Train_Segmentation_Model_FPN+inceptionv4.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "790dec7b",
   "metadata": {},
   "source": [
    "# IMPORTS\n",
    "\n",
    "First we started with some python imports. For model building we used the <mark>**segmentation_models_pytorch library**</mark>. This has library has a lot segmentation model architechtures. Before you can import the library it needs to be ensured that the library is installed in the conda environment. For this in the anaconda terminal just type the below command to install the library:-\n",
    "\n",
    "<div class=\"alert alert-block alert-success\">\n",
    "<b>Conda Command:</b> conda install -q segmentation-models-pytorch\n",
    "</div>\n",
    "\n",
    "Once, the libray is installed it can be imported in the script safely. We also imported the utility script Trainer for model training."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "82452039",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import segmentation_models_pytorch as smp\n",
    "from Utilities.Trainer import Trainer\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "from torchvision import models\n",
    "import torch.multiprocessing as mp\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "69ec7cbf",
   "metadata": {},
   "source": [
    "The below statement helps in better error tracking when training is done in cuda. More details about this can be found in the below url:-\n",
    "\n",
    "https://pytorch.org/docs/stable/notes/cuda.html"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "73740900",
   "metadata": {},
   "outputs": [],
   "source": [
    "CUDA_LAUNCH_BLOCKING=\"1\""
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5a6a5342",
   "metadata": {},
   "source": [
    "# Model Building\n",
    "\n",
    "Next a FPN model with inceptionv3 encoder was created with imagenet weights from the smp library."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "04921d03",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "FPN(\n",
      "  (encoder): InceptionV4Encoder(\n",
      "    (features): Sequential(\n",
      "      (0): BasicConv2d(\n",
      "        (conv): Conv2d(3, 32, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "        (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (1): BasicConv2d(\n",
      "        (conv): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn): BatchNorm2d(32, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): BasicConv2d(\n",
      "        (conv): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (3): Mixed_3a(\n",
      "        (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=(1, 1), dilation=1, ceil_mode=False)\n",
      "        (conv): BasicConv2d(\n",
      "          (conv): Conv2d(64, 96, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "      )\n",
      "      (4): Mixed_4a(\n",
      "        (branch0): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(160, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(64, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch1): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(160, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(64, 64, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(64, 64, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (3): BasicConv2d(\n",
      "            (conv): Conv2d(64, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (5): Mixed_5a(\n",
      "        (conv): BasicConv2d(\n",
      "          (conv): Conv2d(192, 192, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=(1, 1), dilation=1, ceil_mode=False)\n",
      "      )\n",
      "      (6): Inception_A(\n",
      "        (branch0): BasicConv2d(\n",
      "          (conv): Conv2d(384, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(384, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(64, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch2): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(384, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(64, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch3): Sequential(\n",
      "          (0): AvgPool2d(kernel_size=3, stride=1, padding=1)\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(384, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (7): Inception_A(\n",
      "        (branch0): BasicConv2d(\n",
      "          (conv): Conv2d(384, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(384, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(64, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch2): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(384, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(64, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch3): Sequential(\n",
      "          (0): AvgPool2d(kernel_size=3, stride=1, padding=1)\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(384, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (8): Inception_A(\n",
      "        (branch0): BasicConv2d(\n",
      "          (conv): Conv2d(384, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(384, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(64, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch2): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(384, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(64, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch3): Sequential(\n",
      "          (0): AvgPool2d(kernel_size=3, stride=1, padding=1)\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(384, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (9): Inception_A(\n",
      "        (branch0): BasicConv2d(\n",
      "          (conv): Conv2d(384, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(384, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(64, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch2): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(384, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(64, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(96, 96, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch3): Sequential(\n",
      "          (0): AvgPool2d(kernel_size=3, stride=1, padding=1)\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(384, 96, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(96, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (10): Reduction_A(\n",
      "        (branch0): BasicConv2d(\n",
      "          (conv): Conv2d(384, 384, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(384, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(384, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(192, 224, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(224, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch2): MaxPool2d(kernel_size=3, stride=2, padding=(1, 1), dilation=1, ceil_mode=False)\n",
      "      )\n",
      "      (11): Inception_B(\n",
      "        (branch0): BasicConv2d(\n",
      "          (conv): Conv2d(1024, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(384, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(192, 224, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(224, 256, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch2): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(192, 192, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(192, 224, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (3): BasicConv2d(\n",
      "            (conv): Conv2d(224, 224, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (4): BasicConv2d(\n",
      "            (conv): Conv2d(224, 256, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch3): Sequential(\n",
      "          (0): AvgPool2d(kernel_size=3, stride=1, padding=1)\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (12): Inception_B(\n",
      "        (branch0): BasicConv2d(\n",
      "          (conv): Conv2d(1024, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(384, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(192, 224, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(224, 256, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch2): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(192, 192, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(192, 224, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (3): BasicConv2d(\n",
      "            (conv): Conv2d(224, 224, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (4): BasicConv2d(\n",
      "            (conv): Conv2d(224, 256, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch3): Sequential(\n",
      "          (0): AvgPool2d(kernel_size=3, stride=1, padding=1)\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (13): Inception_B(\n",
      "        (branch0): BasicConv2d(\n",
      "          (conv): Conv2d(1024, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(384, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(192, 224, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(224, 256, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch2): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(192, 192, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(192, 224, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (3): BasicConv2d(\n",
      "            (conv): Conv2d(224, 224, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (4): BasicConv2d(\n",
      "            (conv): Conv2d(224, 256, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch3): Sequential(\n",
      "          (0): AvgPool2d(kernel_size=3, stride=1, padding=1)\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (14): Inception_B(\n",
      "        (branch0): BasicConv2d(\n",
      "          (conv): Conv2d(1024, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(384, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(192, 224, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(224, 256, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch2): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(192, 192, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(192, 224, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (3): BasicConv2d(\n",
      "            (conv): Conv2d(224, 224, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (4): BasicConv2d(\n",
      "            (conv): Conv2d(224, 256, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch3): Sequential(\n",
      "          (0): AvgPool2d(kernel_size=3, stride=1, padding=1)\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (15): Inception_B(\n",
      "        (branch0): BasicConv2d(\n",
      "          (conv): Conv2d(1024, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(384, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(192, 224, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(224, 256, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch2): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(192, 192, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(192, 224, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (3): BasicConv2d(\n",
      "            (conv): Conv2d(224, 224, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (4): BasicConv2d(\n",
      "            (conv): Conv2d(224, 256, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch3): Sequential(\n",
      "          (0): AvgPool2d(kernel_size=3, stride=1, padding=1)\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (16): Inception_B(\n",
      "        (branch0): BasicConv2d(\n",
      "          (conv): Conv2d(1024, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(384, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(192, 224, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(224, 256, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch2): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(192, 192, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(192, 224, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (3): BasicConv2d(\n",
      "            (conv): Conv2d(224, 224, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (4): BasicConv2d(\n",
      "            (conv): Conv2d(224, 256, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch3): Sequential(\n",
      "          (0): AvgPool2d(kernel_size=3, stride=1, padding=1)\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (17): Inception_B(\n",
      "        (branch0): BasicConv2d(\n",
      "          (conv): Conv2d(1024, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(384, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(192, 224, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(224, 256, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch2): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(192, 192, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(192, 224, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (3): BasicConv2d(\n",
      "            (conv): Conv2d(224, 224, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(224, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (4): BasicConv2d(\n",
      "            (conv): Conv2d(224, 256, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch3): Sequential(\n",
      "          (0): AvgPool2d(kernel_size=3, stride=1, padding=1)\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (18): Reduction_B(\n",
      "        (branch0): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 192, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(192, 192, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(192, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch1): Sequential(\n",
      "          (0): BasicConv2d(\n",
      "            (conv): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(256, 256, kernel_size=(1, 7), stride=(1, 1), padding=(0, 3), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (2): BasicConv2d(\n",
      "            (conv): Conv2d(256, 320, kernel_size=(7, 1), stride=(1, 1), padding=(3, 0), bias=False)\n",
      "            (bn): BatchNorm2d(320, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "          (3): BasicConv2d(\n",
      "            (conv): Conv2d(320, 320, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(320, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "        (branch2): MaxPool2d(kernel_size=3, stride=2, padding=(1, 1), dilation=1, ceil_mode=False)\n",
      "      )\n",
      "      (19): Inception_C(\n",
      "        (branch0): BasicConv2d(\n",
      "          (conv): Conv2d(1536, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1_0): BasicConv2d(\n",
      "          (conv): Conv2d(1536, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(384, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1_1a): BasicConv2d(\n",
      "          (conv): Conv2d(384, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1_1b): BasicConv2d(\n",
      "          (conv): Conv2d(384, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch2_0): BasicConv2d(\n",
      "          (conv): Conv2d(1536, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(384, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch2_1): BasicConv2d(\n",
      "          (conv): Conv2d(384, 448, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)\n",
      "          (bn): BatchNorm2d(448, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch2_2): BasicConv2d(\n",
      "          (conv): Conv2d(448, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)\n",
      "          (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch2_3a): BasicConv2d(\n",
      "          (conv): Conv2d(512, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch2_3b): BasicConv2d(\n",
      "          (conv): Conv2d(512, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch3): Sequential(\n",
      "          (0): AvgPool2d(kernel_size=3, stride=1, padding=1)\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(1536, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (20): Inception_C(\n",
      "        (branch0): BasicConv2d(\n",
      "          (conv): Conv2d(1536, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1_0): BasicConv2d(\n",
      "          (conv): Conv2d(1536, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(384, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1_1a): BasicConv2d(\n",
      "          (conv): Conv2d(384, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1_1b): BasicConv2d(\n",
      "          (conv): Conv2d(384, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch2_0): BasicConv2d(\n",
      "          (conv): Conv2d(1536, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(384, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch2_1): BasicConv2d(\n",
      "          (conv): Conv2d(384, 448, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)\n",
      "          (bn): BatchNorm2d(448, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch2_2): BasicConv2d(\n",
      "          (conv): Conv2d(448, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)\n",
      "          (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch2_3a): BasicConv2d(\n",
      "          (conv): Conv2d(512, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch2_3b): BasicConv2d(\n",
      "          (conv): Conv2d(512, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch3): Sequential(\n",
      "          (0): AvgPool2d(kernel_size=3, stride=1, padding=1)\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(1536, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (21): Inception_C(\n",
      "        (branch0): BasicConv2d(\n",
      "          (conv): Conv2d(1536, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1_0): BasicConv2d(\n",
      "          (conv): Conv2d(1536, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(384, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1_1a): BasicConv2d(\n",
      "          (conv): Conv2d(384, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch1_1b): BasicConv2d(\n",
      "          (conv): Conv2d(384, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch2_0): BasicConv2d(\n",
      "          (conv): Conv2d(1536, 384, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(384, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch2_1): BasicConv2d(\n",
      "          (conv): Conv2d(384, 448, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)\n",
      "          (bn): BatchNorm2d(448, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch2_2): BasicConv2d(\n",
      "          (conv): Conv2d(448, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)\n",
      "          (bn): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch2_3a): BasicConv2d(\n",
      "          (conv): Conv2d(512, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch2_3b): BasicConv2d(\n",
      "          (conv): Conv2d(512, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (relu): ReLU(inplace=True)\n",
      "        )\n",
      "        (branch3): Sequential(\n",
      "          (0): AvgPool2d(kernel_size=3, stride=1, padding=1)\n",
      "          (1): BasicConv2d(\n",
      "            (conv): Conv2d(1536, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "            (bn): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "            (relu): ReLU(inplace=True)\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "    )\n",
      "    (avg_pool): AvgPool2d(kernel_size=8, stride=8, padding=0)\n",
      "  )\n",
      "  (decoder): FPNDecoder(\n",
      "    (p5): Conv2d(1536, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "    (p4): FPNBlock(\n",
      "      (skip_conv): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "    )\n",
      "    (p3): FPNBlock(\n",
      "      (skip_conv): Conv2d(384, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "    )\n",
      "    (p2): FPNBlock(\n",
      "      (skip_conv): Conv2d(192, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "    )\n",
      "    (seg_blocks): ModuleList(\n",
      "      (0): SegmentationBlock(\n",
      "        (block): Sequential(\n",
      "          (0): Conv3x3GNReLU(\n",
      "            (block): Sequential(\n",
      "              (0): Conv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "              (1): GroupNorm(32, 128, eps=1e-05, affine=True)\n",
      "              (2): ReLU(inplace=True)\n",
      "            )\n",
      "          )\n",
      "          (1): Conv3x3GNReLU(\n",
      "            (block): Sequential(\n",
      "              (0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "              (1): GroupNorm(32, 128, eps=1e-05, affine=True)\n",
      "              (2): ReLU(inplace=True)\n",
      "            )\n",
      "          )\n",
      "          (2): Conv3x3GNReLU(\n",
      "            (block): Sequential(\n",
      "              (0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "              (1): GroupNorm(32, 128, eps=1e-05, affine=True)\n",
      "              (2): ReLU(inplace=True)\n",
      "            )\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (1): SegmentationBlock(\n",
      "        (block): Sequential(\n",
      "          (0): Conv3x3GNReLU(\n",
      "            (block): Sequential(\n",
      "              (0): Conv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "              (1): GroupNorm(32, 128, eps=1e-05, affine=True)\n",
      "              (2): ReLU(inplace=True)\n",
      "            )\n",
      "          )\n",
      "          (1): Conv3x3GNReLU(\n",
      "            (block): Sequential(\n",
      "              (0): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "              (1): GroupNorm(32, 128, eps=1e-05, affine=True)\n",
      "              (2): ReLU(inplace=True)\n",
      "            )\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (2): SegmentationBlock(\n",
      "        (block): Sequential(\n",
      "          (0): Conv3x3GNReLU(\n",
      "            (block): Sequential(\n",
      "              (0): Conv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "              (1): GroupNorm(32, 128, eps=1e-05, affine=True)\n",
      "              (2): ReLU(inplace=True)\n",
      "            )\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "      (3): SegmentationBlock(\n",
      "        (block): Sequential(\n",
      "          (0): Conv3x3GNReLU(\n",
      "            (block): Sequential(\n",
      "              (0): Conv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "              (1): GroupNorm(32, 128, eps=1e-05, affine=True)\n",
      "              (2): ReLU(inplace=True)\n",
      "            )\n",
      "          )\n",
      "        )\n",
      "      )\n",
      "    )\n",
      "    (merge): MergeBlock()\n",
      "    (dropout): Dropout2d(p=0.2, inplace=True)\n",
      "  )\n",
      "  (segmentation_head): SegmentationHead(\n",
      "    (0): Conv2d(128, 6, kernel_size=(1, 1), stride=(1, 1))\n",
      "    (1): UpsamplingBilinear2d(scale_factor=4.0, mode=bilinear)\n",
      "    (2): Activation(\n",
      "      (activation): Identity()\n",
      "    )\n",
      "  )\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "torch.manual_seed(42)\n",
    "ENCODER = 'inceptionv4'\n",
    "ENCODER_WEIGHTS = 'imagenet'\n",
    "CLASSES = 6\n",
    "ACTIVATION = None # could be None for logits or 'softmax2d' for multiclass segmentation\n",
    "\n",
    "# create segmentation model with pretrained encoder\n",
    "model = smp.FPN(\n",
    "    encoder_name=ENCODER, \n",
    "    encoder_weights=ENCODER_WEIGHTS, \n",
    "    in_channels=3,\n",
    "    classes=CLASSES, \n",
    "    activation=ACTIVATION,\n",
    ")\n",
    "print(model)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "22b27de4",
   "metadata": {},
   "source": [
    "# Model Training\n",
    "\n",
    "Next the model training was started with 30 epochs and initial learning rate as 5e-04. The model checkpoint was saved whenever the validation dice was reported the greater than the previous best validation dice.\n",
    "\n",
    "In the below results you can see the training starting from epoch 6. This is because the kernel was initally interuptted after epoch 5. So, I had to start all over again. Since, checkpoint was saved upto epoch number 5 so it started from epoch 6."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "6f528cc3",
   "metadata": {},
   "outputs": [],
   "source": [
    "lr = 5e-04\n",
    "epochs = 20\n",
    "path = \"./Models/model_fpn_inceptionv4.pth\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "82b95ac9",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Current Device is GPU: 0\n",
      "Total memory allocated in GB: 0.5\n",
      "Total memory cached in GB: 1.0\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 0 | phase: train | ⏰: 11:14:41\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████| 310/310 [04:03<00:00,  1.27it/s, loss=0.0182]\n",
      "C:\\Users\\SIDDHARTA\\OneDrive\\Documents\\NEU-DET\\Utilities\\Meter.py:40: RuntimeWarning: Mean of empty slice\n",
      "  dice_neg = np.nanmean(self.dice_neg_scores)\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.1455 | IoU: 0.3235 | dice: 0.4151 | dice_neg: nan | dice_pos: 0.4151\n",
      "Starting epoch: 0 | phase: val | ⏰: 11:18:45\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████| 133/133 [01:23<00:00,  1.58it/s, loss=0.0122]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0972 | IoU: 0.5192 | dice: 0.6472 | dice_neg: nan | dice_pos: 0.6472\n",
      "******** New optimal found, saving state ********\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 1 | phase: train | ⏰: 11:20:10\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████| 310/310 [11:17<00:00,  2.19s/it, loss=0.0158]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.1267 | IoU: 0.4099 | dice: 0.5232 | dice_neg: nan | dice_pos: 0.5232\n",
      "Starting epoch: 1 | phase: val | ⏰: 11:31:28\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|█████████████████████████████████████████████████████████████████████| 133/133 [01:24<00:00,  1.57it/s, loss=0.01]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0803 | IoU: 0.5853 | dice: 0.7152 | dice_neg: nan | dice_pos: 0.7152\n",
      "******** New optimal found, saving state ********\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 2 | phase: train | ⏰: 11:32:54\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████| 310/310 [11:17<00:00,  2.19s/it, loss=0.0135]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.1077 | IoU: 0.4745 | dice: 0.5949 | dice_neg: nan | dice_pos: 0.5949\n",
      "Starting epoch: 2 | phase: val | ⏰: 11:44:12\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████| 133/133 [01:23<00:00,  1.59it/s, loss=0.0093]\n",
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0744 | IoU: 0.5291 | dice: 0.6454 | dice_neg: nan | dice_pos: 0.6454\n",
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 3 | phase: train | ⏰: 11:45:35\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████| 310/310 [11:16<00:00,  2.18s/it, loss=0.0123]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0983 | IoU: 0.5043 | dice: 0.6252 | dice_neg: nan | dice_pos: 0.6252\n",
      "Starting epoch: 3 | phase: val | ⏰: 11:56:52\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████| 133/133 [01:25<00:00,  1.55it/s, loss=0.0085]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0680 | IoU: 0.6090 | dice: 0.7333 | dice_neg: nan | dice_pos: 0.7333\n",
      "******** New optimal found, saving state ********\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 4 | phase: train | ⏰: 11:58:20\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████| 310/310 [11:20<00:00,  2.19s/it, loss=0.0112]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0899 | IoU: 0.5305 | dice: 0.6541 | dice_neg: nan | dice_pos: 0.6541\n",
      "Starting epoch: 4 | phase: val | ⏰: 12:09:40\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 133/133 [01:27<00:00,  1.53it/s, loss=0.00826]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0661 | IoU: 0.6250 | dice: 0.7465 | dice_neg: nan | dice_pos: 0.7465\n",
      "******** New optimal found, saving state ********\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 5 | phase: train | ⏰: 12:11:09\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████| 310/310 [11:21<00:00,  2.20s/it, loss=0.0101]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0810 | IoU: 0.5670 | dice: 0.6911 | dice_neg: nan | dice_pos: 0.6911\n",
      "Starting epoch: 5 | phase: val | ⏰: 12:22:30\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 133/133 [01:27<00:00,  1.53it/s, loss=0.00786]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0629 | IoU: 0.6540 | dice: 0.7751 | dice_neg: nan | dice_pos: 0.7751\n",
      "******** New optimal found, saving state ********\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 6 | phase: train | ⏰: 12:23:59\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 310/310 [11:23<00:00,  2.21s/it, loss=0.00979]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0783 | IoU: 0.5759 | dice: 0.7004 | dice_neg: nan | dice_pos: 0.7004\n",
      "Starting epoch: 6 | phase: val | ⏰: 12:35:23\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 133/133 [01:24<00:00,  1.58it/s, loss=0.00789]\n",
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0631 | IoU: 0.6117 | dice: 0.7381 | dice_neg: nan | dice_pos: 0.7381\n",
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 7 | phase: train | ⏰: 12:36:47\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 310/310 [11:18<00:00,  2.19s/it, loss=0.00991]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0793 | IoU: 0.5757 | dice: 0.6991 | dice_neg: nan | dice_pos: 0.6991\n",
      "Starting epoch: 7 | phase: val | ⏰: 12:48:05\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████| 133/133 [01:24<00:00,  1.58it/s, loss=0.0085]\n",
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0680 | IoU: 0.6022 | dice: 0.7281 | dice_neg: nan | dice_pos: 0.7281\n",
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 8 | phase: train | ⏰: 12:49:29\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 310/310 [11:19<00:00,  2.19s/it, loss=0.00954]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0763 | IoU: 0.5872 | dice: 0.7120 | dice_neg: nan | dice_pos: 0.7120\n",
      "Starting epoch: 8 | phase: val | ⏰: 13:00:49\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████| 133/133 [01:24<00:00,  1.58it/s, loss=0.0074]\n",
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0592 | IoU: 0.6423 | dice: 0.7643 | dice_neg: nan | dice_pos: 0.7643\n",
      "\n",
      "Learning Rate =  0.0005\n",
      "Starting epoch: 9 | phase: train | ⏰: 13:02:13\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 310/310 [11:20<00:00,  2.19s/it, loss=0.00844]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0675 | IoU: 0.6207 | dice: 0.7430 | dice_neg: nan | dice_pos: 0.7430\n",
      "Starting epoch: 9 | phase: val | ⏰: 13:13:34\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 133/133 [01:24<00:00,  1.58it/s, loss=0.00793]\n",
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0634 | IoU: 0.6315 | dice: 0.7573 | dice_neg: nan | dice_pos: 0.7573\n",
      "Epoch    10: reducing learning rate of group 0 to 5.0000e-05.\n",
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 10 | phase: train | ⏰: 13:14:58\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 310/310 [11:21<00:00,  2.20s/it, loss=0.00807]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0646 | IoU: 0.6383 | dice: 0.7587 | dice_neg: nan | dice_pos: 0.7587\n",
      "Starting epoch: 10 | phase: val | ⏰: 13:26:19\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 133/133 [01:24<00:00,  1.57it/s, loss=0.00693]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0554 | IoU: 0.6621 | dice: 0.7822 | dice_neg: nan | dice_pos: 0.7822\n",
      "******** New optimal found, saving state ********\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 11 | phase: train | ⏰: 13:27:45\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 310/310 [11:21<00:00,  2.20s/it, loss=0.00729]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0583 | IoU: 0.6489 | dice: 0.7696 | dice_neg: nan | dice_pos: 0.7696\n",
      "Starting epoch: 11 | phase: val | ⏰: 13:39:06\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 133/133 [01:33<00:00,  1.43it/s, loss=0.00686]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0549 | IoU: 0.6624 | dice: 0.7829 | dice_neg: nan | dice_pos: 0.7829\n",
      "******** New optimal found, saving state ********\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 12 | phase: train | ⏰: 13:40:41\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 310/310 [11:18<00:00,  2.19s/it, loss=0.00727]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0581 | IoU: 0.6529 | dice: 0.7720 | dice_neg: nan | dice_pos: 0.7720\n",
      "Starting epoch: 12 | phase: val | ⏰: 13:51:59\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 133/133 [01:07<00:00,  1.96it/s, loss=0.00687]\n",
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0549 | IoU: 0.6605 | dice: 0.7808 | dice_neg: nan | dice_pos: 0.7808\n",
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 13 | phase: train | ⏰: 13:53:07\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 310/310 [11:17<00:00,  2.18s/it, loss=0.00716]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0573 | IoU: 0.6557 | dice: 0.7752 | dice_neg: nan | dice_pos: 0.7752\n",
      "Starting epoch: 13 | phase: val | ⏰: 14:04:25\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 133/133 [01:26<00:00,  1.53it/s, loss=0.00684]\n",
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0547 | IoU: 0.6600 | dice: 0.7804 | dice_neg: nan | dice_pos: 0.7804\n",
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 14 | phase: train | ⏰: 14:05:51\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 310/310 [11:17<00:00,  2.18s/it, loss=0.00715]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0572 | IoU: 0.6593 | dice: 0.7785 | dice_neg: nan | dice_pos: 0.7785\n",
      "Starting epoch: 14 | phase: val | ⏰: 14:17:09\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 133/133 [01:23<00:00,  1.58it/s, loss=0.00675]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0540 | IoU: 0.6715 | dice: 0.7897 | dice_neg: nan | dice_pos: 0.7897\n",
      "******** New optimal found, saving state ********\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 15 | phase: train | ⏰: 14:18:35\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 310/310 [11:17<00:00,  2.19s/it, loss=0.00699]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0559 | IoU: 0.6623 | dice: 0.7806 | dice_neg: nan | dice_pos: 0.7806\n",
      "Starting epoch: 15 | phase: val | ⏰: 14:29:53\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 133/133 [01:25<00:00,  1.55it/s, loss=0.00694]\n",
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0555 | IoU: 0.6656 | dice: 0.7844 | dice_neg: nan | dice_pos: 0.7844\n",
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 16 | phase: train | ⏰: 14:31:18\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 310/310 [11:17<00:00,  2.19s/it, loss=0.00689]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0552 | IoU: 0.6659 | dice: 0.7842 | dice_neg: nan | dice_pos: 0.7842\n",
      "Starting epoch: 16 | phase: val | ⏰: 14:42:36\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 133/133 [01:26<00:00,  1.53it/s, loss=0.00686]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0549 | IoU: 0.6729 | dice: 0.7904 | dice_neg: nan | dice_pos: 0.7904\n",
      "******** New optimal found, saving state ********\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 17 | phase: train | ⏰: 14:44:07\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|███████████████████████████████████████████████████████████████████| 310/310 [11:19<00:00,  2.19s/it, loss=0.0068]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0544 | IoU: 0.6689 | dice: 0.7860 | dice_neg: nan | dice_pos: 0.7860\n",
      "Starting epoch: 17 | phase: val | ⏰: 14:55:27\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 133/133 [01:23<00:00,  1.59it/s, loss=0.00686]\n",
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0549 | IoU: 0.6747 | dice: 0.7902 | dice_neg: nan | dice_pos: 0.7902\n",
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 18 | phase: train | ⏰: 14:56:50\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 310/310 [11:16<00:00,  2.18s/it, loss=0.00699]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0559 | IoU: 0.6650 | dice: 0.7823 | dice_neg: nan | dice_pos: 0.7823\n",
      "Starting epoch: 18 | phase: val | ⏰: 15:08:07\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 133/133 [01:23<00:00,  1.60it/s, loss=0.00691]\n",
      "  0%|                                                                                          | 0/310 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0553 | IoU: 0.6594 | dice: 0.7799 | dice_neg: nan | dice_pos: 0.7799\n",
      "\n",
      "Learning Rate =  5e-05\n",
      "Starting epoch: 19 | phase: train | ⏰: 15:09:30\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 310/310 [11:16<00:00,  2.18s/it, loss=0.00685]\n",
      "  0%|                                                                                          | 0/133 [00:00<?, ?it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0548 | IoU: 0.6728 | dice: 0.7898 | dice_neg: nan | dice_pos: 0.7898\n",
      "Starting epoch: 19 | phase: val | ⏰: 15:20:47\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████████████████████████████████████████████████████████████| 133/133 [01:24<00:00,  1.58it/s, loss=0.00691]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0553 | IoU: 0.6633 | dice: 0.7812 | dice_neg: nan | dice_pos: 0.7812\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "model_trainer = Trainer(model, lr, epochs, path)\n",
    "print(f\"Current Device is GPU: {torch.cuda.current_device()}\")\n",
    "print(f\"Total memory allocated in GB: {round(torch.cuda.memory_allocated(0)/1024**3,1)}\")\n",
    "print(f\"Total memory cached in GB: {round(torch.cuda.memory_reserved(0)/1024**3,1)}\")\n",
    "model_trainer.start()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ec46db95",
   "metadata": {},
   "source": [
    "After the model training was done the loss, dice_scores and iou_scores were plotted. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "75baafda",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "losses = model_trainer.losses\n",
    "dice_scores = model_trainer.dice_scores # overall dice\n",
    "iou_scores = model_trainer.iou_scores\n",
    "\n",
    "def plot(scores, name):\n",
    "    plt.figure(figsize=(15,5))\n",
    "    plt.plot(range(len(scores[\"train\"])), scores[\"train\"], label=f'train {name}')\n",
    "    plt.plot(range(len(scores[\"train\"])), scores[\"val\"], label=f'val {name}')\n",
    "    plt.title(f'{name} plot'); plt.xlabel('Epoch'); plt.ylabel(f'{name}');\n",
    "    plt.legend(); \n",
    "    plt.show()\n",
    "\n",
    "plot(losses, \"BCE loss\")\n",
    "plot(dice_scores, \"Dice score\")\n",
    "plot(iou_scores, \"IoU score\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a958bb96",
   "metadata": {},
   "source": [
    "The above results the overall training and validation dice and IoU goes up with epochs and training and validation loss goes down with epochs."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4e5f9cad",
   "metadata": {},
   "source": [
    "# Convert to ONNX\n",
    "\n",
    "Lastly the trained model is converted to ONNX format using the Utility script ONNX_Converter.py."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "edc7a53d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model name :FPN\n",
      "Model checkpoint path is :./Models/model_fpn_inceptionv4.pth\n",
      "Model output path is :./ONNX_models/fpn_inceptionv3.onnx\n",
      "Successfully converted the model to onnx format.\n"
     ]
    }
   ],
   "source": [
    "! python ./Utilities/ONNX_converter.py FPN inceptionv4  imagenet  6 ./Models/model_fpn_inceptionv4.pth ./ONNX_models/fpn_inceptionv3.onnx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "95ae9325",
   "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": "3.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}