765111502345 |  val_acc_mean_pos: 2.1154005281242405  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 33\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:45<00:00,  7.57it/s, Train_Loss=2.6]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 33 Results: |  train_loss: 2.6008729254695733 |  val_loss: 2.6233557725736234  train_acc_top1: 0.6062114876760564 |  val_acc_top1: 0.5996625585735124  train_acc_top5: 0.9421673268779343 |  val_acc_top5: 0.9314377933320864  train_acc_mean_pos: 2.084109692926138 |  val_acc_mean_pos: 2.1674633216410175  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 34\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:44<00:00,  7.58it/s, Train_Loss=2.6]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 34 Results: |  train_loss: 2.6018013735192485 |  val_loss: 2.6180024012713363  train_acc_top1: 0.6058049687127552 |  val_acc_top1: 0.6072344483344208  train_acc_top5: 0.9407460387323944 |  val_acc_top5: 0.9325748238485184  train_acc_mean_pos: 2.0878570031532098 |  val_acc_mean_pos: 2.1456235328190765  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 35\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:44<00:00,  7.58it/s, Train_Loss=2.6] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 35 Results: |  train_loss: 2.602679522664334 |  val_loss: 2.62048743868098  train_acc_top1: 0.60182841842723 |  val_acc_top1: 0.6043940726580195  train_acc_top5: 0.940865243544601 |  val_acc_top5: 0.9319146125809128  train_acc_mean_pos: 2.1041666666666665 |  val_acc_mean_pos: 2.1599809272748205  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 36\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:43<00:00,  7.61it/s, Train_Loss=2.6]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 36 Results: |  train_loss: 2.602301324966928 |  val_loss: 2.6146485811108153  train_acc_top1: 0.6034972857981221 |  val_acc_top1: 0.6181015257544361  train_acc_top5: 0.9401133362676056 |  val_acc_top5: 0.9353990609377203  train_acc_mean_pos: 2.098004694835681 |  val_acc_mean_pos: 2.1114656690140845  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 37\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:42<00:00,  7.66it/s, Train_Loss=2.6]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 37 Results: |  train_loss: 2.5998161454016055 |  val_loss: 2.615796873267268  train_acc_top1: 0.6063643143700322 |  val_acc_top1: 0.6131499412473939  train_acc_top5: 0.9418463908450704 |  val_acc_top5: 0.9333817487311475  train_acc_mean_pos: 2.0812946254378755 |  val_acc_mean_pos: 2.125806924882629  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 38\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:45<00:00,  7.57it/s, Train_Loss=2.6]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 38 Results: |  train_loss: 2.6023901626817496 |  val_loss: 2.6170345186627526  train_acc_top1: 0.6023021811540697 |  val_acc_top1: 0.6099955985243891  train_acc_top5: 0.940278389084507 |  val_acc_top5: 0.9328315726748095  train_acc_mean_pos: 2.0972038829970248 |  val_acc_mean_pos: 2.1380208333333335  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 39\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:41<00:00,  7.71it/s, Train_Loss=2.6]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 39 Results: |  train_loss: 2.59935524779866 |  val_loss: 2.6191948439593604  train_acc_top1: 0.6037265258215962 |  val_acc_top1: 0.6029533450032624  train_acc_top5: 0.9416446596244131 |  val_acc_top5: 0.9310710092945278  train_acc_mean_pos: 2.0822452073785622 |  val_acc_mean_pos: 2.1580472417840375  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 40\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:46<00:00,  7.54it/s, Train_Loss=2.6]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 40 Results: |  train_loss: 2.6002822543533757 |  val_loss: 2.6119645800389035  train_acc_top1: 0.6032588761737089 |  val_acc_top1: 0.6215757042029654  train_acc_top5: 0.9413695715962441 |  val_acc_top5: 0.9351056337076733  train_acc_mean_pos: 2.0854209458604105 |  val_acc_mean_pos: 2.107221244086682  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.6000e-07.\n",
      "\u001b[1;45m CROSS VAL 2 \u001b[0m\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 1\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:42<00:00,  7.65it/s, Train_Loss=3.29]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 1 Results: |  train_loss: 3.285621945007306 |  val_loss: 3.0722562035484495  train_acc_top1: 0.10071278366964188 |  val_acc_top1: 0.10756015256397042  train_acc_top5: 0.40654342723004694 |  val_acc_top5: 0.47599031681745824  train_acc_mean_pos: 11.398535309053363 |  val_acc_mean_pos: 7.4308348004247105  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 2\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:44<00:00,  7.61it/s, Train_Loss=2.97]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 2 Results: |  train_loss: 2.967476977988588 |  val_loss: 3.023717703393927  train_acc_top1: 0.20825386348184846 |  val_acc_top1: 0.1396229460030934  train_acc_top5: 0.6393320862676056 |  val_acc_top5: 0.5439275234517916  train_acc_mean_pos: 5.703996112089202 |  val_acc_mean_pos: 6.568277581756663  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 3\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:45<00:00,  7.56it/s, Train_Loss=2.91]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 3 Results: |  train_loss: 2.9081614556167046 |  val_loss: 2.8855977931492767  train_acc_top1: 0.2527386541587646 |  val_acc_top1: 0.2555912558181745  train_acc_top5: 0.6998973004694836 |  val_acc_top5: 0.7121068074109969  train_acc_mean_pos: 4.9129040737946825 |  val_acc_mean_pos: 4.688235035524682  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 4\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:42<00:00,  7.64it/s, Train_Loss=2.86]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 4 Results: |  train_loss: 2.8641920066635373 |  val_loss: 2.90826252704495  train_acc_top1: 0.29143131359921937 |  val_acc_top1: 0.22365610329478  train_acc_top5: 0.7391890404929577 |  val_acc_top5: 0.6923224765090316  train_acc_mean_pos: 4.4193136248510205 |  val_acc_mean_pos: 4.838896713346383  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 5\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.74it/s, Train_Loss=2.84]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 5 Results: |  train_loss: 2.8354858770756652 |  val_loss: 2.812800298834071  train_acc_top1: 0.320596757629108 |  val_acc_top1: 0.34933538727917024  train_acc_top5: 0.7652582159624414 |  val_acc_top5: 0.7913057511681122  train_acc_mean_pos: 4.109845706204853 |  val_acc_mean_pos: 3.8028521123626424  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 6\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:45<00:00,  7.56it/s, Train_Loss=2.82]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 6 Results: |  train_loss: 2.8158291579692016 |  val_loss: 2.7932601428367723  train_acc_top1: 0.3413750733568075 |  val_acc_top1: 0.3618632629163948  train_acc_top5: 0.7862107541079812 |  val_acc_top5: 0.8064744717358423  train_acc_mean_pos: 3.884169600938967 |  val_acc_mean_pos: 3.6468309858035592  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 7\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:47<00:00,  7.48it/s, Train_Loss=2.79]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 7 Results: |  train_loss: 2.787897429136043 |  val_loss: 2.75954923887208  train_acc_top1: 0.3716072476525822 |  val_acc_top1: 0.40662705394583687  train_acc_top5: 0.8115280223004695 |  val_acc_top5: 0.8323694247874855  train_acc_mean_pos: 3.5965222760405338 |  val_acc_mean_pos: 3.305941901296517  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 8\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:44<00:00,  7.61it/s, Train_Loss=2.77]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 8 Results: |  train_loss: 2.7736417632987242 |  val_loss: 2.744148767050443  train_acc_top1: 0.38957966549295775 |  val_acc_top1: 0.4253051643360389  train_acc_top5: 0.825053183685446 |  val_acc_top5: 0.8492781689189409  train_acc_mean_pos: 3.431600890109237 |  val_acc_mean_pos: 3.15295774640052  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 9\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:42<00:00,  7.65it/s, Train_Loss=2.76]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 9 Results: |  train_loss: 2.759133478845229 |  val_loss: 2.735452655335547  train_acc_top1: 0.4078913585680751 |  val_acc_top1: 0.43977259386313355  train_acc_top5: 0.8394494571596244 |  val_acc_top5: 0.8551833919236358  train_acc_mean_pos: 3.2748465620575935 |  val_acc_mean_pos: 3.0648239436843587  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.0000e-04.\n",
      "\u001b[92m Epoch 10\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:45<00:00,  7.55it/s, Train_Loss=2.74]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 10 Results: |  train_loss: 2.7430064392411655 |  val_loss: 2.720027638153291  train_acc_top1: 0.42925652875586856 |  val_acc_top1: 0.46107247653701494  train_acc_top5: 0.8541483274647887 |  val_acc_top5: 0.8693779341771569  train_acc_mean_pos: 3.1193148474178405 |  val_acc_mean_pos: 2.9021874999216464  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 11\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:47<00:00,  7.49it/s, Train_Loss=2.68]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 11 Results: |  train_loss: 2.6824657616900724 |  val_loss: 2.673573914268207  train_acc_top1: 0.5017605633802817 |  val_acc_top1: 0.5234301642912654  train_acc_top5: 0.8949988996478874 |  val_acc_top5: 0.9013248238485184  train_acc_mean_pos: 2.638690214201878 |  val_acc_mean_pos: 2.5453447769505315  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 12\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:46<00:00,  7.53it/s, Train_Loss=2.67]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 12 Results: |  train_loss: 2.6700516186409713 |  val_loss: 2.660785133290179  train_acc_top1: 0.5206529978715199 |  val_acc_top1: 0.54887470658956  train_acc_top5: 0.9013992811032864 |  val_acc_top5: 0.9096874999048564  train_acc_mean_pos: 2.5408169503363087 |  val_acc_mean_pos: 2.432866784048752  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 13\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:46<00:00,  7.51it/s, Train_Loss=2.66]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 13 Results: |  train_loss: 2.6593039750773024 |  val_loss: 2.6740799521056697  train_acc_top1: 0.5347252787558685 |  val_acc_top1: 0.5148474178123922  train_acc_top5: 0.9099911971830986 |  val_acc_top5: 0.9027186031912414  train_acc_mean_pos: 2.4574927866039142 |  val_acc_mean_pos: 2.5440287556446775  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 14\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:43<00:00,  7.62it/s, Train_Loss=2.65]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 14 Results: |  train_loss: 2.6543068619261327 |  val_loss: 2.6560725152772355  train_acc_top1: 0.5364461072104078 |  val_acc_top1: 0.5521229459366328  train_acc_top5: 0.912384463028169 |  val_acc_top5: 0.9141255867593165  train_acc_mean_pos: 2.4272361600622885 |  val_acc_mean_pos: 2.398180751173709  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 15\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:43<00:00,  7.61it/s, Train_Loss=2.65]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 15 Results: |  train_loss: 2.6479592920049257 |  val_loss: 2.6556075665872423  train_acc_top1: 0.545606538548436 |  val_acc_top1: 0.5536531689861011  train_acc_top5: 0.916556631455399 |  val_acc_top5: 0.9137221243180019  train_acc_mean_pos: 2.3784049785053227 |  val_acc_mean_pos: 2.3928403753629874  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 16\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:41<00:00,  7.69it/s, Train_Loss=2.65]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 16 Results: |  train_loss: 2.6453278220735243 |  val_loss: 2.650645611431677  train_acc_top1: 0.5504022398395158 |  val_acc_top1: 0.5565404929745366  train_acc_top5: 0.9173452171361502 |  val_acc_top5: 0.9175469483568075  train_acc_mean_pos: 2.3645038634818483 |  val_acc_mean_pos: 2.35625146700183  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 17\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:42<00:00,  7.66it/s, Train_Loss=2.64]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 17 Results: |  train_loss: 2.6395356983925815 |  val_loss: 2.647883158334544  train_acc_top1: 0.5592723004694836 |  val_acc_top1: 0.5596214788900294  train_acc_top5: 0.9219483568075117 |  val_acc_top5: 0.9184903168062649  train_acc_mean_pos: 2.3177969728557155 |  val_acc_mean_pos: 2.3556117955507805  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 18\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:42<00:00,  7.67it/s, Train_Loss=2.64]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 18 Results: |  train_loss: 2.636588417168514 |  val_loss: 2.647712320229257  train_acc_top1: 0.5611001076133039 |  val_acc_top1: 0.5610255281410307  train_acc_top5: 0.9245066754694836 |  val_acc_top5: 0.9191872064776264  train_acc_mean_pos: 2.2955637470657275 |  val_acc_mean_pos: 2.3424237087858675  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 19\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.73it/s, Train_Loss=2.63]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 19 Results: |  train_loss: 2.6298952876402177 |  val_loss: 2.6322623177873137  train_acc_top1: 0.5683013008723796 |  val_acc_top1: 0.5947535211491473  train_acc_top5: 0.9267899061032864 |  val_acc_top5: 0.9291637322992227  train_acc_mean_pos: 2.2636841011579047 |  val_acc_mean_pos: 2.2216945422087475  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 2.0000e-05.\n",
      "\u001b[92m Epoch 20\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.72it/s, Train_Loss=2.63]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 20 Results: |  train_loss: 2.6283961038494335 |  val_loss: 2.639957078745667  train_acc_top1: 0.5748545090217545 |  val_acc_top1: 0.5772109741056469  train_acc_top5: 0.9278352406103286 |  val_acc_top5: 0.9221581571818518  train_acc_mean_pos: 2.2511706524331805 |  val_acc_mean_pos: 2.2933582744687935  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 21\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:41<00:00,  7.71it/s, Train_Loss=2.61]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 21 Results: |  train_loss: 2.6102077321547297 |  val_loss: 2.624087307374802  train_acc_top1: 0.5973781054498444 |  val_acc_top1: 0.6028227699754384  train_acc_top5: 0.9371057071596244 |  val_acc_top5: 0.9318779341771569  train_acc_mean_pos: 2.1355022496320832 |  val_acc_mean_pos: 2.1729958919292325  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 22\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.73it/s, Train_Loss=2.61]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 22 Results: |  train_loss: 2.6079987367274056 |  val_loss: 2.615013663757575  train_acc_top1: 0.5994932267447592 |  val_acc_top1: 0.6213820422759079  train_acc_top5: 0.9385911825117371 |  val_acc_top5: 0.9363526994353729  train_acc_mean_pos: 2.124147227112676 |  val_acc_mean_pos: 2.108237969483568  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 23\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.73it/s, Train_Loss=2.61]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 23 Results: |  train_loss: 2.6063021713159453 |  val_loss: 2.6270638642736444  train_acc_top1: 0.6044784330985915 |  val_acc_top1: 0.5893456572098351  train_acc_top5: 0.9393247505868545 |  val_acc_top5: 0.9301173707968752  train_acc_mean_pos: 2.1058538732394365 |  val_acc_mean_pos: 2.204182805029999  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 24\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.72it/s, Train_Loss=2.61]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 24 Results: |  train_loss: 2.605344902531642 |  val_loss: 2.6143595523117855  train_acc_top1: 0.6047901995305164 |  val_acc_top1: 0.6245833333109466  train_acc_top5: 0.9387745745305164 |  val_acc_top5: 0.9368295186841992  train_acc_mean_pos: 2.0971641481062617 |  val_acc_mean_pos: 2.105864143147715  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 25\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.73it/s, Train_Loss=2.6] \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 25 Results: |  train_loss: 2.6040473812622644 |  val_loss: 2.617852362108902  train_acc_top1: 0.6084427572611911 |  val_acc_top1: 0.6129724177676188  train_acc_top5: 0.9409936179577465 |  val_acc_top5: 0.9352523473226968  train_acc_mean_pos: 2.0887036629266023 |  val_acc_mean_pos: 2.1333934857811725  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 26\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:41<00:00,  7.71it/s, Train_Loss=2.6]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 26 Results: |  train_loss: 2.6011033554522083 |  val_loss: 2.6195851553213987  train_acc_top1: 0.6108604753521126 |  val_acc_top1: 0.6049501173932788  train_acc_top5: 0.9417822036384976 |  val_acc_top5: 0.9348488848813823  train_acc_mean_pos: 2.0742921068075115 |  val_acc_mean_pos: 2.1421757628660245  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 27\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.72it/s, Train_Loss=2.6]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 27 Results: |  train_loss: 2.6015150840153716 |  val_loss: 2.617949940229246  train_acc_top1: 0.6128686179577465 |  val_acc_top1: 0.613449237016445  train_acc_top5: 0.9398565874413145 |  val_acc_top5: 0.9353990609377203  train_acc_mean_pos: 2.0843694982394365 |  val_acc_mean_pos: 2.125460680661627  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 28\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.73it/s, Train_Loss=2.6]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 28 Results: |  train_loss: 2.6000218756601843 |  val_loss: 2.6165339241565113  train_acc_top1: 0.6104233910640081 |  val_acc_top1: 0.6159536385200393  train_acc_top5: 0.9428642165492958 |  val_acc_top5: 0.9369762322992227  train_acc_mean_pos: 2.0705631357823178 |  val_acc_mean_pos: 2.1148870304716585  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 29\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.73it/s, Train_Loss=2.6]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 29 Results: |  train_loss: 2.5988580797199914 |  val_loss: 2.609667058282055  train_acc_top1: 0.6105303697183099 |  val_acc_top1: 0.6253799882852975  train_acc_top5: 0.9425707893192489 |  val_acc_top5: 0.939103579717063  train_acc_mean_pos: 2.0703125 |  val_acc_mean_pos: 2.0753213027273545  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 4.0000e-06.\n",
      "\u001b[92m Epoch 30\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.72it/s, Train_Loss=2.6]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 30 Results: |  train_loss: 2.59670340923916 |  val_loss: 2.614923965763038  train_acc_top1: 0.6144763546790315 |  val_acc_top1: 0.6168236501619849  train_acc_top5: 0.9435060886150235 |  val_acc_top5: 0.9370495891067344  train_acc_mean_pos: 2.053027802230047 |  val_acc_mean_pos: 2.1140478284146305  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 31\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.72it/s, Train_Loss=2.59]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 31 Results: |  train_loss: 2.592178656461653 |  val_loss: 2.609877687664659  train_acc_top1: 0.6217906396713615 |  val_acc_top1: 0.625023474156017  train_acc_top5: 0.9461744424882629 |  val_acc_top5: 0.9384433684494574  train_acc_mean_pos: 2.0227222711267605 |  val_acc_mean_pos: 2.082610035166494  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 32\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.73it/s, Train_Loss=2.59]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 32 Results: |  train_loss: 2.592850559054406 |  val_loss: 2.622704675499822  train_acc_top1: 0.6215583431259007 |  val_acc_top1: 0.5952039319024959  train_acc_top5: 0.9457434712441315 |  val_acc_top5: 0.9345554576513353  train_acc_mean_pos: 2.028065091987171 |  val_acc_mean_pos: 2.16422975343158  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 33\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:41<00:00,  7.71it/s, Train_Loss=2.59]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 33 Results: |  train_loss: 2.5920708188918273 |  val_loss: 2.610428360146536  train_acc_top1: 0.6220840669014085 |  val_acc_top1: 0.6206484742007905  train_acc_top5: 0.9452116343896714 |  val_acc_top5: 0.9398371477921804  train_acc_mean_pos: 2.0279947916666665 |  val_acc_mean_pos: 2.0885357979877455  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 34\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.72it/s, Train_Loss=2.59]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 34 Results: |  train_loss: 2.5926642617989035 |  val_loss: 2.608286613030053  train_acc_top1: 0.6238690825625205 |  val_acc_top1: 0.6260196595964297  train_acc_top5: 0.9455509096244131 |  val_acc_top5: 0.9405340374635419  train_acc_mean_pos: 2.020934198943662 |  val_acc_mean_pos: 2.0713174880390435  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 35\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.74it/s, Train_Loss=2.59]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 35 Results: |  train_loss: 2.5908158622647117 |  val_loss: 2.606694862316472  train_acc_top1: 0.6218181484741784 |  val_acc_top1: 0.630468016359168  train_acc_top5: 0.946156103286385 |  val_acc_top5: 0.9406073942710537  train_acc_mean_pos: 2.019595437206573 |  val_acc_mean_pos: 2.0597329811311105  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 36\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.72it/s, Train_Loss=2.59]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 36 Results: |  train_loss: 2.5920656621456146 |  val_loss: 2.612158538030347  train_acc_top1: 0.6235206377268397 |  val_acc_top1: 0.6185944835904619  train_acc_top5: 0.9454225352112676 |  val_acc_top5: 0.9394703637546217  train_acc_mean_pos: 2.024718187644448 |  val_acc_mean_pos: 2.0962177229599215  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 37\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.73it/s, Train_Loss=2.59]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 37 Results: |  train_loss: 2.5898757814801354 |  val_loss: 2.623444229224478  train_acc_top1: 0.6251100352112676 |  val_acc_top1: 0.5926261737313069  train_acc_top5: 0.9472197769953051 |  val_acc_top5: 0.9340052815949973  train_acc_mean_pos: 2.0041079812206575 |  val_acc_mean_pos: 2.1730897885533005  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 38\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.73it/s, Train_Loss=2.59]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 38 Results: |  train_loss: 2.5901616389902546 |  val_loss: 2.6059945849745487  train_acc_top1: 0.6213015942996096 |  val_acc_top1: 0.6305046947629239  train_acc_top5: 0.9467888057511737 |  val_acc_top5: 0.9415243543649503  train_acc_mean_pos: 2.0176117469167485 |  val_acc_mean_pos: 2.06052376742654  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 39\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.72it/s, Train_Loss=2.59]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 39 Results: |  train_loss: 2.5904224128790303 |  val_loss: 2.6101262614200933  train_acc_top1: 0.6230866099607217 |  val_acc_top1: 0.6230956572042384  train_acc_top5: 0.9477974618544601 |  val_acc_top5: 0.9395437205621334  train_acc_mean_pos: 2.014842527363222 |  val_acc_mean_pos: 2.0855017603842865  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 8.0000e-07.\n",
      "\u001b[92m Epoch 40\n",
      "------------------------------- \u001b[0m\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 1704/1704 [03:40<00:00,  7.71it/s, Train_Loss=2.59]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[96m Epoch 40 Results: |  train_loss: 2.5899335702820006 |  val_loss: 2.6045105988990533  train_acc_top1: 0.6265160407118954 |  val_acc_top1: 0.6325689554214478  train_acc_top5: 0.9461194248826291 |  val_acc_top5: 0.9413042839424152  train_acc_mean_pos: 2.0153560250158042 |  val_acc_mean_pos: 2.05139671357025  \u001b[0m\n",
      "\n",
      "Adjusting learning rate of group 0 to 1.6000e-07.\n"
     ]
    }
   ],
   "source": [
    "best_mae = 1000 # just a big number, since our data is normalized between 0 and 1, mae is between 0 and 1 too.\n",
    "best_seed = SEEDS[0]\n",
    "for idx, seed in enumerate(SEEDS):\n",
    "    print(tc.BOLD_BAKGROUNDs.PURPLE, f\"CROSS VAL {idx+1}\", tc.ENDC)\n",
    "    \n",
    "    \n",
    "    train_dl = DataLoader(train_ds, batch_size=TRAIN_BATCH_SIZE, shuffle=True, num_workers=NUM_WORKERS)\n",
    "    test_dl = DataLoader(test_ds, batch_size=TEST_BATCH_SIZE, shuffle=False, num_workers=NUM_WORKERS)\n",
    "    val_dl = DataLoader(val_ds, batch_size=TEST_BATCH_SIZE, shuffle=False, num_workers=NUM_WORKERS)\n",
    "    \n",
    "    #model = SoilNetFC(cnn_in_channels=12, regresor_input_from_cnn=1024, hidden_size=128).to(device)\n",
    "    # architecture = \"101+GLAM\" if USE_SPATIAL_ATTENTION else \"101\"\n",
    "    if USE_LSTM_BRANCH:\n",
    "        model = SoilNetSimCLR(use_glam=USE_SPATIAL_ATTENTION, cnn_arch= CNN_ARCHITECTURE, reg_version= REG_VERSION,\n",
    "                           cnn_in_channels=len(bands), regresor_input_from_cnn=128,\n",
    "                           lstm_n_features= NUM_CLIMATE_FEATURES, lstm_n_layers= 2, lstm_out= 128,\n",
    "                           hidden_size=128, rnn_arch=RNN_ARCHITECTURE,seq_len=SEQ_LEN).to(device)\n",
    "    else:\n",
    "        model = SoilNet(use_glam=USE_SPATIAL_ATTENTION, cnn_arch = CNN_ARCHITECTURE, reg_version= REG_VERSION,\n",
    "                       cnn_in_channels=len(bands), regresor_input_from_cnn=128, hidden_size=128).to(device)\n",
    "    \n",
    "    random.seed(seed)\n",
    "    np.random.seed(seed)\n",
    "    torch.manual_seed(seed)\n",
    "    \n",
    "    # Saving the model on the last epoch\n",
    "    save_model_path = f\"results/RUN_{EXP_NAME}_{run_name}_{USER}.pth.tar\"\n",
    "    \n",
    "    results = train(model, train_dl, test_dl, val_dl,\n",
    "                    torch.optim.Adam(model.parameters(), lr=LEARNING_RATE),\n",
    "                    SimCLR(temperature=0.5), epochs=NUM_EPOCHS, lr_scheduler=LR_SCHEDULER,\n",
    "                    save_model_path= save_model_path,\n",
    "                    save_model_if_mae_lower_than= best_mae,\n",
    "                    )\n",
    "\n",
    "    \n",
    "    cv_results['train_loss'].append(results['train_loss'])\n",
    "    cv_results['train_acc_top1'].append(results['train_acc_top1'])\n",
    "    cv_results['train_acc_top5'].append(results['train_acc_top5'])\n",
    "    cv_results['train_acc_mean_pos'].append(results['train_acc_mean_pos'])\n",
    "    cv_results['val_loss'].append(results['val_loss'])\n",
    "    cv_results['val_acc_top1'].append(results['val_acc_top1'])\n",
    "    cv_results['val_acc_top5'].append(results['val_acc_top5'])\n",
    "    cv_results['val_acc_mean_pos'].append(results['val_acc_mean_pos'])\n",
    "    \n",
    "\n",
    "    # if results['MAE'][0] < best_mae:\n",
    "    #     best_mae = results['MAE'][0]\n",
    "    #     best_seed = seed\n",
    "    #     print(tc.BOLD_BAKGROUNDs.GREEN, f\"MAE improved to {best_mae}\", tc.ENDC)\n",
    "    \n",
    "        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "torch.save(model, f\"results/RUN_{EXP_NAME}_{run_name}_{USER}_SelfSupervised.pth\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "train_arr = np.asarray(cv_results['train_acc_mean_pos'])\n",
    "val_arr = np.asarray(cv_results['val_acc_mean_pos'])\n",
    "train_arr.shape, val_arr.shape\n",
    "plot_train_test_losses(train_arr,val_arr, title=\"Average Self-Rank\", x_label=\"Epochs\", y_label=\"Rank\",\n",
    "                       min_max_bounds= True, tight_x_lim= True,\n",
    "                       train_legend = \"Train\", test_legend = \"Validation\",\n",
    "                       save_path=f\"results/RUN_{run_name}_{USER}.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = torch.load(f\"results/RUN_{EXP_NAME}_{run_name}_{USER}_SelfSupervised.pth\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "train_arr = np.asarray(cv_results['train_acc_top5'])\n",
    "val_arr = np.asarray(cv_results['val_acc_top5'])\n",
    "train_arr.shape, val_arr.shape\n",
    "plot_train_test_losses(train_arr,val_arr, title=\"Top 5 Prob\", x_label=\"Epochs\", y_label=\"Top 5 Prob\",\n",
    "                       min_max_bounds= True, tight_x_lim= True,\n",
    "                       train_legend = \"Train\", test_legend = \"Validation\",\n",
    "                       save_path=f\"results/RUN_{EXP_NAME}_{run_name}_{USER}.png\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "train_arr = np.asarray(cv_results['train_acc_top1'])\n",
    "val_arr = np.asarray(cv_results['val_acc_top1'])\n",
    "train_arr.shape, val_arr.shape\n",
    "plot_train_test_losses(train_arr,val_arr, title=\"Top 1 Prob\", x_label=\"Epochs\", y_label=\"Top 1 Prob\",\n",
    "                       min_max_bounds= True, tight_x_lim= True,\n",
    "                       train_legend = \"Train\", test_legend = \"Validation\",\n",
    "                       save_path=f\"results/RUN_{EXP_NAME}_{run_name}_{USER}.png\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# END\n",
    "We change the rest if the code worked on full dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import csv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Format the date and time\n",
    "now = datetime.now()\n",
    "finish_string = now.strftime(\"%Y-%m-%d %H:%M:%S\")\n",
    "print(\"Current Date and Time:\", finish_string)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(f\"results/RUN_{EXP_NAME}_{run_name}_{USER}.json\", \"w\") as fp:\n",
    "    json.dump(cv_results, fp, indent=4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "cv_results[\"MAE\"],cv_results['RMSE'],cv_results[\"R2\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "cv_results_full = {}\n",
    "cv_results_full['MAE_MEAN'] = np.mean(cv_results['MAE'])\n",
    "cv_results_full['RMSE_MEAN'] = np.mean(cv_results['RMSE'])\n",
    "cv_results_full['R2_MEAN'] = np.mean(cv_results['R2'])\n",
    "cv_results_full['MAE_MEAN'],cv_results_full['RMSE_MEAN'],cv_results_full['R2_MEAN']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "cv_results_full['USE_LSTM_BRANCH'] = USE_LSTM_BRANCH\n",
    "cv_results_full['NUM_CLIMATE_FEATURES'] = NUM_CLIMATE_FEATURES if USE_LSTM_BRANCH else None\n",
    "cv_results_full['CSV_FILES'] = CSV_FILES if USE_LSTM_BRANCH else None\n",
    "cv_results_full['NUM_WORKERS'] = NUM_WORKERS\n",
    "cv_results_full['TRAIN_BATCH_SIZE'] = TRAIN_BATCH_SIZE\n",
    "cv_results_full['TEST_BATCH_SIZE'] = TEST_BATCH_SIZE\n",
    "cv_results_full['LEARNING_RATE'] = LEARNING_RATE\n",
    "cv_results_full['NUM_EPOCHS'] = NUM_EPOCHS\n",
    "cv_results_full['LR_SCHEDULER'] = LR_SCHEDULER\n",
    "cv_results_full['CNN_ARCHITECTURE'] = CNN_ARCHITECTURE\n",
    "cv_results_full['REG_VERSION'] = REG_VERSION\n",
    "cv_results_full['USE_SPATIAL_ATTENTION'] = USE_SPATIAL_ATTENTION\n",
    "cv_results_full['Best Seed'] = best_seed\n",
    "cv_results_full['SEEDS'] = SEEDS\n",
    "cv_results_full['OC_MAX'] = OC_MAX\n",
    "cv_results_full['USE_SRTM'] = USE_SRTM\n",
    "cv_results_full['TIME'] = {\"start\": start_string, \"finish\": finish_string}\n",
    "cv_results_full['cv_results'] = cv_results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(f\"results/RUN_{EXP_NAME}_{run_name}_{USER}.json\", \"w\") as fp:\n",
    "    json.dump(cv_results_full, fp, indent=4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# READ JSON FILE\n",
    "# with open(\"results/my_dict.json\", \"r\") as fp:\n",
    "#     my_dict = json.load(fp)\n",
    "\n",
    "# print(my_dict)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Analysis"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Save Outputs"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Loading the Best Model\n",
    "Loading the best model from all the coross validation runs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "load_checkpoint(model=model, optimizer=torch.optim.Adam(model.parameters(), lr=LEARNING_RATE),filename=f\"results/RUN_{run_name}_{USER}.pth.tar\")\n",
    "model.eval()\n",
    "print(\"Model loaded\")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "load saved model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# load_checkpoint(model=model, optimizer=torch.optim.Adam(model.parameters(), lr=LEARNING_RATE),\n",
    "#                 filename='C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-5\\\\SoilNet\\\\results\\\\RUN_D_2023_06_05_T_16_02_Nafiseh.pth.tar') #LSTM \n",
    "\n",
    "# load_checkpoint(model=model, optimizer=torch.optim.Adam(model.parameters(), lr=LEARNING_RATE),\n",
    "#                 filename='C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-5\\\\SoilNet\\\\results\\\\RUN_D_2023_06_03_T_09_41_Nafiseh.pth.tar') #GLAM \n",
    "\n",
    "# load_checkpoint(model=model, optimizer=torch.optim.Adam(model.parameters(), lr=LEARNING_RATE),\n",
    "#                 filename='C:\\\\Users\\\\nkakhani\\\\_Multimodal\\\\SoilNet-5\\\\SoilNet\\\\results\\\\RUN_D_2023_06_02_T_20_57_Nafiseh.pth.tar') #Base\n",
    "\n",
    "# model.eval()\n",
    "# print(\"Model loaded\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_dl_w_id = DataLoader(test_ds_w_id, batch_size=TEST_BATCH_SIZE, shuffle=False, num_workers=NUM_WORKERS)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# test_step_w_id(model=model, data_loader=test_dl_w_id, loss_fn=nn.L1Loss(), verbose=False, csv_file=f\"results/RUN_{run_name}_{USER}.csv\")\n",
    "# test_step_w_id(model=model, data_loader=test_dl_w_id, loss_fn=nn.L1Loss(), verbose=False, csv_file=f\"results/onlytest_{run_name}_{USER}.csv\")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Visualize Spatial Attention"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "if not USE_SPATIAL_ATTENTION or USE_LSTM_BRANCH:\n",
    "    raise Exception(\"The Next Part of the Notebook is only available for the model with Spatial Attention and no LSTM Branch\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "BATCH_N = 1\n",
    "\n",
    "IMG_IN_BATCH = 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch_loader = BatchLoader(test_dl_w_id)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "batch = batch_loader(BATCH_N)\n",
    "batch[0].shape, batch[1].shape, len(batch[2])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "import importlib\n",
    "importlib.reload(sys.modules['plot_utils'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from plot_utils.plot import *\n",
    "from PIL import Image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# check if results/attention_maps folder exists if not create it\n",
    "if not os.path.exists(\"results/attention_maps\"):\n",
    "    os.makedirs(\"results/attention_maps\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ABS_ATT = False\n",
    "ALPHA = 0.4\n",
    "with torch.no_grad():\n",
    "    pid = batch[2][IMG_IN_BATCH]\n",
    "    print(\"Point Id:\", pid)\n",
    "    out = model(batch[0].to(device))\n",
    "    print(out.shape)\n",
    "    print(model.cnn.glam.local_spatial_att.att_map.shape)\n",
    "    img_rgb = batch[0][0].permute(1,2,0)[:,:,[3,2,1]].detach().cpu().numpy()\n",
    "    img_rgb = convert2uint8(normalize(img_rgb))\n",
    "    \n",
    "    att_map = model.cnn.glam.local_spatial_att.att_map\n",
    "    att_map = F.interpolate(att_map, size=(64, 64), mode='bicubic', align_corners=True)[IMG_IN_BATCH].detach().cpu().numpy()[0]\n",
    "    att_map = np.abs(att_map) if ABS_ATT else att_map\n",
    "    att_map = convert2uint8(normalize(att_map))   \n",
    "    \n",
    "    colormap = cv2.applyColorMap(att_map, cv2.COLORMAP_JET)\n",
    "    \n",
    "    img_colormaped = cv2.addWeighted(img_rgb, 1 - ALPHA, colormap, ALPHA, 0)\n",
    "\n",
    "    display_images(img_rgb, img_colormaped, [\"Image\", \"Attention Map\"], f\"Local Attention Map Visualization | Point ID: {pid}\", figsize=(10,5))\n",
    "    \n",
    "\n",
    "    \n",
    "    \n",
    "    cv2.imwrite(f\"results/attention_maps/final/_{run_name}_{USER}_pid{pid}_img_lcl_colormaped.png\", cv2.cvtColor(img_colormaped, cv2.COLOR_RGB2BGR))\n",
    "    \n",
    "    \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "GLOB_ATT_CHANNEL = 510 # ranges between 0 and 511\n",
    "with torch.no_grad():\n",
    "    pid = batch[2][IMG_IN_BATCH]\n",
    "    print(\"Point Id:\", pid)\n",
    "    out = model(batch[0].to(device))\n",
    "    print(out.shape)\n",
    "    print(model.cnn.glam.global_spatial_att.att.shape)\n",
    "    img_rgb = batch[0][0].permute(1,2,0)[:,:,[3,2,1]].detach().cpu().numpy()\n",
    "    img_rgb = convert2uint8(normalize(img_rgb))\n",
    "    \n",
    "    att_map = model.cnn.glam.global_spatial_att.att\n",
    "    att_map = F.interpolate(att_map, size=(64, 64), mode='bicubic', align_corners=True)[IMG_IN_BATCH].detach().cpu().numpy()[GLOB_ATT_CHANNEL]\n",
    "    att_map = np.abs(att_map) if ABS_ATT else att_map\n",
    "    att_map = convert2uint8(normalize(att_map))   \n",
    "    \n",
    "    colormap = cv2.applyColorMap(att_map, cv2.COLORMAP_JET)\n",
    "    \n",
    "    img_colormaped = cv2.addWeighted(img_rgb, 1 - ALPHA, colormap, ALPHA, 0)\n",
    "\n",
    "    display_images(img_rgb, img_colormaped, [\"Image\", \"Attention Map\"], f\"Global{GLOB_ATT_CHANNEL} Attention Map Visualization | Point ID: {pid}\") \n",
    "                #    figsize=(10,5), savepath=f\"results/attention_maps/final/RUN_{run_name}_{USER}_pid{pid}_glb{GLOB_ATT_CHANNEL}.png\")\n",
    "    \n",
    "    # # Save img_rgband img_colormaped as jpg files\n",
    "    # cv2.imwrite(f\"results/attention_maps/RUN_{run_name}_{USER}_pid{pid}_img_glb{GLOB_ATT_CHANNEL}_rgb.png\", cv2.cvtColor(img_rgb, cv2.COLOR_RGB2BGR))\n",
    "    # cv2.imwrite(f\"results/attention_maps/RUN_{run_name}_{USER}_pid{pid}_img_glb{GLOB_ATT_CHANNEL}_colormaped.png\", cv2.cvtColor(img_colormaped, cv2.COLOR_RGB2BGR))\n",
    "    \n",
    "    \n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "pytorchGPU",
   "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": "3.9.16"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "5c74e58d50e1361e9a22e65e4f4b85ac48544f43f48a101259d9c47c27371742"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}