human_segmentation/notebooks/models.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "os.chdir(\"..\")\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import cv2\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import segmentation_models_pytorch as smp\n",
    "\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from lib import unet\n",
    "from lib.data import *\n",
    "from lib.models import *\n",
    "from lib.train import *\n",
    "from lib.simple_fcn import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "TRAIN = os.path.join(\"data\", \"train\")\n",
    "TRAIN_MSK = os.path.join(\"data\", \"train_mask\")\n",
    "VAL= os.path.join(\"data\", \"valid\")\n",
    "VAL_MSK = os.path.join(\"data\", \"valid_mask\")\n",
    "\n",
    "DEVICE = \"cuda:0\" if torch.cuda.is_available() else \"cpu\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "logs = {}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "aug = load_augmentations(0.5, 0.1, 0.5, 0.8, 0.3, 0.8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = smp.Unet(\"resnet18\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "optimizer = optim.Adam([\n",
    "    {\"params\": model.encoder.parameters(), \"lr\" : 1e-6},\n",
    "    {\"params\": model.decoder.parameters(), \"lr\" : 1e-4},\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "trainer = Trainer(model, \"models/unet_resnet18_noaug.pth\", \n",
    "                  TRAIN, TRAIN_MSK, \n",
    "                  VAL, VAL_MSK,\n",
    "                  optimizer=optimizer,\n",
    "                  device=DEVICE\n",
    "                 )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch: 0\n",
      "train: 100%|█████████████████| 165/165 [01:02<00:00,  2.64it/s, bce_dice_loss - 0.5244, iou - 0.7845, f-score - 0.8746]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 32.83it/s, bce_dice_loss - 0.3153, iou - 0.8553, f-score - 0.9163]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 1\n",
      "train: 100%|█████████████████| 165/165 [01:04<00:00,  2.55it/s, bce_dice_loss - 0.2501, iou - 0.8949, f-score - 0.9443]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.44it/s, bce_dice_loss - 0.2532, iou - 0.8877, f-score - 0.9367]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 2\n",
      "train: 100%|█████████████████| 165/165 [01:06<00:00,  2.50it/s, bce_dice_loss - 0.1902, iou - 0.9196, f-score - 0.9579]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.27it/s, bce_dice_loss - 0.2084, iou - 0.9061, f-score - 0.9481]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 3\n",
      "train: 100%|██████████████████| 165/165 [01:06<00:00,  2.47it/s, bce_dice_loss - 0.1593, iou - 0.932, f-score - 0.9647]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 34.22it/s, bce_dice_loss - 0.1983, iou - 0.9107, f-score - 0.951]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 4\n",
      "train: 100%|███████████████████| 165/165 [01:06<00:00,  2.47it/s, bce_dice_loss - 0.1408, iou - 0.94, f-score - 0.9689]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.01it/s, bce_dice_loss - 0.1764, iou - 0.9172, f-score - 0.9548]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 5\n",
      "train: 100%|██████████████████| 165/165 [01:07<00:00,  2.45it/s, bce_dice_loss - 0.1261, iou - 0.9456, f-score - 0.972]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 33.11it/s, bce_dice_loss - 0.1751, iou - 0.9216, f-score - 0.9572]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 6\n",
      "train: 100%|█████████████████| 165/165 [01:06<00:00,  2.46it/s, bce_dice_loss - 0.1212, iou - 0.9486, f-score - 0.9734]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.74it/s, bce_dice_loss - 0.1919, iou - 0.9158, f-score - 0.9538]\n",
      "\n",
      "Epoch: 7\n",
      "train: 100%|██████████████████| 165/165 [01:07<00:00,  2.44it/s, bce_dice_loss - 0.1091, iou - 0.9532, f-score - 0.976]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.50it/s, bce_dice_loss - 0.1856, iou - 0.9113, f-score - 0.9505]\n",
      "\n",
      "Epoch: 8\n",
      "train: 100%|█████████████████| 165/165 [01:07<00:00,  2.44it/s, bce_dice_loss - 0.09551, iou - 0.9589, f-score - 0.979]\n",
      "valid: 100%|███████████████████| 100/100 [00:02<00:00, 33.95it/s, bce_dice_loss - 0.157, iou - 0.928, f-score - 0.9611]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 9\n",
      "train: 100%|████████████████| 165/165 [01:08<00:00,  2.42it/s, bce_dice_loss - 0.08978, iou - 0.9613, f-score - 0.9803]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 33.81it/s, bce_dice_loss - 0.163, iou - 0.9256, f-score - 0.9592]\n",
      "\n",
      "Epoch: 10\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.44it/s, bce_dice_loss - 0.08461, iou - 0.9638, f-score - 0.9815]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 33.81it/s, bce_dice_loss - 0.1617, iou - 0.9281, f-score - 0.961]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 11\n",
      "train: 100%|█████████████████| 165/165 [01:07<00:00,  2.43it/s, bce_dice_loss - 0.08196, iou - 0.9647, f-score - 0.982]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.97it/s, bce_dice_loss - 0.1608, iou - 0.9286, f-score - 0.9612]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 12\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.44it/s, bce_dice_loss - 0.07743, iou - 0.9664, f-score - 0.9829]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.70it/s, bce_dice_loss - 0.1586, iou - 0.9319, f-score - 0.9631]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 13\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.43it/s, bce_dice_loss - 0.07256, iou - 0.9683, f-score - 0.9839]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.61it/s, bce_dice_loss - 0.1656, iou - 0.9293, f-score - 0.9616]\n",
      "\n",
      "Epoch: 14\n",
      "train: 100%|██████████████████| 165/165 [01:07<00:00,  2.43it/s, bce_dice_loss - 0.066, iou - 0.9713, f-score - 0.9854]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.70it/s, bce_dice_loss - 0.1587, iou - 0.9309, f-score - 0.9626]\n",
      "\n",
      "Epoch: 15\n",
      "train: 100%|█████████████████| 165/165 [01:07<00:00,  2.44it/s, bce_dice_loss - 0.06301, iou - 0.9724, f-score - 0.986]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.64it/s, bce_dice_loss - 0.1567, iou - 0.9343, f-score - 0.9642]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 16\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.43it/s, bce_dice_loss - 0.06276, iou - 0.9725, f-score - 0.9861]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 33.91it/s, bce_dice_loss - 0.1532, iou - 0.9338, f-score - 0.964]\n",
      "\n",
      "Epoch: 17\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.43it/s, bce_dice_loss - 0.05962, iou - 0.9738, f-score - 0.9867]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.82it/s, bce_dice_loss - 0.1619, iou - 0.9297, f-score - 0.9615]\n",
      "\n",
      "Epoch: 18\n",
      "train: 100%|████████████████| 165/165 [01:08<00:00,  2.43it/s, bce_dice_loss - 0.05718, iou - 0.9748, f-score - 0.9873]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.34it/s, bce_dice_loss - 0.1718, iou - 0.9333, f-score - 0.9635]\n",
      "\n",
      "Epoch: 19\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.44it/s, bce_dice_loss - 0.05667, iou - 0.9749, f-score - 0.9873]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 33.99it/s, bce_dice_loss - 0.1811, iou - 0.9291, f-score - 0.961]\n",
      "\n",
      "Epoch: 20\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.43it/s, bce_dice_loss - 0.05613, iou - 0.9752, f-score - 0.9874]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.39it/s, bce_dice_loss - 0.1624, iou - 0.9326, f-score - 0.9633]\n",
      "\n",
      "Epoch: 21\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.44it/s, bce_dice_loss - 0.05216, iou - 0.9769, f-score - 0.9883]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.92it/s, bce_dice_loss - 0.1537, iou - 0.9369, f-score - 0.9657]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 22\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.44it/s, bce_dice_loss - 0.05097, iou - 0.9773, f-score - 0.9885]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 33.99it/s, bce_dice_loss - 0.1622, iou - 0.9353, f-score - 0.965]\n",
      "\n",
      "Epoch: 23\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.46it/s, bce_dice_loss - 0.05394, iou - 0.9761, f-score - 0.9879]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.83it/s, bce_dice_loss - 0.1695, iou - 0.9353, f-score - 0.9643]\n",
      "\n",
      "Epoch: 24\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.43it/s, bce_dice_loss - 0.04947, iou - 0.9779, f-score - 0.9888]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.61it/s, bce_dice_loss - 0.1588, iou - 0.9357, f-score - 0.9652]\n",
      "\n",
      "Epoch: 25\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.45it/s, bce_dice_loss - 0.04786, iou - 0.9786, f-score - 0.9892]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.68it/s, bce_dice_loss - 0.1612, iou - 0.9353, f-score - 0.9648]\n",
      "\n",
      "Epoch: 26\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.44it/s, bce_dice_loss - 0.05099, iou - 0.9772, f-score - 0.9885]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.91it/s, bce_dice_loss - 0.1625, iou - 0.9332, f-score - 0.9636]\n",
      "\n",
      "Epoch: 27\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.44it/s, bce_dice_loss - 0.04661, iou - 0.9791, f-score - 0.9894]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.76it/s, bce_dice_loss - 0.1631, iou - 0.9355, f-score - 0.9647]\n",
      "\n",
      "Epoch: 28\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.46it/s, bce_dice_loss - 0.04471, iou - 0.9798, f-score - 0.9898]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 33.51it/s, bce_dice_loss - 0.1592, iou - 0.9361, f-score - 0.965]\n",
      "\n",
      "Epoch: 29\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.44it/s, bce_dice_loss - 0.04948, iou - 0.9779, f-score - 0.9888]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.66it/s, bce_dice_loss - 0.1689, iou - 0.9344, f-score - 0.9642]\n",
      "\n",
      "Epoch: 30\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.43it/s, bce_dice_loss - 0.04326, iou - 0.9804, f-score - 0.9901]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 33.94it/s, bce_dice_loss - 0.1687, iou - 0.936, f-score - 0.9649]\n",
      "\n",
      "Epoch: 31\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.45it/s, bce_dice_loss - 0.04452, iou - 0.9798, f-score - 0.9898]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 33.58it/s, bce_dice_loss - 0.157, iou - 0.9369, f-score - 0.9659]\n",
      "\n",
      "Epoch: 32\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.43it/s, bce_dice_loss - 0.04222, iou - 0.9809, f-score - 0.9903]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 33.74it/s, bce_dice_loss - 0.178, iou - 0.9358, f-score - 0.9648]\n",
      "\n",
      "Epoch: 33\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.46it/s, bce_dice_loss - 0.04426, iou - 0.9801, f-score - 0.9899]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 33.28it/s, bce_dice_loss - 0.1754, iou - 0.9333, f-score - 0.9637]\n",
      "\n",
      "Epoch: 34\n",
      "train: 100%|██████████████████| 165/165 [01:07<00:00,  2.44it/s, bce_dice_loss - 0.04363, iou - 0.9803, f-score - 0.99]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.68it/s, bce_dice_loss - 0.1693, iou - 0.9339, f-score - 0.9639]\n",
      "\n",
      "Epoch: 35\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.45it/s, bce_dice_loss - 0.04141, iou - 0.9812, f-score - 0.9905]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.90it/s, bce_dice_loss - 0.1604, iou - 0.9364, f-score - 0.9652]\n",
      "\n",
      "Epoch: 36\n",
      "train: 100%|█████████████████| 165/165 [01:07<00:00,  2.43it/s, bce_dice_loss - 0.03963, iou - 0.982, f-score - 0.9909]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 33.29it/s, bce_dice_loss - 0.1584, iou - 0.9383, f-score - 0.9665]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 37\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.46it/s, bce_dice_loss - 0.03905, iou - 0.9823, f-score - 0.9911]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.40it/s, bce_dice_loss - 0.1619, iou - 0.9381, f-score - 0.9664]\n",
      "\n",
      "Epoch: 38\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.45it/s, bce_dice_loss - 0.03821, iou - 0.9827, f-score - 0.9913]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.49it/s, bce_dice_loss - 0.1677, iou - 0.9356, f-score - 0.9645]\n",
      "\n",
      "Epoch: 39\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.45it/s, bce_dice_loss - 0.03834, iou - 0.9825, f-score - 0.9911]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.89it/s, bce_dice_loss - 0.1783, iou - 0.9338, f-score - 0.9636]\n",
      "\n",
      "Epoch: 40\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.43it/s, bce_dice_loss - 0.04068, iou - 0.9816, f-score - 0.9907]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.02it/s, bce_dice_loss - 0.1662, iou - 0.9383, f-score - 0.9664]\n",
      "\n",
      "Epoch: 41\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.45it/s, bce_dice_loss - 0.03827, iou - 0.9826, f-score - 0.9912]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.49it/s, bce_dice_loss - 0.1632, iou - 0.9382, f-score - 0.9664]\n",
      "\n",
      "Epoch: 42\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.45it/s, bce_dice_loss - 0.03684, iou - 0.9832, f-score - 0.9915]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.67it/s, bce_dice_loss - 0.1647, iou - 0.9392, f-score - 0.9669]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 43\n",
      "train: 100%|█████████████████| 165/165 [01:07<00:00,  2.45it/s, bce_dice_loss - 0.03502, iou - 0.984, f-score - 0.9919]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.71it/s, bce_dice_loss - 0.1668, iou - 0.9392, f-score - 0.9669]\n",
      "\n",
      "Epoch: 44\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.45it/s, bce_dice_loss - 0.03625, iou - 0.9835, f-score - 0.9917]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.84it/s, bce_dice_loss - 0.1648, iou - 0.9384, f-score - 0.9663]\n",
      "\n",
      "Epoch: 45\n",
      "train: 100%|█████████████████| 165/165 [01:07<00:00,  2.45it/s, bce_dice_loss - 0.03505, iou - 0.984, f-score - 0.9919]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.46it/s, bce_dice_loss - 0.1683, iou - 0.9388, f-score - 0.9668]\n",
      "\n",
      "Epoch: 46\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.44it/s, bce_dice_loss - 0.03533, iou - 0.9837, f-score - 0.9918]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.44it/s, bce_dice_loss - 0.1728, iou - 0.9391, f-score - 0.9667]\n",
      "\n",
      "Epoch: 47\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.46it/s, bce_dice_loss - 0.03443, iou - 0.9842, f-score - 0.9921]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.02it/s, bce_dice_loss - 0.1709, iou - 0.9377, f-score - 0.9661]\n",
      "\n",
      "Epoch: 48\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.45it/s, bce_dice_loss - 0.03312, iou - 0.9848, f-score - 0.9924]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.48it/s, bce_dice_loss - 0.1839, iou - 0.9349, f-score - 0.9643]\n",
      "\n",
      "Epoch: 49\n",
      "train: 100%|████████████████| 165/165 [01:07<00:00,  2.46it/s, bce_dice_loss - 0.03294, iou - 0.9849, f-score - 0.9924]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.68it/s, bce_dice_loss - 0.1639, iou - 0.9396, f-score - 0.9669]\n",
      "Model saved!\n",
      "Training completed sucessfully.\n",
      "Best model IoU: 0.9396085813641547\n"
     ]
    }
   ],
   "source": [
    "results = trainer.fit(50, 8, 2)\n",
    "logs[\"resnet18_noaug\"] = results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = smp.Unet(\"resnet18\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "optimizer = optim.Adam([\n",
    "    {\"params\": model.encoder.parameters(), \"lr\" : 1e-6},\n",
    "    {\"params\": model.decoder.parameters(), \"lr\" : 1e-4},\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "trainer = Trainer(model, \"models/unet_resnet18.pth\", \n",
    "                  TRAIN, TRAIN_MSK, \n",
    "                  VAL, VAL_MSK,\n",
    "                  augmentations=aug,\n",
    "                  optimizer=optimizer,\n",
    "                  device=DEVICE\n",
    "                 )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch: 0\n",
      "train: 100%|██████████████████| 165/165 [01:22<00:00,  1.99it/s, bce_dice_loss - 0.8773, iou - 0.6536, f-score - 0.777]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 34.72it/s, bce_dice_loss - 0.358, iou - 0.8393, f-score - 0.9084]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 1\n",
      "train: 100%|██████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.4473, iou - 0.8044, f-score - 0.891]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 34.91it/s, bce_dice_loss - 0.298, iou - 0.8646, f-score - 0.9233]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 2\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.3958, iou - 0.8254, f-score - 0.9037]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 34.43it/s, bce_dice_loss - 0.2673, iou - 0.875, f-score - 0.9291]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 3\n",
      "train: 100%|██████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.379, iou - 0.8336, f-score - 0.9085]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.75it/s, bce_dice_loss - 0.2329, iou - 0.8912, f-score - 0.9395]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 4\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.3315, iou - 0.8537, f-score - 0.9206]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.88it/s, bce_dice_loss - 0.2305, iou - 0.8918, f-score - 0.9388]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 5\n",
      "train: 100%|██████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.335, iou - 0.8522, f-score - 0.9197]\n",
      "valid: 100%|████████████████████| 100/100 [00:02<00:00, 34.61it/s, bce_dice_loss - 0.2, iou - 0.9057, f-score - 0.9482]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 6\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.3099, iou - 0.8624, f-score - 0.9255]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 35.10it/s, bce_dice_loss - 0.1892, iou - 0.9108, f-score - 0.9511]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 7\n",
      "train: 100%|██████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.3015, iou - 0.867, f-score - 0.9283]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.31it/s, bce_dice_loss - 0.1854, iou - 0.9136, f-score - 0.9526]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 8\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.2864, iou - 0.8733, f-score - 0.9319]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 34.94it/s, bce_dice_loss - 0.1992, iou - 0.9046, f-score - 0.947]\n",
      "\n",
      "Epoch: 9\n",
      "train: 100%|███████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.279, iou - 0.8768, f-score - 0.934]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 34.43it/s, bce_dice_loss - 0.1728, iou - 0.918, f-score - 0.9555]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 10\n",
      "train: 100%|███████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.2754, iou - 0.877, f-score - 0.934]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.30it/s, bce_dice_loss - 0.1833, iou - 0.9152, f-score - 0.9538]\n",
      "\n",
      "Epoch: 11\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.2622, iou - 0.8832, f-score - 0.9376]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 34.61it/s, bce_dice_loss - 0.1665, iou - 0.919, f-score - 0.9557]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 12\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.2609, iou - 0.8834, f-score - 0.9377]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 35.36it/s, bce_dice_loss - 0.1629, iou - 0.9234, f-score - 0.9584]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 13\n",
      "train: 100%|██████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.2556, iou - 0.8858, f-score - 0.939]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.60it/s, bce_dice_loss - 0.1619, iou - 0.9256, f-score - 0.9596]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 14\n",
      "train: 100%|█████████████████| 165/165 [01:21<00:00,  2.01it/s, bce_dice_loss - 0.2448, iou - 0.8916, f-score - 0.9423]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.77it/s, bce_dice_loss - 0.1678, iou - 0.9208, f-score - 0.9566]\n",
      "\n",
      "Epoch: 15\n",
      "train: 100%|██████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.244, iou - 0.8913, f-score - 0.9422]\n",
      "valid: 100%|███████████████████| 100/100 [00:02<00:00, 34.63it/s, bce_dice_loss - 0.157, iou - 0.9242, f-score - 0.959]\n",
      "\n",
      "Epoch: 16\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.2391, iou - 0.8938, f-score - 0.9436]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.56it/s, bce_dice_loss - 0.1555, iou - 0.9298, f-score - 0.9619]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 17\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.2335, iou - 0.8955, f-score - 0.9445]\n",
      "valid: 100%|███████████████████| 100/100 [00:02<00:00, 34.72it/s, bce_dice_loss - 0.148, iou - 0.9318, f-score - 0.963]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 18\n",
      "train: 100%|██████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.2331, iou - 0.897, f-score - 0.9453]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.49it/s, bce_dice_loss - 0.1421, iou - 0.9331, f-score - 0.9639]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 19\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.2278, iou - 0.8983, f-score - 0.9461]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.55it/s, bce_dice_loss - 0.1399, iou - 0.9331, f-score - 0.9636]\n",
      "\n",
      "Epoch: 20\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.2256, iou - 0.8998, f-score - 0.9469]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.50it/s, bce_dice_loss - 0.1382, iou - 0.9339, f-score - 0.9641]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 21\n",
      "train: 100%|██████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.2268, iou - 0.8999, f-score - 0.947]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.55it/s, bce_dice_loss - 0.1394, iou - 0.9331, f-score - 0.9636]\n",
      "\n",
      "Epoch: 22\n",
      "train: 100%|██████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.225, iou - 0.9008, f-score - 0.9475]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 34.56it/s, bce_dice_loss - 0.146, iou - 0.9318, f-score - 0.9629]\n",
      "\n",
      "Epoch: 23\n",
      "train: 100%|██████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.2215, iou - 0.9018, f-score - 0.948]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 34.36it/s, bce_dice_loss - 0.1385, iou - 0.9334, f-score - 0.964]\n",
      "\n",
      "Epoch: 24\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.2183, iou - 0.9033, f-score - 0.9488]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 34.84it/s, bce_dice_loss - 0.1377, iou - 0.9333, f-score - 0.964]\n",
      "\n",
      "Epoch: 25\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.2148, iou - 0.9039, f-score - 0.9492]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 34.63it/s, bce_dice_loss - 0.1371, iou - 0.934, f-score - 0.9643]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 26\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.2058, iou - 0.9091, f-score - 0.9521]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 35.03it/s, bce_dice_loss - 0.1422, iou - 0.9347, f-score - 0.9645]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 27\n",
      "train: 100%|██████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.2129, iou - 0.905, f-score - 0.9499]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 34.82it/s, bce_dice_loss - 0.1383, iou - 0.934, f-score - 0.9641]\n",
      "\n",
      "Epoch: 28\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.2119, iou - 0.9052, f-score - 0.9499]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.48it/s, bce_dice_loss - 0.1371, iou - 0.9352, f-score - 0.9647]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 29\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.2022, iou - 0.9105, f-score - 0.9528]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.85it/s, bce_dice_loss - 0.1279, iou - 0.9385, f-score - 0.9667]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 30\n",
      "train: 100%|██████████████████| 165/165 [01:21<00:00,  2.02it/s, bce_dice_loss - 0.2067, iou - 0.9073, f-score - 0.951]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 34.48it/s, bce_dice_loss - 0.128, iou - 0.9378, f-score - 0.9662]\n",
      "\n",
      "Epoch: 31\n",
      "train: 100%|█████████████████| 165/165 [01:21<00:00,  2.01it/s, bce_dice_loss - 0.1989, iou - 0.9109, f-score - 0.9531]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.84it/s, bce_dice_loss - 0.1269, iou - 0.9397, f-score - 0.9674]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 32\n",
      "train: 100%|█████████████████| 165/165 [01:21<00:00,  2.01it/s, bce_dice_loss - 0.1961, iou - 0.9129, f-score - 0.9542]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 35.03it/s, bce_dice_loss - 0.132, iou - 0.9387, f-score - 0.9667]\n",
      "\n",
      "Epoch: 33\n",
      "train: 100%|██████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.196, iou - 0.9121, f-score - 0.9537]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.67it/s, bce_dice_loss - 0.1364, iou - 0.9371, f-score - 0.9657]\n",
      "\n",
      "Epoch: 34\n",
      "train: 100%|█████████████████| 165/165 [01:21<00:00,  2.02it/s, bce_dice_loss - 0.1938, iou - 0.9118, f-score - 0.9536]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.56it/s, bce_dice_loss - 0.1386, iou - 0.9368, f-score - 0.9657]\n",
      "\n",
      "Epoch: 35\n",
      "train: 100%|██████████████████| 165/165 [01:21<00:00,  2.02it/s, bce_dice_loss - 0.1905, iou - 0.915, f-score - 0.9554]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.23it/s, bce_dice_loss - 0.1392, iou - 0.9364, f-score - 0.9654]\n",
      "\n",
      "Epoch: 36\n",
      "train: 100%|██████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.187, iou - 0.9163, f-score - 0.9561]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 34.92it/s, bce_dice_loss - 0.125, iou - 0.9392, f-score - 0.9669]\n",
      "\n",
      "Epoch: 37\n",
      "train: 100%|███████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.191, iou - 0.9142, f-score - 0.955]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.71it/s, bce_dice_loss - 0.1285, iou - 0.9395, f-score - 0.9672]\n",
      "\n",
      "Epoch: 38\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.1907, iou - 0.9146, f-score - 0.9551]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 32.43it/s, bce_dice_loss - 0.1301, iou - 0.9403, f-score - 0.9675]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 39\n",
      "train: 100%|██████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.1986, iou - 0.9108, f-score - 0.953]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 33.82it/s, bce_dice_loss - 0.1305, iou - 0.9383, f-score - 0.9667]\n",
      "\n",
      "Epoch: 40\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.1912, iou - 0.9139, f-score - 0.9548]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 34.29it/s, bce_dice_loss - 0.1274, iou - 0.9396, f-score - 0.967]\n",
      "\n",
      "Epoch: 41\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.1811, iou - 0.9188, f-score - 0.9575]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.66it/s, bce_dice_loss - 0.1296, iou - 0.9406, f-score - 0.9677]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 42\n",
      "train: 100%|█████████████████| 165/165 [01:21<00:00,  2.01it/s, bce_dice_loss - 0.1793, iou - 0.9192, f-score - 0.9577]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.60it/s, bce_dice_loss - 0.1227, iou - 0.9416, f-score - 0.9683]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 43\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.1863, iou - 0.9173, f-score - 0.9566]\n",
      "valid: 100%|███████████████████| 100/100 [00:02<00:00, 34.95it/s, bce_dice_loss - 0.1238, iou - 0.941, f-score - 0.968]\n",
      "\n",
      "Epoch: 44\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.1778, iou - 0.9205, f-score - 0.9584]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.88it/s, bce_dice_loss - 0.1249, iou - 0.9423, f-score - 0.9687]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 45\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.01it/s, bce_dice_loss - 0.1845, iou - 0.9169, f-score - 0.9564]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.94it/s, bce_dice_loss - 0.1223, iou - 0.9425, f-score - 0.9688]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 46\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.1818, iou - 0.9184, f-score - 0.9572]\n",
      "valid: 100%|███████████████████| 100/100 [00:02<00:00, 34.94it/s, bce_dice_loss - 0.1248, iou - 0.94, f-score - 0.9674]\n",
      "\n",
      "Epoch: 47\n",
      "train: 100%|█████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.1835, iou - 0.9171, f-score - 0.9565]\n",
      "valid: 100%|██████████████████| 100/100 [00:02<00:00, 34.90it/s, bce_dice_loss - 0.1183, iou - 0.9429, f-score - 0.969]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 48\n",
      "train: 100%|██████████████████| 165/165 [01:22<00:00,  2.00it/s, bce_dice_loss - 0.1832, iou - 0.918, f-score - 0.9569]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.53it/s, bce_dice_loss - 0.1261, iou - 0.9403, f-score - 0.9675]\n",
      "\n",
      "Epoch: 49\n",
      "train: 100%|█████████████████| 165/165 [01:21<00:00,  2.02it/s, bce_dice_loss - 0.1724, iou - 0.9228, f-score - 0.9597]\n",
      "valid: 100%|█████████████████| 100/100 [00:02<00:00, 34.97it/s, bce_dice_loss - 0.1161, iou - 0.9446, f-score - 0.9702]\n",
      "Model saved!\n",
      "Training completed sucessfully.\n",
      "Best model IoU: 0.9445689922571183\n"
     ]
    }
   ],
   "source": [
    "results = trainer.fit(50, 8, 2)\n",
    "logs[\"resnet18\"] = results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = smp.Unet(\"vgg11\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "optimizer = optim.Adam([\n",
    "    {\"params\": model.encoder.parameters(), \"lr\" : 1e-6},\n",
    "    {\"params\": model.decoder.parameters(), \"lr\" : 1e-4},\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "trainer = Trainer(model, \"models/unet_vgg11.pth\", \n",
    "                  TRAIN, TRAIN_MSK, \n",
    "                  VAL, VAL_MSK,\n",
    "                  augmentations=aug,\n",
    "                  optimizer=optimizer,\n",
    "                  device=DEVICE\n",
    "                 )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch: 0\n",
      "train: 100%|███████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 1.28, iou - 0.5588, f-score - 0.7146]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.22it/s, bce_dice_loss - 0.9943, iou - 0.6049, f-score - 0.7386]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 1\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.9589, iou - 0.6181, f-score - 0.7624]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.26it/s, bce_dice_loss - 0.7616, iou - 0.6749, f-score - 0.7934]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 2\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.7636, iou - 0.6801, f-score - 0.8084]\n",
      "valid: 100%|██████████████████| 100/100 [00:03<00:00, 25.15it/s, bce_dice_loss - 0.584, iou - 0.7379, f-score - 0.8394]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 3\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.6355, iou - 0.7259, f-score - 0.8402]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.16it/s, bce_dice_loss - 0.5029, iou - 0.7682, f-score - 0.8602]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 4\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.5682, iou - 0.7566, f-score - 0.8605]\n",
      "valid: 100%|██████████████████| 100/100 [00:03<00:00, 25.17it/s, bce_dice_loss - 0.4371, iou - 0.802, f-score - 0.8833]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 5\n",
      "train: 100%|██████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.528, iou - 0.7749, f-score - 0.8723]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.31it/s, bce_dice_loss - 0.4053, iou - 0.8229, f-score - 0.8959]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 6\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.4813, iou - 0.7962, f-score - 0.8859]\n",
      "valid: 100%|██████████████████| 100/100 [00:04<00:00, 24.91it/s, bce_dice_loss - 0.3556, iou - 0.843, f-score - 0.9092]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 7\n",
      "train: 100%|██████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.4567, iou - 0.8062, f-score - 0.892]\n",
      "valid: 100%|█████████████████| 100/100 [00:04<00:00, 24.84it/s, bce_dice_loss - 0.3687, iou - 0.8444, f-score - 0.9094]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 8\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.4415, iou - 0.8124, f-score - 0.8958]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.23it/s, bce_dice_loss - 0.3541, iou - 0.8511, f-score - 0.9134]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 9\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.4267, iou - 0.8181, f-score - 0.8994]\n",
      "valid: 100%|█████████████████| 100/100 [00:04<00:00, 24.76it/s, bce_dice_loss - 0.3197, iou - 0.8547, f-score - 0.9154]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 10\n",
      "train: 100%|██████████████████| 165/165 [02:01<00:00,  1.36it/s, bce_dice_loss - 0.414, iou - 0.8238, f-score - 0.9027]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.32it/s, bce_dice_loss - 0.3007, iou - 0.8685, f-score - 0.9247]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 11\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.3919, iou - 0.8325, f-score - 0.9081]\n",
      "valid: 100%|██████████████████| 100/100 [00:03<00:00, 25.22it/s, bce_dice_loss - 0.2851, iou - 0.872, f-score - 0.9271]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 12\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.3925, iou - 0.8324, f-score - 0.9079]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.22it/s, bce_dice_loss - 0.2878, iou - 0.8742, f-score - 0.9277]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 13\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.3776, iou - 0.8385, f-score - 0.9117]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.21it/s, bce_dice_loss - 0.2728, iou - 0.8781, f-score - 0.9308]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 14\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.3743, iou - 0.8401, f-score - 0.9124]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.27it/s, bce_dice_loss - 0.3474, iou - 0.8581, f-score - 0.9175]\n",
      "\n",
      "Epoch: 15\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.3598, iou - 0.8446, f-score - 0.9152]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.14it/s, bce_dice_loss - 0.2649, iou - 0.8783, f-score - 0.9312]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 16\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.3476, iou - 0.8517, f-score - 0.9192]\n",
      "valid: 100%|███████████████████| 100/100 [00:03<00:00, 25.22it/s, bce_dice_loss - 0.25, iou - 0.8871, f-score - 0.9365]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 17\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.3475, iou - 0.8499, f-score - 0.9182]\n",
      "valid: 100%|██████████████████| 100/100 [00:03<00:00, 25.23it/s, bce_dice_loss - 0.2516, iou - 0.8882, f-score - 0.937]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 18\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.3469, iou - 0.8509, f-score - 0.9189]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.16it/s, bce_dice_loss - 0.2612, iou - 0.8849, f-score - 0.9348]\n",
      "\n",
      "Epoch: 19\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.3364, iou - 0.8549, f-score - 0.9213]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.14it/s, bce_dice_loss - 0.2356, iou - 0.8933, f-score - 0.9397]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 20\n",
      "train: 100%|█████████████████| 165/165 [01:59<00:00,  1.38it/s, bce_dice_loss - 0.3357, iou - 0.8556, f-score - 0.9217]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.22it/s, bce_dice_loss - 0.2391, iou - 0.8929, f-score - 0.9396]\n",
      "\n",
      "Epoch: 21\n",
      "train: 100%|██████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.3271, iou - 0.8579, f-score - 0.923]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.30it/s, bce_dice_loss - 0.2374, iou - 0.8926, f-score - 0.9396]\n",
      "\n",
      "Epoch: 22\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.3289, iou - 0.8575, f-score - 0.9227]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.19it/s, bce_dice_loss - 0.2377, iou - 0.8926, f-score - 0.9397]\n",
      "\n",
      "Epoch: 23\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.3247, iou - 0.8591, f-score - 0.9238]\n",
      "valid: 100%|██████████████████| 100/100 [00:03<00:00, 25.13it/s, bce_dice_loss - 0.2314, iou - 0.897, f-score - 0.9427]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 24\n",
      "train: 100%|██████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.3156, iou - 0.863, f-score - 0.9259]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.14it/s, bce_dice_loss - 0.2305, iou - 0.8965, f-score - 0.9421]\n",
      "\n",
      "Epoch: 25\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.3133, iou - 0.8641, f-score - 0.9266]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.10it/s, bce_dice_loss - 0.2247, iou - 0.8976, f-score - 0.9425]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 26\n",
      "train: 100%|█████████████████| 165/165 [01:59<00:00,  1.38it/s, bce_dice_loss - 0.3141, iou - 0.8637, f-score - 0.9263]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.24it/s, bce_dice_loss - 0.2148, iou - 0.9015, f-score - 0.9454]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 27\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.3054, iou - 0.8671, f-score - 0.9284]\n",
      "valid: 100%|███████████████████| 100/100 [00:03<00:00, 25.09it/s, bce_dice_loss - 0.234, iou - 0.896, f-score - 0.9412]\n",
      "\n",
      "Epoch: 28\n",
      "train: 100%|█████████████████| 165/165 [01:59<00:00,  1.38it/s, bce_dice_loss - 0.2946, iou - 0.8713, f-score - 0.9309]\n",
      "valid: 100%|█████████████████| 100/100 [00:04<00:00, 24.84it/s, bce_dice_loss - 0.2379, iou - 0.8955, f-score - 0.9407]\n",
      "\n",
      "Epoch: 29\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.2936, iou - 0.8723, f-score - 0.9313]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.16it/s, bce_dice_loss - 0.2092, iou - 0.9052, f-score - 0.9471]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 30\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.2985, iou - 0.8696, f-score - 0.9298]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.14it/s, bce_dice_loss - 0.2011, iou - 0.9087, f-score - 0.9498]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 31\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.2908, iou - 0.8728, f-score - 0.9316]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.22it/s, bce_dice_loss - 0.2042, iou - 0.9067, f-score - 0.9484]\n",
      "\n",
      "Epoch: 32\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.2919, iou - 0.8724, f-score - 0.9315]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.23it/s, bce_dice_loss - 0.2122, iou - 0.9045, f-score - 0.9469]\n",
      "\n",
      "Epoch: 33\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.2921, iou - 0.8732, f-score - 0.9318]\n",
      "valid: 100%|██████████████████| 100/100 [00:03<00:00, 25.17it/s, bce_dice_loss - 0.2136, iou - 0.904, f-score - 0.9463]\n",
      "\n",
      "Epoch: 34\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.2848, iou - 0.8771, f-score - 0.9341]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.29it/s, bce_dice_loss - 0.2026, iou - 0.9088, f-score - 0.9491]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 35\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.2767, iou - 0.8791, f-score - 0.9353]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.26it/s, bce_dice_loss - 0.2059, iou - 0.9085, f-score - 0.9491]\n",
      "\n",
      "Epoch: 36\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.2783, iou - 0.8785, f-score - 0.9349]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.14it/s, bce_dice_loss - 0.2052, iou - 0.9069, f-score - 0.9484]\n",
      "\n",
      "Epoch: 37\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.2729, iou - 0.8795, f-score - 0.9354]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.19it/s, bce_dice_loss - 0.2022, iou - 0.9071, f-score - 0.9481]\n",
      "\n",
      "Epoch: 38\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.2727, iou - 0.8805, f-score - 0.9361]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.28it/s, bce_dice_loss - 0.1907, iou - 0.9131, f-score - 0.9521]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 39\n",
      "train: 100%|█████████████████| 165/165 [01:59<00:00,  1.38it/s, bce_dice_loss - 0.2663, iou - 0.8837, f-score - 0.9378]\n",
      "valid: 100%|███████████████████| 100/100 [00:03<00:00, 25.19it/s, bce_dice_loss - 0.1897, iou - 0.913, f-score - 0.952]\n",
      "\n",
      "Epoch: 40\n",
      "train: 100%|██████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.265, iou - 0.8837, f-score - 0.9378]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.23it/s, bce_dice_loss - 0.2002, iou - 0.9095, f-score - 0.9495]\n",
      "\n",
      "Epoch: 41\n",
      "train: 100%|███████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.2583, iou - 0.8873, f-score - 0.94]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.18it/s, bce_dice_loss - 0.1828, iou - 0.9155, f-score - 0.9536]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 42\n",
      "train: 100%|█████████████████| 165/165 [01:59<00:00,  1.38it/s, bce_dice_loss - 0.2665, iou - 0.8845, f-score - 0.9383]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.20it/s, bce_dice_loss - 0.1842, iou - 0.9174, f-score - 0.9546]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 43\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.2626, iou - 0.8853, f-score - 0.9387]\n",
      "valid: 100%|██████████████████| 100/100 [00:03<00:00, 25.11it/s, bce_dice_loss - 0.1892, iou - 0.911, f-score - 0.9508]\n",
      "\n",
      "Epoch: 44\n",
      "train: 100%|██████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.264, iou - 0.8848, f-score - 0.9385]\n",
      "valid: 100%|█████████████████| 100/100 [00:04<00:00, 23.53it/s, bce_dice_loss - 0.1809, iou - 0.9179, f-score - 0.9548]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 45\n",
      "train: 100%|█████████████████| 165/165 [02:01<00:00,  1.36it/s, bce_dice_loss - 0.2601, iou - 0.8859, f-score - 0.9391]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.21it/s, bce_dice_loss - 0.1805, iou - 0.9171, f-score - 0.9544]\n",
      "\n",
      "Epoch: 46\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.2637, iou - 0.8843, f-score - 0.9382]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.28it/s, bce_dice_loss - 0.1884, iou - 0.9164, f-score - 0.9538]\n",
      "\n",
      "Epoch: 47\n",
      "train: 100%|██████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.244, iou - 0.8939, f-score - 0.9436]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.18it/s, bce_dice_loss - 0.1947, iou - 0.9144, f-score - 0.9523]\n",
      "\n",
      "Epoch: 48\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.2599, iou - 0.8859, f-score - 0.9392]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.36it/s, bce_dice_loss - 0.1815, iou - 0.9183, f-score - 0.9549]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 49\n",
      "train: 100%|█████████████████| 165/165 [02:00<00:00,  1.37it/s, bce_dice_loss - 0.2459, iou - 0.8924, f-score - 0.9427]\n",
      "valid: 100%|█████████████████| 100/100 [00:03<00:00, 25.21it/s, bce_dice_loss - 0.1802, iou - 0.9147, f-score - 0.9531]\n",
      "Training completed sucessfully.\n",
      "Best model IoU: 0.9183234661817551\n"
     ]
    }
   ],
   "source": [
    "results = trainer.fit(50, 8, 2)\n",
    "logs[\"vgg11\"] = results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = smp.Unet(\"se_resnext50_32x4d\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "optimizer = optim.Adam([\n",
    "    {\"params\": model.encoder.parameters(), \"lr\" : 1e-6},\n",
    "    {\"params\": model.decoder.parameters(), \"lr\" : 1e-4},\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "trainer = Trainer(model, \"models/unet_seresnext50.pth\", \n",
    "                  TRAIN, TRAIN_MSK, \n",
    "                  VAL, VAL_MSK,\n",
    "                  augmentations=aug,\n",
    "                  optimizer=optimizer,\n",
    "                  device=DEVICE\n",
    "                 )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch: 0\n",
      "train: 100%|█████████████████| 165/165 [03:39<00:00,  1.33s/it, bce_dice_loss - 0.5956, iou - 0.7432, f-score - 0.8453]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.59it/s, bce_dice_loss - 0.2446, iou - 0.8908, f-score - 0.9393]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 1\n",
      "train: 100%|███████████████████| 165/165 [03:40<00:00,  1.34s/it, bce_dice_loss - 0.325, iou - 0.858, f-score - 0.9232]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.65it/s, bce_dice_loss - 0.1977, iou - 0.9106, f-score - 0.9506]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 2\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.2823, iou - 0.8751, f-score - 0.9329]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.47it/s, bce_dice_loss - 0.1649, iou - 0.9218, f-score - 0.9575]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 3\n",
      "train: 100%|███████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.2541, iou - 0.8874, f-score - 0.94]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.20it/s, bce_dice_loss - 0.1556, iou - 0.9266, f-score - 0.9601]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 4\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.2501, iou - 0.8891, f-score - 0.9409]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.48it/s, bce_dice_loss - 0.1424, iou - 0.9339, f-score - 0.9642]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 5\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.2326, iou - 0.8977, f-score - 0.9458]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 15.36it/s, bce_dice_loss - 0.1317, iou - 0.9371, f-score - 0.966]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 6\n",
      "train: 100%|█████████████████| 165/165 [03:42<00:00,  1.35s/it, bce_dice_loss - 0.2181, iou - 0.9041, f-score - 0.9494]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 15.22it/s, bce_dice_loss - 0.1339, iou - 0.937, f-score - 0.9659]\n",
      "\n",
      "Epoch: 7\n",
      "train: 100%|███████████████████| 165/165 [03:42<00:00,  1.35s/it, bce_dice_loss - 0.2177, iou - 0.9053, f-score - 0.95]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.66it/s, bce_dice_loss - 0.1315, iou - 0.9381, f-score - 0.9663]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 8\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.35s/it, bce_dice_loss - 0.2058, iou - 0.9093, f-score - 0.9522]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.45it/s, bce_dice_loss - 0.1179, iou - 0.9445, f-score - 0.9705]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 9\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1986, iou - 0.9123, f-score - 0.9538]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 15.19it/s, bce_dice_loss - 0.126, iou - 0.9427, f-score - 0.9693]\n",
      "\n",
      "Epoch: 10\n",
      "train: 100%|█████████████████| 165/165 [03:42<00:00,  1.35s/it, bce_dice_loss - 0.2074, iou - 0.9079, f-score - 0.9515]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.47it/s, bce_dice_loss - 0.1135, iou - 0.9481, f-score - 0.9724]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 11\n",
      "train: 100%|██████████████████| 165/165 [03:42<00:00,  1.35s/it, bce_dice_loss - 0.197, iou - 0.9131, f-score - 0.9543]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.44it/s, bce_dice_loss - 0.1093, iou - 0.9479, f-score - 0.9723]\n",
      "\n",
      "Epoch: 12\n",
      "train: 100%|█████████████████| 165/165 [03:42<00:00,  1.35s/it, bce_dice_loss - 0.1917, iou - 0.9156, f-score - 0.9557]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 15.16it/s, bce_dice_loss - 0.1023, iou - 0.951, f-score - 0.9741]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 13\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1876, iou - 0.9177, f-score - 0.9568]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.23it/s, bce_dice_loss - 0.1038, iou - 0.9516, f-score - 0.9744]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 14\n",
      "train: 100%|███████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1856, iou - 0.918, f-score - 0.957]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 15.37it/s, bce_dice_loss - 0.1111, iou - 0.9473, f-score - 0.972]\n",
      "\n",
      "Epoch: 15\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1777, iou - 0.9217, f-score - 0.9591]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.70it/s, bce_dice_loss - 0.104, iou - 0.9516, f-score - 0.9744]\n",
      "\n",
      "Epoch: 16\n",
      "train: 100%|█████████████████| 165/165 [03:42<00:00,  1.35s/it, bce_dice_loss - 0.1733, iou - 0.9238, f-score - 0.9602]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.24it/s, bce_dice_loss - 0.1012, iou - 0.9525, f-score - 0.9749]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 17\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1678, iou - 0.9263, f-score - 0.9615]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 15.28it/s, bce_dice_loss - 0.1018, iou - 0.952, f-score - 0.9745]\n",
      "\n",
      "Epoch: 18\n",
      "train: 100%|██████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.172, iou - 0.9243, f-score - 0.9605]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.29it/s, bce_dice_loss - 0.1011, iou - 0.9519, f-score - 0.9745]\n",
      "\n",
      "Epoch: 19\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1671, iou - 0.9265, f-score - 0.9616]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.34it/s, bce_dice_loss - 0.1002, iou - 0.9519, f-score - 0.9745]\n",
      "\n",
      "Epoch: 20\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1682, iou - 0.9247, f-score - 0.9607]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.29it/s, bce_dice_loss - 0.1025, iou - 0.9533, f-score - 0.9752]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 21\n",
      "train: 100%|██████████████████| 165/165 [03:42<00:00,  1.35s/it, bce_dice_loss - 0.1672, iou - 0.9253, f-score - 0.961]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.25it/s, bce_dice_loss - 0.09707, iou - 0.9542, f-score - 0.9756]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 22\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1623, iou - 0.9285, f-score - 0.9627]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.43it/s, bce_dice_loss - 0.1077, iou - 0.9519, f-score - 0.9741]\n",
      "\n",
      "Epoch: 23\n",
      "train: 100%|██████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1566, iou - 0.9309, f-score - 0.964]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.46it/s, bce_dice_loss - 0.09528, iou - 0.9559, f-score - 0.9766]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 24\n",
      "train: 100%|██████████████████| 165/165 [03:42<00:00,  1.35s/it, bce_dice_loss - 0.167, iou - 0.9256, f-score - 0.9611]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.30it/s, bce_dice_loss - 0.09361, iou - 0.9553, f-score - 0.9761]\n",
      "\n",
      "Epoch: 25\n",
      "train: 100%|██████████████████| 165/165 [03:42<00:00,  1.35s/it, bce_dice_loss - 0.1598, iou - 0.9289, f-score - 0.963]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.49it/s, bce_dice_loss - 0.09507, iou - 0.9542, f-score - 0.9756]\n",
      "\n",
      "Epoch: 26\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1532, iou - 0.9327, f-score - 0.9651]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.34it/s, bce_dice_loss - 0.09498, iou - 0.9556, f-score - 0.9763]\n",
      "\n",
      "Epoch: 27\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1568, iou - 0.9299, f-score - 0.9635]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 15.37it/s, bce_dice_loss - 0.095, iou - 0.9552, f-score - 0.9761]\n",
      "\n",
      "Epoch: 28\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1475, iou - 0.9341, f-score - 0.9658]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.41it/s, bce_dice_loss - 0.09301, iou - 0.956, f-score - 0.9766]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 29\n",
      "train: 100%|███████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.147, iou - 0.935, f-score - 0.9663]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.25it/s, bce_dice_loss - 0.09255, iou - 0.9565, f-score - 0.9766]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 30\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1423, iou - 0.9367, f-score - 0.9672]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.30it/s, bce_dice_loss - 0.09271, iou - 0.9562, f-score - 0.9767]\n",
      "\n",
      "Epoch: 31\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1471, iou - 0.9339, f-score - 0.9657]\n",
      "valid: 100%|███████████████████| 100/100 [00:06<00:00, 15.63it/s, bce_dice_loss - 0.091, iou - 0.9568, f-score - 0.977]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 32\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1441, iou - 0.9361, f-score - 0.9668]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.46it/s, bce_dice_loss - 0.09426, iou - 0.956, f-score - 0.9764]\n",
      "\n",
      "Epoch: 33\n",
      "train: 100%|██████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1446, iou - 0.9345, f-score - 0.966]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.32it/s, bce_dice_loss - 0.08895, iou - 0.9585, f-score - 0.9779]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 34\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1424, iou - 0.9367, f-score - 0.9671]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.23it/s, bce_dice_loss - 0.08986, iou - 0.9591, f-score - 0.9782]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 35\n",
      "train: 100%|██████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.138, iou - 0.9384, f-score - 0.9681]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.25it/s, bce_dice_loss - 0.09786, iou - 0.9562, f-score - 0.9766]\n",
      "\n",
      "Epoch: 36\n",
      "train: 100%|███████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.144, iou - 0.9364, f-score - 0.967]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.32it/s, bce_dice_loss - 0.08969, iou - 0.9582, f-score - 0.9779]\n",
      "\n",
      "Epoch: 37\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.35s/it, bce_dice_loss - 0.1362, iou - 0.9385, f-score - 0.9681]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.37it/s, bce_dice_loss - 0.08981, iou - 0.9577, f-score - 0.9775]\n",
      "\n",
      "Epoch: 38\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1279, iou - 0.9433, f-score - 0.9707]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.42it/s, bce_dice_loss - 0.08326, iou - 0.9602, f-score - 0.979]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 39\n",
      "train: 100%|██████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1361, iou - 0.939, f-score - 0.9684]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 15.37it/s, bce_dice_loss - 0.0868, iou - 0.9586, f-score - 0.978]\n",
      "\n",
      "Epoch: 40\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1307, iou - 0.9417, f-score - 0.9699]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.45it/s, bce_dice_loss - 0.08429, iou - 0.959, f-score - 0.9782]\n",
      "\n",
      "Epoch: 41\n",
      "train: 100%|██████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1371, iou - 0.939, f-score - 0.9683]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.48it/s, bce_dice_loss - 0.08144, iou - 0.9601, f-score - 0.9789]\n",
      "\n",
      "Epoch: 42\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.35s/it, bce_dice_loss - 0.1351, iou - 0.9396, f-score - 0.9687]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.49it/s, bce_dice_loss - 0.08443, iou - 0.9594, f-score - 0.9785]\n",
      "\n",
      "Epoch: 43\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1342, iou - 0.9396, f-score - 0.9687]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.41it/s, bce_dice_loss - 0.08277, iou - 0.9606, f-score - 0.9792]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 44\n",
      "train: 100%|██████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1267, iou - 0.943, f-score - 0.9705]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.39it/s, bce_dice_loss - 0.07864, iou - 0.9621, f-score - 0.9801]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 45\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.35s/it, bce_dice_loss - 0.1347, iou - 0.9397, f-score - 0.9687]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.30it/s, bce_dice_loss - 0.07802, iou - 0.9619, f-score - 0.9799]\n",
      "\n",
      "Epoch: 46\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1314, iou - 0.9405, f-score - 0.9692]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.39it/s, bce_dice_loss - 0.07931, iou - 0.9614, f-score - 0.9796]\n",
      "\n",
      "Epoch: 47\n",
      "train: 100%|█████████████████| 165/165 [03:42<00:00,  1.35s/it, bce_dice_loss - 0.1304, iou - 0.9411, f-score - 0.9695]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.26it/s, bce_dice_loss - 0.0798, iou - 0.9612, f-score - 0.9795]\n",
      "\n",
      "Epoch: 48\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1224, iou - 0.9451, f-score - 0.9717]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.37it/s, bce_dice_loss - 0.08764, iou - 0.9579, f-score - 0.9776]\n",
      "\n",
      "Epoch: 49\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.1245, iou - 0.9438, f-score - 0.9709]\n",
      "valid: 100%|████████████████| 100/100 [00:06<00:00, 15.13it/s, bce_dice_loss - 0.08751, iou - 0.9585, f-score - 0.9777]\n",
      "Training completed sucessfully.\n",
      "Best model IoU: 0.9620702105760575\n"
     ]
    }
   ],
   "source": [
    "results = trainer.fit(50, 8, 2)\n",
    "logs[\"seresnext\"] = results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = unet.unet_model.UNet(3, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "optimizer = optim.Adam(model.parameters(), lr=1e-4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "trainer = Trainer(model, \"models/unet_pure2.pth\", \n",
    "                  TRAIN, TRAIN_MSK, \n",
    "                  VAL, VAL_MSK,\n",
    "                  augmentations=aug,\n",
    "                  optimizer=optimizer,\n",
    "                  device=DEVICE\n",
    "                 )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch: 0\n",
      "train: 100%|█████████████████| 165/165 [04:03<00:00,  1.48s/it, bce_dice_loss - 0.9737, iou - 0.5692, f-score - 0.7237]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.42it/s, bce_dice_loss - 0.9436, iou - 0.5712, f-score - 0.713]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 1\n",
      "train: 100%|█████████████████| 165/165 [04:03<00:00,  1.47s/it, bce_dice_loss - 0.9266, iou - 0.5694, f-score - 0.7237]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.44it/s, bce_dice_loss - 0.9164, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 2\n",
      "train: 100%|█████████████████| 165/165 [04:03<00:00,  1.47s/it, bce_dice_loss - 0.9097, iou - 0.5676, f-score - 0.7223]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.62it/s, bce_dice_loss - 0.9022, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 3\n",
      "train: 100%|█████████████████| 165/165 [04:02<00:00,  1.47s/it, bce_dice_loss - 0.8928, iou - 0.5719, f-score - 0.7259]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.50it/s, bce_dice_loss - 0.8888, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 4\n",
      "train: 100%|██████████████████| 165/165 [04:02<00:00,  1.47s/it, bce_dice_loss - 0.884, iou - 0.5702, f-score - 0.7245]\n",
      "valid: 100%|███████████████████| 100/100 [00:06<00:00, 14.46it/s, bce_dice_loss - 0.883, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 5\n",
      "train: 100%|█████████████████| 165/165 [04:02<00:00,  1.47s/it, bce_dice_loss - 0.8806, iou - 0.5683, f-score - 0.7227]\n",
      "valid: 100%|████████████████████| 100/100 [00:06<00:00, 14.64it/s, bce_dice_loss - 0.88, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 6\n",
      "train: 100%|█████████████████| 165/165 [04:02<00:00,  1.47s/it, bce_dice_loss - 0.8736, iou - 0.5686, f-score - 0.7233]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.44it/s, bce_dice_loss - 0.8747, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 7\n",
      "train: 100%|██████████████████| 165/165 [04:03<00:00,  1.48s/it, bce_dice_loss - 0.871, iou - 0.5681, f-score - 0.7226]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.45it/s, bce_dice_loss - 0.8639, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 8\n",
      "train: 100%|█████████████████| 165/165 [04:03<00:00,  1.47s/it, bce_dice_loss - 0.8713, iou - 0.5681, f-score - 0.7225]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.47it/s, bce_dice_loss - 0.8718, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 9\n",
      "train: 100%|███████████████████| 165/165 [04:02<00:00,  1.47s/it, bce_dice_loss - 0.86, iou - 0.5698, f-score - 0.7245]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.63it/s, bce_dice_loss - 0.8619, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 10\n",
      "train: 100%|██████████████████| 165/165 [04:02<00:00,  1.47s/it, bce_dice_loss - 0.8603, iou - 0.5696, f-score - 0.724]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.47it/s, bce_dice_loss - 0.8657, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 11\n",
      "train: 100%|█████████████████| 165/165 [04:02<00:00,  1.47s/it, bce_dice_loss - 0.8603, iou - 0.5677, f-score - 0.7227]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.47it/s, bce_dice_loss - 0.8579, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 12\n",
      "train: 100%|█████████████████| 165/165 [04:03<00:00,  1.47s/it, bce_dice_loss - 0.8553, iou - 0.5702, f-score - 0.7244]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.47it/s, bce_dice_loss - 0.8554, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 13\n",
      "train: 100%|█████████████████| 165/165 [04:03<00:00,  1.48s/it, bce_dice_loss - 0.8561, iou - 0.5689, f-score - 0.7235]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.46it/s, bce_dice_loss - 0.8539, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 14\n",
      "train: 100%|█████████████████| 165/165 [04:04<00:00,  1.48s/it, bce_dice_loss - 0.8536, iou - 0.5712, f-score - 0.7252]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.49it/s, bce_dice_loss - 0.8532, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 15\n",
      "train: 100%|██████████████████| 165/165 [04:04<00:00,  1.48s/it, bce_dice_loss - 0.854, iou - 0.5696, f-score - 0.7241]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.46it/s, bce_dice_loss - 0.8503, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 16\n",
      "train: 100%|███████████████████| 165/165 [04:03<00:00,  1.48s/it, bce_dice_loss - 0.8547, iou - 0.567, f-score - 0.722]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.46it/s, bce_dice_loss - 0.8466, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 17\n",
      "train: 100%|█████████████████| 165/165 [04:03<00:00,  1.48s/it, bce_dice_loss - 0.8526, iou - 0.5668, f-score - 0.7217]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.50it/s, bce_dice_loss - 0.8491, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 18\n",
      "train: 100%|██████████████████| 165/165 [04:03<00:00,  1.48s/it, bce_dice_loss - 0.8488, iou - 0.5711, f-score - 0.725]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.47it/s, bce_dice_loss - 0.8497, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 19\n",
      "train: 100%|█████████████████| 165/165 [04:03<00:00,  1.48s/it, bce_dice_loss - 0.8477, iou - 0.5695, f-score - 0.7237]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.45it/s, bce_dice_loss - 0.8489, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 20\n",
      "train: 100%|██████████████████| 165/165 [04:03<00:00,  1.47s/it, bce_dice_loss - 0.845, iou - 0.5722, f-score - 0.7263]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.55it/s, bce_dice_loss - 0.8479, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 21\n",
      "train: 100%|██████████████████| 165/165 [04:03<00:00,  1.47s/it, bce_dice_loss - 0.8452, iou - 0.573, f-score - 0.7265]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.46it/s, bce_dice_loss - 0.8458, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 22\n",
      "train: 100%|██████████████████| 165/165 [04:03<00:00,  1.48s/it, bce_dice_loss - 0.8465, iou - 0.5711, f-score - 0.725]\n",
      "valid: 100%|███████████████████| 100/100 [00:06<00:00, 14.46it/s, bce_dice_loss - 0.843, iou - 0.5712, f-score - 0.713]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 23\n",
      "train: 100%|█████████████████| 165/165 [04:03<00:00,  1.47s/it, bce_dice_loss - 0.8457, iou - 0.5679, f-score - 0.7226]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.47it/s, bce_dice_loss - 0.8468, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 24\n",
      "train: 100%|█████████████████| 165/165 [04:03<00:00,  1.48s/it, bce_dice_loss - 0.8444, iou - 0.5705, f-score - 0.7248]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.45it/s, bce_dice_loss - 0.8443, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 25\n",
      "train: 100%|██████████████████| 165/165 [04:04<00:00,  1.48s/it, bce_dice_loss - 0.8365, iou - 0.5733, f-score - 0.727]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.34it/s, bce_dice_loss - 0.8407, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 26\n",
      "train: 100%|██████████████████| 165/165 [04:04<00:00,  1.48s/it, bce_dice_loss - 0.8417, iou - 0.5697, f-score - 0.724]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.33it/s, bce_dice_loss - 0.8405, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 27\n",
      "train: 100%|█████████████████| 165/165 [04:04<00:00,  1.48s/it, bce_dice_loss - 0.8394, iou - 0.5713, f-score - 0.7254]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.46it/s, bce_dice_loss - 0.8398, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 28\n",
      "train: 100%|██████████████████| 165/165 [04:04<00:00,  1.48s/it, bce_dice_loss - 0.839, iou - 0.5723, f-score - 0.7257]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.37it/s, bce_dice_loss - 0.8414, iou - 0.5713, f-score - 0.713]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 29\n",
      "train: 100%|█████████████████| 165/165 [04:04<00:00,  1.48s/it, bce_dice_loss - 0.8429, iou - 0.5651, f-score - 0.7203]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.38it/s, bce_dice_loss - 0.8383, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 30\n",
      "train: 100%|█████████████████| 165/165 [04:05<00:00,  1.49s/it, bce_dice_loss - 0.8389, iou - 0.5701, f-score - 0.7244]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.31it/s, bce_dice_loss - 0.8393, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 31\n",
      "train: 100%|█████████████████| 165/165 [04:05<00:00,  1.49s/it, bce_dice_loss - 0.8361, iou - 0.5726, f-score - 0.7266]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.39it/s, bce_dice_loss - 0.8378, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 32\n",
      "train: 100%|█████████████████| 165/165 [04:05<00:00,  1.49s/it, bce_dice_loss - 0.8365, iou - 0.5707, f-score - 0.7248]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.39it/s, bce_dice_loss - 0.8389, iou - 0.5713, f-score - 0.713]\n",
      "\n",
      "Epoch: 33\n",
      "train: 100%|█████████████████| 165/165 [04:06<00:00,  1.49s/it, bce_dice_loss - 0.8375, iou - 0.5705, f-score - 0.7248]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.35it/s, bce_dice_loss - 0.8348, iou - 0.5713, f-score - 0.7131]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 34\n",
      "train: 100%|██████████████████| 165/165 [04:07<00:00,  1.50s/it, bce_dice_loss - 0.8346, iou - 0.5708, f-score - 0.725]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.34it/s, bce_dice_loss - 0.837, iou - 0.5713, f-score - 0.7131]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 35\n",
      "train: 100%|█████████████████| 165/165 [04:06<00:00,  1.50s/it, bce_dice_loss - 0.8344, iou - 0.5727, f-score - 0.7265]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.31it/s, bce_dice_loss - 0.8351, iou - 0.5712, f-score - 0.713]\n",
      "\n",
      "Epoch: 36\n",
      "train: 100%|██████████████████| 165/165 [04:06<00:00,  1.50s/it, bce_dice_loss - 0.8379, iou - 0.571, f-score - 0.7248]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.33it/s, bce_dice_loss - 0.8391, iou - 0.5713, f-score - 0.7131]\n",
      "\n",
      "Epoch: 37\n",
      "train: 100%|█████████████████| 165/165 [04:07<00:00,  1.50s/it, bce_dice_loss - 0.8364, iou - 0.5721, f-score - 0.7257]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.29it/s, bce_dice_loss - 0.8379, iou - 0.5713, f-score - 0.713]\n",
      "\n",
      "Epoch: 38\n",
      "train: 100%|█████████████████| 165/165 [04:06<00:00,  1.49s/it, bce_dice_loss - 0.8345, iou - 0.5727, f-score - 0.7265]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.33it/s, bce_dice_loss - 0.8363, iou - 0.5723, f-score - 0.7139]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 39\n",
      "train: 100%|██████████████████| 165/165 [04:05<00:00,  1.49s/it, bce_dice_loss - 0.8327, iou - 0.575, f-score - 0.7281]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.34it/s, bce_dice_loss - 0.837, iou - 0.5718, f-score - 0.7135]\n",
      "\n",
      "Epoch: 40\n",
      "train: 100%|█████████████████| 165/165 [04:06<00:00,  1.49s/it, bce_dice_loss - 0.8339, iou - 0.5752, f-score - 0.7281]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.30it/s, bce_dice_loss - 0.8355, iou - 0.572, f-score - 0.7137]\n",
      "\n",
      "Epoch: 41\n",
      "train: 100%|█████████████████| 165/165 [04:06<00:00,  1.49s/it, bce_dice_loss - 0.8309, iou - 0.5808, f-score - 0.7323]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.39it/s, bce_dice_loss - 0.834, iou - 0.5714, f-score - 0.7132]\n",
      "\n",
      "Epoch: 42\n",
      "train: 100%|█████████████████| 165/165 [04:05<00:00,  1.49s/it, bce_dice_loss - 0.8295, iou - 0.5812, f-score - 0.7327]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.39it/s, bce_dice_loss - 0.839, iou - 0.5713, f-score - 0.7131]\n",
      "\n",
      "Epoch: 43\n",
      "train: 100%|██████████████████| 165/165 [04:05<00:00,  1.49s/it, bce_dice_loss - 0.8324, iou - 0.5835, f-score - 0.734]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.33it/s, bce_dice_loss - 0.8321, iou - 0.581, f-score - 0.7211]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 44\n",
      "train: 100%|███████████████████| 165/165 [04:05<00:00,  1.49s/it, bce_dice_loss - 0.8346, iou - 0.575, f-score - 0.728]\n",
      "valid: 100%|██████████████████| 100/100 [00:07<00:00, 14.04it/s, bce_dice_loss - 0.8328, iou - 0.6271, f-score - 0.757]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 45\n",
      "train: 100%|██████████████████| 165/165 [04:07<00:00,  1.50s/it, bce_dice_loss - 0.832, iou - 0.5827, f-score - 0.7337]\n",
      "valid: 100%|█████████████████| 100/100 [00:07<00:00, 14.26it/s, bce_dice_loss - 0.8336, iou - 0.5714, f-score - 0.7132]\n",
      "\n",
      "Epoch: 46\n",
      "train: 100%|██████████████████| 165/165 [04:06<00:00,  1.50s/it, bce_dice_loss - 0.8327, iou - 0.5834, f-score - 0.734]\n",
      "valid: 100%|█████████████████| 100/100 [00:07<00:00, 14.26it/s, bce_dice_loss - 0.8377, iou - 0.5728, f-score - 0.7143]\n",
      "\n",
      "Epoch: 47\n",
      "train: 100%|██████████████████| 165/165 [04:06<00:00,  1.49s/it, bce_dice_loss - 0.8308, iou - 0.584, f-score - 0.7346]\n",
      "valid: 100%|█████████████████| 100/100 [00:07<00:00, 14.11it/s, bce_dice_loss - 0.8316, iou - 0.5964, f-score - 0.7329]\n",
      "\n",
      "Epoch: 48\n",
      "train: 100%|█████████████████| 165/165 [04:05<00:00,  1.49s/it, bce_dice_loss - 0.8288, iou - 0.5927, f-score - 0.7403]\n",
      "valid: 100%|█████████████████| 100/100 [00:07<00:00, 14.27it/s, bce_dice_loss - 0.8321, iou - 0.5834, f-score - 0.7228]\n",
      "\n",
      "Epoch: 49\n",
      "train: 100%|█████████████████| 165/165 [04:06<00:00,  1.49s/it, bce_dice_loss - 0.8358, iou - 0.5755, f-score - 0.7283]\n",
      "valid: 100%|█████████████████| 100/100 [00:07<00:00, 14.13it/s, bce_dice_loss - 0.8353, iou - 0.5718, f-score - 0.7135]\n",
      "Training completed sucessfully.\n",
      "Best model IoU: 0.6270632593333723\n"
     ]
    }
   ],
   "source": [
    "results = trainer.fit(50, 8, 2)\n",
    "logs[\"unet\"] = results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = SimpleFCN(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "optimizer = optim.Adam(model.parameters(), lr=1e-4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "trainer = Trainer(model, \"models/fcn.pth\", \n",
    "                  TRAIN, TRAIN_MSK, \n",
    "                  VAL, VAL_MSK,\n",
    "                  augmentations=aug,\n",
    "                  optimizer=optimizer,\n",
    "                  device=DEVICE\n",
    "                 )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch: 0\n",
      "train: 100%|██████████████████| 165/165 [03:37<00:00,  1.32s/it, bce_dice_loss - 1.073, iou - 0.5382, f-score - 0.6605]\n",
      "valid: 100%|███████████████████| 100/100 [00:05<00:00, 18.49it/s, bce_dice_loss - 1.039, iou - 0.6394, f-score - 0.765]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 1\n",
      "train: 100%|█████████████████| 165/165 [03:34<00:00,  1.30s/it, bce_dice_loss - 0.8653, iou - 0.6738, f-score - 0.8039]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 18.42it/s, bce_dice_loss - 0.8192, iou - 0.6735, f-score - 0.7929]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 2\n",
      "train: 100%|██████████████████| 165/165 [03:34<00:00,  1.30s/it, bce_dice_loss - 0.7683, iou - 0.687, f-score - 0.8135]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 18.51it/s, bce_dice_loss - 0.7703, iou - 0.6867, f-score - 0.802]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 3\n",
      "train: 100%|█████████████████| 165/165 [03:34<00:00,  1.30s/it, bce_dice_loss - 0.7299, iou - 0.6961, f-score - 0.8198]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 18.38it/s, bce_dice_loss - 0.743, iou - 0.6956, f-score - 0.8083]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 4\n",
      "train: 100%|█████████████████| 165/165 [03:34<00:00,  1.30s/it, bce_dice_loss - 0.6975, iou - 0.7057, f-score - 0.8264]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 18.43it/s, bce_dice_loss - 0.7312, iou - 0.706, f-score - 0.8151]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 5\n",
      "train: 100%|█████████████████| 165/165 [03:39<00:00,  1.33s/it, bce_dice_loss - 0.6833, iou - 0.7105, f-score - 0.8296]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.83it/s, bce_dice_loss - 0.6924, iou - 0.7056, f-score - 0.8166]\n",
      "\n",
      "Epoch: 6\n",
      "train: 100%|█████████████████| 165/165 [03:43<00:00,  1.35s/it, bce_dice_loss - 0.6618, iou - 0.7199, f-score - 0.8361]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.55it/s, bce_dice_loss - 0.6698, iou - 0.7236, f-score - 0.8278]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 7\n",
      "train: 100%|█████████████████| 165/165 [03:42<00:00,  1.35s/it, bce_dice_loss - 0.6562, iou - 0.7209, f-score - 0.8368]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.98it/s, bce_dice_loss - 0.6484, iou - 0.7303, f-score - 0.8337]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 8\n",
      "train: 100%|██████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.6379, iou - 0.7268, f-score - 0.841]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.75it/s, bce_dice_loss - 0.6761, iou - 0.7183, f-score - 0.8248]\n",
      "\n",
      "Epoch: 9\n",
      "train: 100%|█████████████████| 165/165 [03:40<00:00,  1.33s/it, bce_dice_loss - 0.6283, iou - 0.7322, f-score - 0.8444]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 18.04it/s, bce_dice_loss - 0.6514, iou - 0.7367, f-score - 0.8366]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 10\n",
      "train: 100%|█████████████████| 165/165 [03:39<00:00,  1.33s/it, bce_dice_loss - 0.6225, iou - 0.7358, f-score - 0.8468]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.97it/s, bce_dice_loss - 0.6149, iou - 0.7298, f-score - 0.8338]\n",
      "\n",
      "Epoch: 11\n",
      "train: 100%|██████████████████| 165/165 [03:40<00:00,  1.34s/it, bce_dice_loss - 0.6166, iou - 0.738, f-score - 0.8483]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 17.91it/s, bce_dice_loss - 0.6086, iou - 0.7423, f-score - 0.841]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 12\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.6062, iou - 0.7436, f-score - 0.8519]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 17.94it/s, bce_dice_loss - 0.5731, iou - 0.754, f-score - 0.8505]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 13\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.5953, iou - 0.7461, f-score - 0.8538]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 18.25it/s, bce_dice_loss - 0.5893, iou - 0.7471, f-score - 0.844]\n",
      "\n",
      "Epoch: 14\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.5997, iou - 0.7455, f-score - 0.8533]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.94it/s, bce_dice_loss - 0.5981, iou - 0.7491, f-score - 0.8461]\n",
      "\n",
      "Epoch: 15\n",
      "train: 100%|█████████████████| 165/165 [03:45<00:00,  1.37s/it, bce_dice_loss - 0.5926, iou - 0.7473, f-score - 0.8546]\n",
      "valid: 100%|████████████████████| 100/100 [00:05<00:00, 17.16it/s, bce_dice_loss - 0.5817, iou - 0.754, f-score - 0.85]\n",
      "\n",
      "Epoch: 16\n",
      "train: 100%|██████████████████| 165/165 [03:45<00:00,  1.37s/it, bce_dice_loss - 0.582, iou - 0.7516, f-score - 0.8573]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.28it/s, bce_dice_loss - 0.5525, iou - 0.7634, f-score - 0.8558]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 17\n",
      "train: 100%|██████████████████| 165/165 [03:42<00:00,  1.35s/it, bce_dice_loss - 0.5869, iou - 0.7497, f-score - 0.856]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.92it/s, bce_dice_loss - 0.5605, iou - 0.7578, f-score - 0.8521]\n",
      "\n",
      "Epoch: 18\n",
      "train: 100%|██████████████████| 165/165 [03:43<00:00,  1.35s/it, bce_dice_loss - 0.5826, iou - 0.7511, f-score - 0.857]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 18.08it/s, bce_dice_loss - 0.5484, iou - 0.764, f-score - 0.8576]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 19\n",
      "train: 100%|█████████████████| 165/165 [03:40<00:00,  1.34s/it, bce_dice_loss - 0.5863, iou - 0.7506, f-score - 0.8567]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.94it/s, bce_dice_loss - 0.5454, iou - 0.7656, f-score - 0.8569]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 20\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.5628, iou - 0.7593, f-score - 0.8624]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.97it/s, bce_dice_loss - 0.5595, iou - 0.7617, f-score - 0.8527]\n",
      "\n",
      "Epoch: 21\n",
      "train: 100%|██████████████████| 165/165 [03:39<00:00,  1.33s/it, bce_dice_loss - 0.5751, iou - 0.755, f-score - 0.8596]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.02it/s, bce_dice_loss - 0.5469, iou - 0.7671, f-score - 0.8584]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 22\n",
      "train: 100%|█████████████████| 165/165 [03:42<00:00,  1.35s/it, bce_dice_loss - 0.5565, iou - 0.7612, f-score - 0.8635]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.96it/s, bce_dice_loss - 0.5402, iou - 0.7663, f-score - 0.8571]\n",
      "\n",
      "Epoch: 23\n",
      "train: 100%|█████████████████| 165/165 [03:43<00:00,  1.35s/it, bce_dice_loss - 0.5678, iou - 0.7563, f-score - 0.8602]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.57it/s, bce_dice_loss - 0.5398, iou - 0.7684, f-score - 0.8589]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 24\n",
      "train: 100%|█████████████████| 165/165 [03:46<00:00,  1.37s/it, bce_dice_loss - 0.5618, iou - 0.7595, f-score - 0.8623]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.21it/s, bce_dice_loss - 0.5367, iou - 0.7667, f-score - 0.8576]\n",
      "\n",
      "Epoch: 25\n",
      "train: 100%|█████████████████| 165/165 [03:50<00:00,  1.39s/it, bce_dice_loss - 0.5596, iou - 0.7621, f-score - 0.8641]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 17.20it/s, bce_dice_loss - 0.5238, iou - 0.7754, f-score - 0.864]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 26\n",
      "train: 100%|██████████████████| 165/165 [03:49<00:00,  1.39s/it, bce_dice_loss - 0.549, iou - 0.7641, f-score - 0.8654]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.78it/s, bce_dice_loss - 0.5139, iou - 0.7757, f-score - 0.8637]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 27\n",
      "train: 100%|█████████████████| 165/165 [03:45<00:00,  1.37s/it, bce_dice_loss - 0.5449, iou - 0.7669, f-score - 0.8671]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.20it/s, bce_dice_loss - 0.5157, iou - 0.7772, f-score - 0.8644]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 28\n",
      "train: 100%|█████████████████| 165/165 [03:49<00:00,  1.39s/it, bce_dice_loss - 0.5468, iou - 0.7654, f-score - 0.8664]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 16.88it/s, bce_dice_loss - 0.5155, iou - 0.779, f-score - 0.8658]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 29\n",
      "train: 100%|█████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.5489, iou - 0.7661, f-score - 0.8666]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.96it/s, bce_dice_loss - 0.5603, iou - 0.7656, f-score - 0.8564]\n",
      "\n",
      "Epoch: 30\n",
      "train: 100%|██████████████████| 165/165 [03:39<00:00,  1.33s/it, bce_dice_loss - 0.5484, iou - 0.764, f-score - 0.8654]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 17.79it/s, bce_dice_loss - 0.5175, iou - 0.774, f-score - 0.8622]\n",
      "\n",
      "Epoch: 31\n",
      "train: 100%|██████████████████| 165/165 [03:38<00:00,  1.32s/it, bce_dice_loss - 0.543, iou - 0.7682, f-score - 0.8679]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 18.46it/s, bce_dice_loss - 0.5155, iou - 0.7736, f-score - 0.8613]\n",
      "\n",
      "Epoch: 32\n",
      "train: 100%|██████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.5469, iou - 0.7668, f-score - 0.867]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.53it/s, bce_dice_loss - 0.5087, iou - 0.7804, f-score - 0.8674]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 33\n",
      "train: 100%|███████████████████| 165/165 [03:41<00:00,  1.34s/it, bce_dice_loss - 0.5403, iou - 0.77, f-score - 0.8691]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.90it/s, bce_dice_loss - 0.5138, iou - 0.7814, f-score - 0.8677]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 34\n",
      "train: 100%|█████████████████| 165/165 [03:38<00:00,  1.33s/it, bce_dice_loss - 0.5376, iou - 0.7713, f-score - 0.8698]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.45it/s, bce_dice_loss - 0.5058, iou - 0.7753, f-score - 0.8636]\n",
      "\n",
      "Epoch: 35\n",
      "train: 100%|█████████████████| 165/165 [03:46<00:00,  1.37s/it, bce_dice_loss - 0.5376, iou - 0.7704, f-score - 0.8692]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 17.45it/s, bce_dice_loss - 0.5098, iou - 0.7805, f-score - 0.867]\n",
      "\n",
      "Epoch: 36\n",
      "train: 100%|█████████████████| 165/165 [03:45<00:00,  1.37s/it, bce_dice_loss - 0.5305, iou - 0.7724, f-score - 0.8708]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 17.67it/s, bce_dice_loss - 0.5071, iou - 0.7743, f-score - 0.862]\n",
      "\n",
      "Epoch: 37\n",
      "train: 100%|█████████████████| 165/165 [03:43<00:00,  1.36s/it, bce_dice_loss - 0.5317, iou - 0.7722, f-score - 0.8706]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 17.61it/s, bce_dice_loss - 0.517, iou - 0.7766, f-score - 0.8638]\n",
      "\n",
      "Epoch: 38\n",
      "train: 100%|█████████████████| 165/165 [03:44<00:00,  1.36s/it, bce_dice_loss - 0.5214, iou - 0.7765, f-score - 0.8734]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.97it/s, bce_dice_loss - 0.4878, iou - 0.7874, f-score - 0.8719]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 39\n",
      "train: 100%|█████████████████| 165/165 [03:45<00:00,  1.37s/it, bce_dice_loss - 0.5195, iou - 0.7777, f-score - 0.8741]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.68it/s, bce_dice_loss - 0.5003, iou - 0.7816, f-score - 0.8664]\n",
      "\n",
      "Epoch: 40\n",
      "train: 100%|█████████████████| 165/165 [04:01<00:00,  1.46s/it, bce_dice_loss - 0.5266, iou - 0.7738, f-score - 0.8717]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 16.05it/s, bce_dice_loss - 0.494, iou - 0.7874, f-score - 0.8722]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 41\n",
      "train: 100%|█████████████████| 165/165 [03:59<00:00,  1.45s/it, bce_dice_loss - 0.5198, iou - 0.7772, f-score - 0.8736]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.03it/s, bce_dice_loss - 0.4957, iou - 0.7789, f-score - 0.8644]\n",
      "\n",
      "Epoch: 42\n",
      "train: 100%|█████████████████| 165/165 [03:59<00:00,  1.45s/it, bce_dice_loss - 0.5277, iou - 0.7742, f-score - 0.8719]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.50it/s, bce_dice_loss - 0.5478, iou - 0.7704, f-score - 0.8586]\n",
      "\n",
      "Epoch: 43\n",
      "train: 100%|█████████████████| 165/165 [03:56<00:00,  1.43s/it, bce_dice_loss - 0.5305, iou - 0.7716, f-score - 0.8701]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.61it/s, bce_dice_loss - 0.4993, iou - 0.7785, f-score - 0.8658]\n",
      "\n",
      "Epoch: 44\n",
      "train: 100%|█████████████████| 165/165 [03:50<00:00,  1.40s/it, bce_dice_loss - 0.5196, iou - 0.7763, f-score - 0.8732]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.23it/s, bce_dice_loss - 0.4919, iou - 0.7878, f-score - 0.8717]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 45\n",
      "train: 100%|███████████████████| 165/165 [03:48<00:00,  1.38s/it, bce_dice_loss - 0.507, iou - 0.7836, f-score - 0.878]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.16it/s, bce_dice_loss - 0.4922, iou - 0.7871, f-score - 0.8713]\n",
      "\n",
      "Epoch: 46\n",
      "train: 100%|█████████████████| 165/165 [03:47<00:00,  1.38s/it, bce_dice_loss - 0.5165, iou - 0.7788, f-score - 0.8747]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 16.71it/s, bce_dice_loss - 0.4761, iou - 0.7914, f-score - 0.8749]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 47\n",
      "train: 100%|█████████████████| 165/165 [03:46<00:00,  1.37s/it, bce_dice_loss - 0.5175, iou - 0.7784, f-score - 0.8745]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.00it/s, bce_dice_loss - 0.4844, iou - 0.7868, f-score - 0.8719]\n",
      "\n",
      "Epoch: 48\n",
      "train: 100%|█████████████████| 165/165 [03:46<00:00,  1.37s/it, bce_dice_loss - 0.5185, iou - 0.7767, f-score - 0.8736]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.58it/s, bce_dice_loss - 0.4845, iou - 0.7914, f-score - 0.8753]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 49\n",
      "train: 100%|██████████████████| 165/165 [03:47<00:00,  1.38s/it, bce_dice_loss - 0.501, iou - 0.7851, f-score - 0.8789]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.20it/s, bce_dice_loss - 0.4697, iou - 0.7897, f-score - 0.8738]\n",
      "Training completed sucessfully.\n",
      "Best model IoU: 0.7914387258887288\n"
     ]
    }
   ],
   "source": [
    "results = trainer.fit(50, 8, 2)\n",
    "logs[\"fcn\"] = results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = smp.PSPNet(\"se_resnext50_32x4d\", classes=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "optimizer = optim.Adam([\n",
    "    {\"params\": model.encoder.parameters(), \"lr\" : 1e-6},\n",
    "    {\"params\": model.decoder.parameters(), \"lr\" : 1e-4},\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "trainer = Trainer(model, \"models/psp_seresnext50.pth\", \n",
    "                  TRAIN, TRAIN_MSK, \n",
    "                  VAL, VAL_MSK,\n",
    "                  augmentations=aug,\n",
    "                  optimizer=optimizer,\n",
    "                  device=DEVICE\n",
    "                 )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch: 0\n",
      "train: 100%|████████████████████| 165/165 [02:17<00:00,  1.20it/s, bce_dice_loss - 3.597, iou - 0.627, f-score - 0.765]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 17.34it/s, bce_dice_loss - 1.805, iou - 0.7484, f-score - 0.8435]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 1\n",
      "train: 100%|██████████████████| 165/165 [02:19<00:00,  1.19it/s, bce_dice_loss - 2.453, iou - 0.7146, f-score - 0.8326]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 17.10it/s, bce_dice_loss - 1.397, iou - 0.7843, f-score - 0.8699]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 2\n",
      "train: 100%|██████████████████| 165/165 [02:19<00:00,  1.18it/s, bce_dice_loss - 2.175, iou - 0.7353, f-score - 0.8463]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 16.96it/s, bce_dice_loss - 1.236, iou - 0.8136, f-score - 0.8886]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 3\n",
      "train: 100%|███████████████████| 165/165 [02:19<00:00,  1.18it/s, bce_dice_loss - 1.95, iou - 0.7513, f-score - 0.8572]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 17.41it/s, bce_dice_loss - 1.199, iou - 0.8128, f-score - 0.8876]\n",
      "\n",
      "Epoch: 4\n",
      "train: 100%|██████████████████| 165/165 [02:20<00:00,  1.18it/s, bce_dice_loss - 1.868, iou - 0.7608, f-score - 0.8633]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 17.03it/s, bce_dice_loss - 1.073, iou - 0.8198, f-score - 0.8925]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 5\n",
      "train: 100%|██████████████████| 165/165 [02:19<00:00,  1.18it/s, bce_dice_loss - 1.841, iou - 0.7629, f-score - 0.8646]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.03it/s, bce_dice_loss - 0.9826, iou - 0.8327, f-score - 0.9013]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 6\n",
      "train: 100%|██████████████████| 165/165 [02:19<00:00,  1.18it/s, bce_dice_loss - 1.641, iou - 0.7774, f-score - 0.8738]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.19it/s, bce_dice_loss - 0.9914, iou - 0.8458, f-score - 0.9096]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 7\n",
      "train: 100%|██████████████████| 165/165 [02:19<00:00,  1.18it/s, bce_dice_loss - 1.643, iou - 0.7754, f-score - 0.8727]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.36it/s, bce_dice_loss - 0.9174, iou - 0.8422, f-score - 0.9078]\n",
      "\n",
      "Epoch: 8\n",
      "train: 100%|██████████████████| 165/165 [02:19<00:00,  1.18it/s, bce_dice_loss - 1.535, iou - 0.7854, f-score - 0.8791]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 16.94it/s, bce_dice_loss - 0.9157, iou - 0.8447, f-score - 0.9089]\n",
      "\n",
      "Epoch: 9\n",
      "train: 100%|███████████████████| 165/165 [02:19<00:00,  1.18it/s, bce_dice_loss - 1.505, iou - 0.789, f-score - 0.8813]\n",
      "valid: 100%|███████████████████| 100/100 [00:05<00:00, 17.21it/s, bce_dice_loss - 0.89, iou - 0.8521, f-score - 0.9137]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 10\n",
      "train: 100%|██████████████████| 165/165 [02:19<00:00,  1.18it/s, bce_dice_loss - 1.545, iou - 0.7814, f-score - 0.8766]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.00it/s, bce_dice_loss - 0.8613, iou - 0.8457, f-score - 0.9093]\n",
      "\n",
      "Epoch: 11\n",
      "train: 100%|██████████████████| 165/165 [02:20<00:00,  1.17it/s, bce_dice_loss - 1.394, iou - 0.7931, f-score - 0.8839]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 16.70it/s, bce_dice_loss - 0.7854, iou - 0.8529, f-score - 0.9134]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 12\n",
      "train: 100%|██████████████████| 165/165 [02:20<00:00,  1.17it/s, bce_dice_loss - 1.401, iou - 0.7913, f-score - 0.8827]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 16.78it/s, bce_dice_loss - 0.7975, iou - 0.8575, f-score - 0.917]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 13\n",
      "train: 100%|██████████████████| 165/165 [02:20<00:00,  1.17it/s, bce_dice_loss - 1.327, iou - 0.7989, f-score - 0.8875]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 16.75it/s, bce_dice_loss - 0.8086, iou - 0.8504, f-score - 0.9121]\n",
      "\n",
      "Epoch: 14\n",
      "train: 100%|██████████████████| 165/165 [02:20<00:00,  1.17it/s, bce_dice_loss - 1.321, iou - 0.7991, f-score - 0.8877]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.32it/s, bce_dice_loss - 0.7278, iou - 0.8645, f-score - 0.9217]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 15\n",
      "train: 100%|██████████████████| 165/165 [02:20<00:00,  1.17it/s, bce_dice_loss - 1.243, iou - 0.8047, f-score - 0.8911]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 16.66it/s, bce_dice_loss - 0.816, iou - 0.8493, f-score - 0.9112]\n",
      "\n",
      "Epoch: 16\n",
      "train: 100%|██████████████████| 165/165 [02:20<00:00,  1.17it/s, bce_dice_loss - 1.261, iou - 0.7984, f-score - 0.8871]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.59it/s, bce_dice_loss - 0.7232, iou - 0.8591, f-score - 0.9182]\n",
      "\n",
      "Epoch: 17\n",
      "train: 100%|██████████████████| 165/165 [02:20<00:00,  1.17it/s, bce_dice_loss - 1.173, iou - 0.8094, f-score - 0.8941]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.42it/s, bce_dice_loss - 0.7382, iou - 0.8679, f-score - 0.9236]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 18\n",
      "train: 100%|██████████████████| 165/165 [02:20<00:00,  1.17it/s, bce_dice_loss - 1.142, iou - 0.8107, f-score - 0.8949]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.38it/s, bce_dice_loss - 0.6909, iou - 0.8687, f-score - 0.9242]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 19\n",
      "train: 100%|███████████████████| 165/165 [02:22<00:00,  1.16it/s, bce_dice_loss - 1.156, iou - 0.806, f-score - 0.8919]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.29it/s, bce_dice_loss - 0.6367, iou - 0.8737, f-score - 0.9283]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 20\n",
      "train: 100%|██████████████████| 165/165 [02:21<00:00,  1.16it/s, bce_dice_loss - 1.118, iou - 0.8091, f-score - 0.8939]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.28it/s, bce_dice_loss - 0.6168, iou - 0.8724, f-score - 0.9268]\n",
      "\n",
      "Epoch: 21\n",
      "train: 100%|██████████████████| 165/165 [02:20<00:00,  1.17it/s, bce_dice_loss - 1.132, iou - 0.8097, f-score - 0.8943]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 16.90it/s, bce_dice_loss - 0.6455, iou - 0.876, f-score - 0.9286]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 22\n",
      "train: 100%|██████████████████| 165/165 [02:20<00:00,  1.17it/s, bce_dice_loss - 1.079, iou - 0.8136, f-score - 0.8966]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 16.80it/s, bce_dice_loss - 0.6072, iou - 0.8744, f-score - 0.9278]\n",
      "\n",
      "Epoch: 23\n",
      "train: 100%|███████████████████| 165/165 [02:20<00:00,  1.18it/s, bce_dice_loss - 1.09, iou - 0.8108, f-score - 0.8948]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.00it/s, bce_dice_loss - 0.6278, iou - 0.8739, f-score - 0.9271]\n",
      "\n",
      "Epoch: 24\n",
      "train: 100%|██████████████████| 165/165 [02:20<00:00,  1.18it/s, bce_dice_loss - 1.011, iou - 0.8187, f-score - 0.8996]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 16.90it/s, bce_dice_loss - 0.6079, iou - 0.8772, f-score - 0.9291]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 25\n",
      "train: 100%|█████████████████| 165/165 [02:19<00:00,  1.19it/s, bce_dice_loss - 0.9885, iou - 0.8207, f-score - 0.9009]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.10it/s, bce_dice_loss - 0.5966, iou - 0.8797, f-score - 0.9311]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 26\n",
      "train: 100%|████████████████████| 165/165 [02:19<00:00,  1.18it/s, bce_dice_loss - 0.968, iou - 0.82, f-score - 0.9004]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 16.74it/s, bce_dice_loss - 0.5393, iou - 0.8802, f-score - 0.9313]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 27\n",
      "train: 100%|████████████████████| 165/165 [02:19<00:00,  1.18it/s, bce_dice_loss - 0.973, iou - 0.82, f-score - 0.9006]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 16.74it/s, bce_dice_loss - 0.6023, iou - 0.8797, f-score - 0.9311]\n",
      "\n",
      "Epoch: 28\n",
      "train: 100%|██████████████████| 165/165 [02:19<00:00,  1.18it/s, bce_dice_loss - 0.952, iou - 0.8186, f-score - 0.8997]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.15it/s, bce_dice_loss - 0.5373, iou - 0.8846, f-score - 0.9341]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 29\n",
      "train: 100%|██████████████████| 165/165 [02:19<00:00,  1.19it/s, bce_dice_loss - 0.879, iou - 0.8259, f-score - 0.9041]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.04it/s, bce_dice_loss - 0.5428, iou - 0.8854, f-score - 0.9347]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 30\n",
      "train: 100%|██████████████████| 165/165 [02:19<00:00,  1.18it/s, bce_dice_loss - 0.9036, iou - 0.8226, f-score - 0.902]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 16.95it/s, bce_dice_loss - 0.5509, iou - 0.8813, f-score - 0.9317]\n",
      "\n",
      "Epoch: 31\n",
      "train: 100%|█████████████████| 165/165 [02:19<00:00,  1.19it/s, bce_dice_loss - 0.8795, iou - 0.8248, f-score - 0.9034]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.50it/s, bce_dice_loss - 0.5001, iou - 0.8829, f-score - 0.9331]\n",
      "\n",
      "Epoch: 32\n",
      "train: 100%|█████████████████| 165/165 [02:17<00:00,  1.20it/s, bce_dice_loss - 0.8673, iou - 0.8231, f-score - 0.9023]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.50it/s, bce_dice_loss - 0.5145, iou - 0.8841, f-score - 0.9336]\n",
      "\n",
      "Epoch: 33\n",
      "train: 100%|████████████████████| 165/165 [02:17<00:00,  1.20it/s, bce_dice_loss - 0.8596, iou - 0.8193, f-score - 0.9]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.55it/s, bce_dice_loss - 0.5066, iou - 0.8801, f-score - 0.9312]\n",
      "\n",
      "Epoch: 34\n",
      "train: 100%|█████████████████| 165/165 [02:17<00:00,  1.20it/s, bce_dice_loss - 0.8118, iou - 0.8282, f-score - 0.9055]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 17.51it/s, bce_dice_loss - 0.4991, iou - 0.886, f-score - 0.9353]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 35\n",
      "train: 100%|█████████████████| 165/165 [02:17<00:00,  1.20it/s, bce_dice_loss - 0.7828, iou - 0.8302, f-score - 0.9068]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.45it/s, bce_dice_loss - 0.5368, iou - 0.8837, f-score - 0.9338]\n",
      "\n",
      "Epoch: 36\n",
      "train: 100%|█████████████████| 165/165 [02:17<00:00,  1.20it/s, bce_dice_loss - 0.7446, iou - 0.8316, f-score - 0.9076]\n",
      "valid: 100%|██████████████████| 100/100 [00:05<00:00, 17.47it/s, bce_dice_loss - 0.4992, iou - 0.885, f-score - 0.9343]\n",
      "\n",
      "Epoch: 37\n",
      "train: 100%|█████████████████| 165/165 [02:19<00:00,  1.19it/s, bce_dice_loss - 0.7401, iou - 0.8342, f-score - 0.9091]\n",
      "valid: 100%|█████████████████| 100/100 [00:07<00:00, 12.66it/s, bce_dice_loss - 0.4572, iou - 0.8919, f-score - 0.9388]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 38\n",
      "train: 100%|█████████████████| 165/165 [02:51<00:00,  1.04s/it, bce_dice_loss - 0.7529, iou - 0.8314, f-score - 0.9073]\n",
      "valid: 100%|███████████████████| 100/100 [00:06<00:00, 15.64it/s, bce_dice_loss - 0.51, iou - 0.8873, f-score - 0.9357]\n",
      "\n",
      "Epoch: 39\n",
      "train: 100%|█████████████████| 165/165 [02:18<00:00,  1.19it/s, bce_dice_loss - 0.7101, iou - 0.8342, f-score - 0.9091]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.69it/s, bce_dice_loss - 0.4314, iou - 0.8905, f-score - 0.9384]\n",
      "\n",
      "Epoch: 40\n",
      "train: 100%|█████████████████| 165/165 [02:20<00:00,  1.17it/s, bce_dice_loss - 0.7157, iou - 0.8325, f-score - 0.9082]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.70it/s, bce_dice_loss - 0.4345, iou - 0.8903, f-score - 0.9376]\n",
      "\n",
      "Epoch: 41\n",
      "train: 100%|██████████████████| 165/165 [02:20<00:00,  1.18it/s, bce_dice_loss - 0.6707, iou - 0.837, f-score - 0.9108]\n",
      "valid: 100%|█████████████████| 100/100 [00:05<00:00, 17.54it/s, bce_dice_loss - 0.4418, iou - 0.8885, f-score - 0.9362]\n",
      "\n",
      "Epoch: 42\n",
      "train: 100%|█████████████████| 165/165 [02:18<00:00,  1.19it/s, bce_dice_loss - 0.6976, iou - 0.8297, f-score - 0.9064]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 15.68it/s, bce_dice_loss - 0.4846, iou - 0.8874, f-score - 0.936]\n",
      "\n",
      "Epoch: 43\n",
      "train: 100%|███████████████████| 165/165 [02:19<00:00,  1.18it/s, bce_dice_loss - 0.6716, iou - 0.834, f-score - 0.909]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.71it/s, bce_dice_loss - 0.3996, iou - 0.8898, f-score - 0.9375]\n",
      "\n",
      "Epoch: 44\n",
      "train: 100%|█████████████████| 165/165 [02:19<00:00,  1.19it/s, bce_dice_loss - 0.6419, iou - 0.8378, f-score - 0.9112]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.50it/s, bce_dice_loss - 0.4019, iou - 0.8912, f-score - 0.9385]\n",
      "\n",
      "Epoch: 45\n",
      "train: 100%|█████████████████| 165/165 [02:18<00:00,  1.19it/s, bce_dice_loss - 0.6184, iou - 0.8382, f-score - 0.9115]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.46it/s, bce_dice_loss - 0.4069, iou - 0.8943, f-score - 0.9403]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 46\n",
      "train: 100%|█████████████████| 165/165 [02:18<00:00,  1.19it/s, bce_dice_loss - 0.6213, iou - 0.8354, f-score - 0.9097]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.85it/s, bce_dice_loss - 0.3763, iou - 0.8949, f-score - 0.9406]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 47\n",
      "train: 100%|█████████████████| 165/165 [02:18<00:00,  1.19it/s, bce_dice_loss - 0.6344, iou - 0.8349, f-score - 0.9094]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.76it/s, bce_dice_loss - 0.3959, iou - 0.8929, f-score - 0.9395]\n",
      "\n",
      "Epoch: 48\n",
      "train: 100%|█████████████████| 165/165 [02:18<00:00,  1.20it/s, bce_dice_loss - 0.5842, iou - 0.8441, f-score - 0.9149]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.11it/s, bce_dice_loss - 0.3995, iou - 0.8932, f-score - 0.9396]\n",
      "\n",
      "Epoch: 49\n",
      "train: 100%|█████████████████| 165/165 [02:18<00:00,  1.19it/s, bce_dice_loss - 0.5648, iou - 0.8429, f-score - 0.9144]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.76it/s, bce_dice_loss - 0.3772, iou - 0.8939, f-score - 0.9401]\n",
      "Training completed sucessfully.\n",
      "Best model IoU: 0.8949280664324757\n"
     ]
    }
   ],
   "source": [
    "results = trainer.fit(50, 8, 2)\n",
    "logs[\"psp_seresnext\"] = results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = smp.FPN(\"se_resnext50_32x4d\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "optimizer = optim.Adam([\n",
    "    {\"params\": model.encoder.parameters(), \"lr\" : 1e-6},\n",
    "    {\"params\": model.decoder.parameters(), \"lr\" : 1e-4},\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "trainer = Trainer(model, \"models/fpn_seresnext50.pth\", \n",
    "                  TRAIN, TRAIN_MSK, \n",
    "                  VAL, VAL_MSK,\n",
    "                  augmentations=aug,\n",
    "                  optimizer=optimizer,\n",
    "                  device=DEVICE\n",
    "                 )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch: 0\n",
      "train: 100%|██████████████████| 165/165 [03:09<00:00,  1.15s/it, bce_dice_loss - 3.073, iou - 0.6783, f-score - 0.8048]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.85it/s, bce_dice_loss - 0.8691, iou - 0.8364, f-score - 0.9058]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 1\n",
      "train: 100%|██████████████████| 165/165 [03:18<00:00,  1.21s/it, bce_dice_loss - 1.598, iou - 0.7398, f-score - 0.8492]\n",
      "valid: 100%|██████████████████| 100/100 [00:07<00:00, 13.14it/s, bce_dice_loss - 0.6131, iou - 0.8638, f-score - 0.921]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 2\n",
      "train: 100%|██████████████████| 165/165 [03:18<00:00,  1.20s/it, bce_dice_loss - 1.278, iou - 0.7631, f-score - 0.8644]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.37it/s, bce_dice_loss - 0.4634, iou - 0.8781, f-score - 0.9319]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 3\n",
      "train: 100%|██████████████████| 165/165 [03:18<00:00,  1.20s/it, bce_dice_loss - 1.104, iou - 0.7728, f-score - 0.8707]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.24it/s, bce_dice_loss - 0.4463, iou - 0.8849, f-score - 0.9356]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 4\n",
      "train: 100%|██████████████████| 165/165 [03:18<00:00,  1.20s/it, bce_dice_loss - 0.9521, iou - 0.7965, f-score - 0.886]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 16.21it/s, bce_dice_loss - 0.399, iou - 0.8889, f-score - 0.9383]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 5\n",
      "train: 100%|█████████████████| 165/165 [03:19<00:00,  1.21s/it, bce_dice_loss - 0.8527, iou - 0.8025, f-score - 0.8897]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.17it/s, bce_dice_loss - 0.4314, iou - 0.8915, f-score - 0.9392]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 6\n",
      "train: 100%|█████████████████| 165/165 [03:19<00:00,  1.21s/it, bce_dice_loss - 0.8902, iou - 0.7991, f-score - 0.8874]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 16.14it/s, bce_dice_loss - 0.339, iou - 0.8949, f-score - 0.9418]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 7\n",
      "train: 100%|█████████████████| 165/165 [03:18<00:00,  1.21s/it, bce_dice_loss - 0.7843, iou - 0.8124, f-score - 0.8957]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 16.32it/s, bce_dice_loss - 0.3424, iou - 0.8967, f-score - 0.943]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 8\n",
      "train: 100%|█████████████████| 165/165 [03:21<00:00,  1.22s/it, bce_dice_loss - 0.7445, iou - 0.8138, f-score - 0.8965]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.74it/s, bce_dice_loss - 0.3287, iou - 0.9004, f-score - 0.9452]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 9\n",
      "train: 100%|█████████████████| 165/165 [03:20<00:00,  1.21s/it, bce_dice_loss - 0.7286, iou - 0.8162, f-score - 0.8978]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.76it/s, bce_dice_loss - 0.3165, iou - 0.8995, f-score - 0.9447]\n",
      "\n",
      "Epoch: 10\n",
      "train: 100%|██████████████████| 165/165 [03:20<00:00,  1.21s/it, bce_dice_loss - 0.6518, iou - 0.8261, f-score - 0.904]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.28it/s, bce_dice_loss - 0.2846, iou - 0.9051, f-score - 0.9481]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 11\n",
      "train: 100%|██████████████████| 165/165 [03:20<00:00,  1.21s/it, bce_dice_loss - 0.6038, iou - 0.837, f-score - 0.9105]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.07it/s, bce_dice_loss - 0.2804, iou - 0.9029, f-score - 0.9468]\n",
      "\n",
      "Epoch: 12\n",
      "train: 100%|█████████████████| 165/165 [03:23<00:00,  1.23s/it, bce_dice_loss - 0.5967, iou - 0.8372, f-score - 0.9109]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.70it/s, bce_dice_loss - 0.2672, iou - 0.9085, f-score - 0.9498]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 13\n",
      "train: 100%|██████████████████| 165/165 [03:23<00:00,  1.23s/it, bce_dice_loss - 0.5805, iou - 0.836, f-score - 0.9099]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.67it/s, bce_dice_loss - 0.3063, iou - 0.904, f-score - 0.9466]\n",
      "\n",
      "Epoch: 14\n",
      "train: 100%|█████████████████| 165/165 [03:22<00:00,  1.22s/it, bce_dice_loss - 0.5539, iou - 0.8384, f-score - 0.9113]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.09it/s, bce_dice_loss - 0.2607, iou - 0.9089, f-score - 0.9498]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 15\n",
      "train: 100%|█████████████████| 165/165 [03:20<00:00,  1.22s/it, bce_dice_loss - 0.4999, iou - 0.8502, f-score - 0.9186]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 15.88it/s, bce_dice_loss - 0.2259, iou - 0.912, f-score - 0.9521]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 16\n",
      "train: 100%|█████████████████| 165/165 [03:20<00:00,  1.22s/it, bce_dice_loss - 0.4654, iou - 0.8548, f-score - 0.9212]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 15.93it/s, bce_dice_loss - 0.2252, iou - 0.914, f-score - 0.9531]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 17\n",
      "train: 100%|█████████████████| 165/165 [03:21<00:00,  1.22s/it, bce_dice_loss - 0.5141, iou - 0.8508, f-score - 0.9188]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.12it/s, bce_dice_loss - 0.2305, iou - 0.9137, f-score - 0.9528]\n",
      "\n",
      "Epoch: 18\n",
      "train: 100%|█████████████████| 165/165 [03:22<00:00,  1.23s/it, bce_dice_loss - 0.4369, iou - 0.8587, f-score - 0.9235]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.34it/s, bce_dice_loss - 0.2156, iou - 0.9147, f-score - 0.9538]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 19\n",
      "train: 100%|███████████████████| 165/165 [03:19<00:00,  1.21s/it, bce_dice_loss - 0.44, iou - 0.8586, f-score - 0.9233]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.60it/s, bce_dice_loss - 0.2234, iou - 0.9169, f-score - 0.9548]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 20\n",
      "train: 100%|█████████████████| 165/165 [03:21<00:00,  1.22s/it, bce_dice_loss - 0.4331, iou - 0.8619, f-score - 0.9252]\n",
      "valid: 100%|██████████████████| 100/100 [00:08<00:00, 12.48it/s, bce_dice_loss - 0.2339, iou - 0.918, f-score - 0.9551]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 21\n",
      "train: 100%|█████████████████| 165/165 [03:23<00:00,  1.23s/it, bce_dice_loss - 0.3948, iou - 0.8697, f-score - 0.9299]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 15.45it/s, bce_dice_loss - 0.2076, iou - 0.918, f-score - 0.9552]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 22\n",
      "train: 100%|█████████████████| 165/165 [03:21<00:00,  1.22s/it, bce_dice_loss - 0.3696, iou - 0.8726, f-score - 0.9317]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.82it/s, bce_dice_loss - 0.1918, iou - 0.9215, f-score - 0.9577]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 23\n",
      "train: 100%|██████████████████| 165/165 [03:20<00:00,  1.21s/it, bce_dice_loss - 0.3774, iou - 0.872, f-score - 0.9312]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.25it/s, bce_dice_loss - 0.1873, iou - 0.9232, f-score - 0.9584]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 24\n",
      "train: 100%|█████████████████| 165/165 [03:21<00:00,  1.22s/it, bce_dice_loss - 0.3529, iou - 0.8752, f-score - 0.9331]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.59it/s, bce_dice_loss - 0.2229, iou - 0.9201, f-score - 0.9562]\n",
      "\n",
      "Epoch: 25\n",
      "train: 100%|█████████████████| 165/165 [03:21<00:00,  1.22s/it, bce_dice_loss - 0.3484, iou - 0.8781, f-score - 0.9347]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.68it/s, bce_dice_loss - 0.1795, iou - 0.9215, f-score - 0.9574]\n",
      "\n",
      "Epoch: 26\n",
      "train: 100%|██████████████████| 165/165 [03:21<00:00,  1.22s/it, bce_dice_loss - 0.354, iou - 0.8755, f-score - 0.9333]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.57it/s, bce_dice_loss - 0.1912, iou - 0.9245, f-score - 0.9587]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 27\n",
      "train: 100%|█████████████████| 165/165 [03:19<00:00,  1.21s/it, bce_dice_loss - 0.3494, iou - 0.8766, f-score - 0.9339]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 15.69it/s, bce_dice_loss - 0.1807, iou - 0.9246, f-score - 0.959]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 28\n",
      "train: 100%|██████████████████| 165/165 [03:19<00:00,  1.21s/it, bce_dice_loss - 0.3347, iou - 0.881, f-score - 0.9364]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.24it/s, bce_dice_loss - 0.1843, iou - 0.9258, f-score - 0.9594]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 29\n",
      "train: 100%|█████████████████| 165/165 [03:24<00:00,  1.24s/it, bce_dice_loss - 0.3025, iou - 0.8881, f-score - 0.9405]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 15.94it/s, bce_dice_loss - 0.162, iou - 0.9288, f-score - 0.9615]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 30\n",
      "train: 100%|██████████████████| 165/165 [03:20<00:00,  1.22s/it, bce_dice_loss - 0.3142, iou - 0.884, f-score - 0.9381]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.61it/s, bce_dice_loss - 0.1593, iou - 0.9314, f-score - 0.9628]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 31\n",
      "train: 100%|█████████████████| 165/165 [03:20<00:00,  1.22s/it, bce_dice_loss - 0.3174, iou - 0.8848, f-score - 0.9384]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.74it/s, bce_dice_loss - 0.1571, iou - 0.9302, f-score - 0.9624]\n",
      "\n",
      "Epoch: 32\n",
      "train: 100%|█████████████████| 165/165 [03:22<00:00,  1.23s/it, bce_dice_loss - 0.3001, iou - 0.8881, f-score - 0.9404]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.62it/s, bce_dice_loss - 0.1727, iou - 0.9258, f-score - 0.9595]\n",
      "\n",
      "Epoch: 33\n",
      "train: 100%|█████████████████| 165/165 [03:23<00:00,  1.24s/it, bce_dice_loss - 0.3048, iou - 0.8889, f-score - 0.9409]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.46it/s, bce_dice_loss - 0.1657, iou - 0.9306, f-score - 0.9623]\n",
      "\n",
      "Epoch: 34\n",
      "train: 100%|█████████████████| 165/165 [03:21<00:00,  1.22s/it, bce_dice_loss - 0.2846, iou - 0.8931, f-score - 0.9432]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 14.74it/s, bce_dice_loss - 0.1597, iou - 0.931, f-score - 0.9626]\n",
      "\n",
      "Epoch: 35\n",
      "train: 100%|█████████████████| 165/165 [03:22<00:00,  1.22s/it, bce_dice_loss - 0.2933, iou - 0.8912, f-score - 0.9421]\n",
      "valid: 100%|█████████████████| 100/100 [00:07<00:00, 14.13it/s, bce_dice_loss - 0.1602, iou - 0.9305, f-score - 0.9626]\n",
      "\n",
      "Epoch: 36\n",
      "train: 100%|██████████████████| 165/165 [03:20<00:00,  1.22s/it, bce_dice_loss - 0.286, iou - 0.8913, f-score - 0.9423]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.13it/s, bce_dice_loss - 0.1501, iou - 0.9339, f-score - 0.9642]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 37\n",
      "train: 100%|█████████████████| 165/165 [03:20<00:00,  1.22s/it, bce_dice_loss - 0.2729, iou - 0.8942, f-score - 0.9438]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.91it/s, bce_dice_loss - 0.1427, iou - 0.9348, f-score - 0.9649]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 38\n",
      "train: 100%|█████████████████| 165/165 [03:19<00:00,  1.21s/it, bce_dice_loss - 0.2707, iou - 0.8974, f-score - 0.9456]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.69it/s, bce_dice_loss - 0.1698, iou - 0.9291, f-score - 0.9611]\n",
      "\n",
      "Epoch: 39\n",
      "train: 100%|█████████████████| 165/165 [03:19<00:00,  1.21s/it, bce_dice_loss - 0.2626, iou - 0.8972, f-score - 0.9456]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.64it/s, bce_dice_loss - 0.1454, iou - 0.9359, f-score - 0.9652]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 40\n",
      "train: 100%|█████████████████| 165/165 [03:20<00:00,  1.21s/it, bce_dice_loss - 0.2649, iou - 0.8973, f-score - 0.9456]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 14.90it/s, bce_dice_loss - 0.1392, iou - 0.9359, f-score - 0.9654]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 41\n",
      "train: 100%|█████████████████| 165/165 [03:18<00:00,  1.20s/it, bce_dice_loss - 0.2666, iou - 0.8974, f-score - 0.9455]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.63it/s, bce_dice_loss - 0.1355, iou - 0.9353, f-score - 0.9652]\n",
      "\n",
      "Epoch: 42\n",
      "train: 100%|█████████████████| 165/165 [03:18<00:00,  1.20s/it, bce_dice_loss - 0.2602, iou - 0.8991, f-score - 0.9467]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 16.24it/s, bce_dice_loss - 0.142, iou - 0.9358, f-score - 0.9652]\n",
      "\n",
      "Epoch: 43\n",
      "train: 100%|█████████████████| 165/165 [03:16<00:00,  1.19s/it, bce_dice_loss - 0.2589, iou - 0.8996, f-score - 0.9469]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 15.95it/s, bce_dice_loss - 0.1337, iou - 0.9375, f-score - 0.9664]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 44\n",
      "train: 100%|█████████████████| 165/165 [03:16<00:00,  1.19s/it, bce_dice_loss - 0.2447, iou - 0.9021, f-score - 0.9482]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.43it/s, bce_dice_loss - 0.1422, iou - 0.9363, f-score - 0.9654]\n",
      "\n",
      "Epoch: 45\n",
      "train: 100%|█████████████████| 165/165 [03:16<00:00,  1.19s/it, bce_dice_loss - 0.2247, iou - 0.9078, f-score - 0.9515]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.17it/s, bce_dice_loss - 0.1362, iou - 0.9388, f-score - 0.9669]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 46\n",
      "train: 100%|█████████████████| 165/165 [03:16<00:00,  1.19s/it, bce_dice_loss - 0.2454, iou - 0.9056, f-score - 0.9501]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.19it/s, bce_dice_loss - 0.1285, iou - 0.9401, f-score - 0.9677]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 47\n",
      "train: 100%|█████████████████| 165/165 [03:16<00:00,  1.19s/it, bce_dice_loss - 0.2292, iou - 0.9078, f-score - 0.9515]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 16.29it/s, bce_dice_loss - 0.1275, iou - 0.9404, f-score - 0.968]\n",
      "Model saved!\n",
      "\n",
      "Epoch: 48\n",
      "train: 100%|█████████████████| 165/165 [03:17<00:00,  1.20s/it, bce_dice_loss - 0.2295, iou - 0.9061, f-score - 0.9505]\n",
      "valid: 100%|██████████████████| 100/100 [00:06<00:00, 15.80it/s, bce_dice_loss - 0.1371, iou - 0.939, f-score - 0.9669]\n",
      "\n",
      "Epoch: 49\n",
      "train: 100%|██████████████████| 165/165 [03:15<00:00,  1.19s/it, bce_dice_loss - 0.229, iou - 0.9061, f-score - 0.9505]\n",
      "valid: 100%|█████████████████| 100/100 [00:06<00:00, 16.31it/s, bce_dice_loss - 0.1324, iou - 0.9378, f-score - 0.9669]\n",
      "Training completed sucessfully.\n",
      "Best model IoU: 0.9403736621141433\n"
     ]
    }
   ],
   "source": [
    "results = trainer.fit(50, 8, 2)\n",
    "logs[\"fpn_seresnext\"] = results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open(\"results/model_logs.json\", \"w\") as f:\n",
    "    json.dump(logs, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}