Explore-Train-Validate-Submit.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "fewer-garbage",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-20T20:00:22.587523Z",
     "start_time": "2021-12-20T20:00:22.584843Z"
    }
   },
   "outputs": [],
   "source": [
    "# !pip install \"cloudpathlib[azure]\" loguru tqdm typer\n",
    "# !pip install xarray xarray-spatial"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "worst-paragraph",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-02-06T09:16:46.577898Z",
     "start_time": "2022-02-06T09:16:42.155032Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "32 cpus\n"
     ]
    }
   ],
   "source": [
    "%reload_ext autoreload\n",
    "%autoreload 2\n",
    "\n",
    "# import sys\n",
    "# sys.path.insert(0, '../')\n",
    "import os\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"1\"\n",
    "import pandas as pd\n",
    "pd.set_option(\"display.max_colwidth\", 250)\n",
    "import matplotlib.pyplot as plt\n",
    "import rasterio\n",
    "import glob\n",
    "import numpy as np\n",
    "import pyproj\n",
    "import shutil\n",
    "import bcolz\n",
    "import xarray\n",
    "import cv2\n",
    "import xrspatial.multispectral as ms\n",
    "from PIL import Image\n",
    "import multiprocessing\n",
    "from concurrent.futures import ProcessPoolExecutor, ThreadPoolExecutor\n",
    "cpus = multiprocessing.cpu_count()\n",
    "cpus = min(32, cpus)\n",
    "print(f\"{cpus} cpus\")\n",
    "\n",
    "from src.train import train\n",
    "from src.hyperparams import HyperParams, open_from_yaml\n",
    "from src.dataloader import *\n",
    "from src.datasets import LoadTifDataset\n",
    "from src.validate import *"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "floating-programmer",
   "metadata": {},
   "source": [
    "# Explore"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "timely-membership",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:23.036480Z",
     "start_time": "2021-12-30T19:08:22.974834Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>chip_id</th>\n",
       "      <th>location</th>\n",
       "      <th>datetime</th>\n",
       "      <th>cloudpath</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>adwp</td>\n",
       "      <td>Chifunfu</td>\n",
       "      <td>2020-04-29T08:20:47Z</td>\n",
       "      <td>az://./train_features/adwp</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>adwu</td>\n",
       "      <td>Chifunfu</td>\n",
       "      <td>2020-04-29T08:20:47Z</td>\n",
       "      <td>az://./train_features/adwu</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>adwz</td>\n",
       "      <td>Chifunfu</td>\n",
       "      <td>2020-04-29T08:20:47Z</td>\n",
       "      <td>az://./train_features/adwz</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>adxp</td>\n",
       "      <td>Chifunfu</td>\n",
       "      <td>2020-04-29T08:20:47Z</td>\n",
       "      <td>az://./train_features/adxp</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>aeaj</td>\n",
       "      <td>Chifunfu</td>\n",
       "      <td>2020-04-29T08:20:47Z</td>\n",
       "      <td>az://./train_features/aeaj</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  chip_id  location              datetime                   cloudpath\n",
       "0    adwp  Chifunfu  2020-04-29T08:20:47Z  az://./train_features/adwp\n",
       "1    adwu  Chifunfu  2020-04-29T08:20:47Z  az://./train_features/adwu\n",
       "2    adwz  Chifunfu  2020-04-29T08:20:47Z  az://./train_features/adwz\n",
       "3    adxp  Chifunfu  2020-04-29T08:20:47Z  az://./train_features/adxp\n",
       "4    aeaj  Chifunfu  2020-04-29T08:20:47Z  az://./train_features/aeaj"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_meta = pd.read_csv(\"data/train_metadata.csv\")\n",
    "train_meta.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "useful-packaging",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:23.158284Z",
     "start_time": "2021-12-30T19:08:23.109913Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "chip_id     11748\n",
       "location       81\n",
       "datetime       91\n",
       "dtype: int64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# how many different chip ids, locations, and datetimes are there?\n",
    "train_meta[[\"chip_id\", \"location\", \"datetime\"]].nunique()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "worst-brooks",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:23.334624Z",
     "start_time": "2021-12-30T19:08:23.288669Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Chifunfu      341\n",
       "Timbuktu      321\n",
       "Chingola      321\n",
       "Lusaka        309\n",
       "Bambesa       267\n",
       "             ... \n",
       "Sudan           9\n",
       "Gabon           9\n",
       "Sfax            6\n",
       "Morocco         4\n",
       "Angkor Wat      2\n",
       "Name: location, Length: 81, dtype: int64"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_meta[\"location\"].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "wireless-toddler",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:23.661723Z",
     "start_time": "2021-12-30T19:08:23.421977Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "train_location_counts = (\n",
    "    train_meta.groupby(\"location\")[\"chip_id\"].nunique().sort_values(ascending=False)\n",
    ")\n",
    "plt.figure(figsize=(12, 4))\n",
    "train_location_counts.head(25).plot(kind=\"bar\")\n",
    "plt.xticks(rotation=90)\n",
    "plt.xlabel(\"Location\")\n",
    "plt.ylabel(\"Number of Chips\")\n",
    "plt.title(\"Number of Train Chips by Location (Top 25)\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "cooperative-conducting",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:23.896566Z",
     "start_time": "2021-12-30T19:08:23.662893Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "train_location_counts = (\n",
    "    train_meta.groupby(\"location\")[\"chip_id\"].nunique().sort_values(ascending=False)\n",
    ")\n",
    "plt.figure(figsize=(12, 4))\n",
    "train_location_counts.tail(25).plot(kind=\"bar\")\n",
    "plt.xticks(rotation=90)\n",
    "plt.xlabel(\"Location\")\n",
    "plt.ylabel(\"Number of Chips\")\n",
    "plt.title(\"Number of Train Chips by Location (Top 25)\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "painful-creator",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:23.948308Z",
     "start_time": "2021-12-30T19:08:23.897759Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>chip_count</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>year</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2018</th>\n",
       "      <td>326</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2019</th>\n",
       "      <td>1015</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2020</th>\n",
       "      <td>10407</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      chip_count\n",
       "year            \n",
       "2018         326\n",
       "2019        1015\n",
       "2020       10407"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_meta[\"datetime\"] = pd.to_datetime(train_meta[\"datetime\"])\n",
    "train_meta[\"year\"] = train_meta.datetime.dt.year\n",
    "train_meta.groupby(\"year\")[[\"chip_id\"]].nunique().sort_index().rename(\n",
    "    columns={\"chip_id\": \"chip_count\"}\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "black-stand",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:24.287169Z",
     "start_time": "2021-12-30T19:08:24.242217Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(Timestamp('2018-03-07 08:46:02+0000', tz='UTC'),\n",
       " Timestamp('2020-09-14 08:28:49+0000', tz='UTC'))"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_meta[\"datetime\"].min(), train_meta[\"datetime\"].max()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "invalid-ticket",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:24.800355Z",
     "start_time": "2021-12-30T19:08:24.748502Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th>chip_count</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>location</th>\n",
       "      <th>datetime</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>South America - Brazil</th>\n",
       "      <th>2020-09-06 15:02:37+00:00</th>\n",
       "      <td>261</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Port Gentil</th>\n",
       "      <th>2020-09-08 09:50:58+00:00</th>\n",
       "      <td>223</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Uganda</th>\n",
       "      <th>2019-04-25 08:29:37+00:00</th>\n",
       "      <td>220</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Australia - Central</th>\n",
       "      <th>2020-08-11 01:24:00+00:00</th>\n",
       "      <td>209</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Malabo</th>\n",
       "      <th>2020-09-06 10:00:03+00:00</th>\n",
       "      <td>206</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Jimma</th>\n",
       "      <th>2020-05-31 08:07:58+00:00</th>\n",
       "      <td>201</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Chifunfu</th>\n",
       "      <th>2020-04-29 08:20:47+00:00</th>\n",
       "      <td>197</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>South America - Suriname</th>\n",
       "      <th>2020-06-03 14:11:18+00:00</th>\n",
       "      <td>197</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Isiro</th>\n",
       "      <th>2020-08-28 08:39:29+00:00</th>\n",
       "      <td>197</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Pibor</th>\n",
       "      <th>2020-08-17 08:18:22+00:00</th>\n",
       "      <td>197</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                    chip_count\n",
       "location                 datetime                             \n",
       "South America - Brazil   2020-09-06 15:02:37+00:00         261\n",
       "Port Gentil              2020-09-08 09:50:58+00:00         223\n",
       "Uganda                   2019-04-25 08:29:37+00:00         220\n",
       "Australia - Central      2020-08-11 01:24:00+00:00         209\n",
       "Malabo                   2020-09-06 10:00:03+00:00         206\n",
       "Jimma                    2020-05-31 08:07:58+00:00         201\n",
       "Chifunfu                 2020-04-29 08:20:47+00:00         197\n",
       "South America - Suriname 2020-06-03 14:11:18+00:00         197\n",
       "Isiro                    2020-08-28 08:39:29+00:00         197\n",
       "Pibor                    2020-08-17 08:18:22+00:00         197"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "chips_per_locationtime = (\n",
    "    train_meta.groupby([\"location\", \"datetime\"])[[\"chip_id\"]]\n",
    "    .nunique()\n",
    "    .sort_values(by=\"chip_id\", ascending=False)\n",
    "    .rename(columns={\"chip_id\": \"chip_count\"})\n",
    ")\n",
    "chips_per_locationtime.head(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "flying-lindsay",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:25.817196Z",
     "start_time": "2021-12-30T19:08:25.752215Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>chip_id</th>\n",
       "      <th>location</th>\n",
       "      <th>datetime</th>\n",
       "      <th>cloudpath</th>\n",
       "      <th>year</th>\n",
       "      <th>B02_path</th>\n",
       "      <th>B03_path</th>\n",
       "      <th>B04_path</th>\n",
       "      <th>B08_path</th>\n",
       "      <th>label_path</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>adwp</td>\n",
       "      <td>Chifunfu</td>\n",
       "      <td>2020-04-29 08:20:47+00:00</td>\n",
       "      <td>az://./train_features/adwp</td>\n",
       "      <td>2020</td>\n",
       "      <td>data/train_features/adwp/B02.tif</td>\n",
       "      <td>data/train_features/adwp/B03.tif</td>\n",
       "      <td>data/train_features/adwp/B04.tif</td>\n",
       "      <td>data/train_features/adwp/B08.tif</td>\n",
       "      <td>data/train_labels/adwp.tif</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>adwu</td>\n",
       "      <td>Chifunfu</td>\n",
       "      <td>2020-04-29 08:20:47+00:00</td>\n",
       "      <td>az://./train_features/adwu</td>\n",
       "      <td>2020</td>\n",
       "      <td>data/train_features/adwu/B02.tif</td>\n",
       "      <td>data/train_features/adwu/B03.tif</td>\n",
       "      <td>data/train_features/adwu/B04.tif</td>\n",
       "      <td>data/train_features/adwu/B08.tif</td>\n",
       "      <td>data/train_labels/adwu.tif</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>adwz</td>\n",
       "      <td>Chifunfu</td>\n",
       "      <td>2020-04-29 08:20:47+00:00</td>\n",
       "      <td>az://./train_features/adwz</td>\n",
       "      <td>2020</td>\n",
       "      <td>data/train_features/adwz/B02.tif</td>\n",
       "      <td>data/train_features/adwz/B03.tif</td>\n",
       "      <td>data/train_features/adwz/B04.tif</td>\n",
       "      <td>data/train_features/adwz/B08.tif</td>\n",
       "      <td>data/train_labels/adwz.tif</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>adxp</td>\n",
       "      <td>Chifunfu</td>\n",
       "      <td>2020-04-29 08:20:47+00:00</td>\n",
       "      <td>az://./train_features/adxp</td>\n",
       "      <td>2020</td>\n",
       "      <td>data/train_features/adxp/B02.tif</td>\n",
       "      <td>data/train_features/adxp/B03.tif</td>\n",
       "      <td>data/train_features/adxp/B04.tif</td>\n",
       "      <td>data/train_features/adxp/B08.tif</td>\n",
       "      <td>data/train_labels/adxp.tif</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>aeaj</td>\n",
       "      <td>Chifunfu</td>\n",
       "      <td>2020-04-29 08:20:47+00:00</td>\n",
       "      <td>az://./train_features/aeaj</td>\n",
       "      <td>2020</td>\n",
       "      <td>data/train_features/aeaj/B02.tif</td>\n",
       "      <td>data/train_features/aeaj/B03.tif</td>\n",
       "      <td>data/train_features/aeaj/B04.tif</td>\n",
       "      <td>data/train_features/aeaj/B08.tif</td>\n",
       "      <td>data/train_labels/aeaj.tif</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  chip_id  location                  datetime                   cloudpath  \\\n",
       "0    adwp  Chifunfu 2020-04-29 08:20:47+00:00  az://./train_features/adwp   \n",
       "1    adwu  Chifunfu 2020-04-29 08:20:47+00:00  az://./train_features/adwu   \n",
       "2    adwz  Chifunfu 2020-04-29 08:20:47+00:00  az://./train_features/adwz   \n",
       "3    adxp  Chifunfu 2020-04-29 08:20:47+00:00  az://./train_features/adxp   \n",
       "4    aeaj  Chifunfu 2020-04-29 08:20:47+00:00  az://./train_features/aeaj   \n",
       "\n",
       "   year                          B02_path                          B03_path  \\\n",
       "0  2020  data/train_features/adwp/B02.tif  data/train_features/adwp/B03.tif   \n",
       "1  2020  data/train_features/adwu/B02.tif  data/train_features/adwu/B03.tif   \n",
       "2  2020  data/train_features/adwz/B02.tif  data/train_features/adwz/B03.tif   \n",
       "3  2020  data/train_features/adxp/B02.tif  data/train_features/adxp/B03.tif   \n",
       "4  2020  data/train_features/aeaj/B02.tif  data/train_features/aeaj/B03.tif   \n",
       "\n",
       "                           B04_path                          B08_path  \\\n",
       "0  data/train_features/adwp/B04.tif  data/train_features/adwp/B08.tif   \n",
       "1  data/train_features/adwu/B04.tif  data/train_features/adwu/B08.tif   \n",
       "2  data/train_features/adwz/B04.tif  data/train_features/adwz/B08.tif   \n",
       "3  data/train_features/adxp/B04.tif  data/train_features/adxp/B08.tif   \n",
       "4  data/train_features/aeaj/B04.tif  data/train_features/aeaj/B08.tif   \n",
       "\n",
       "                   label_path  \n",
       "0  data/train_labels/adwp.tif  \n",
       "1  data/train_labels/adwu.tif  \n",
       "2  data/train_labels/adwz.tif  \n",
       "3  data/train_labels/adxp.tif  \n",
       "4  data/train_labels/aeaj.tif  "
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "BANDS = [\"B02\", \"B03\", \"B04\", \"B08\"]\n",
    "\n",
    "def add_paths(df, feature_dir, label_dir=None, bands=BANDS):\n",
    "    \"\"\"\n",
    "    Given dataframe with a column for chip_id, returns a dataframe with a column\n",
    "    added indicating the path to each band's TIF image as \"{band}_path\", eg \"B02_path\".\n",
    "    A column is also added to the dataframe with paths to the label TIF, if the\n",
    "    path to the labels directory is provided.\n",
    "    \"\"\"\n",
    "    for band in bands:\n",
    "        df[f\"{band}_path\"] = df[\"chip_id\"].apply(lambda x: f\"{feature_dir}/{x}/{band}.tif\")\n",
    "#         assert df[f\"{band}_path\"].path.exists().all()\n",
    "    if label_dir is not None:\n",
    "        df[\"label_path\"] = df[\"chip_id\"].apply(lambda x: f\"{label_dir}/{x}.tif\")\n",
    "#         assert df[\"label_path\"].path.exists().all()\n",
    "\n",
    "    return df\n",
    "\n",
    "\n",
    "train_meta = add_paths(train_meta, \"data/train_features\", \"data/train_labels\")\n",
    "train_meta.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "distinct-occupation",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:29.100034Z",
     "start_time": "2021-12-30T19:08:28.848090Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(11748, 46992)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(glob.glob(\"data/train_features/*\")), len(glob.glob(\"data/train_features/*/*\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "taken-defensive",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:29.182500Z",
     "start_time": "2021-12-30T19:08:29.112159Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "12798"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(glob.glob(\"data/train_labels/*\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "banner-belief",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:29.406895Z",
     "start_time": "2021-12-30T19:08:29.360783Z"
    }
   },
   "outputs": [],
   "source": [
    "example_chip = train_meta.sample(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "flexible-thursday",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:29.668930Z",
     "start_time": "2021-12-30T19:08:29.569197Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'driver': 'GTiff',\n",
       " 'dtype': 'uint16',\n",
       " 'nodata': 0.0,\n",
       " 'width': 512,\n",
       " 'height': 512,\n",
       " 'count': 1,\n",
       " 'crs': CRS.from_epsg(32722),\n",
       " 'transform': Affine(10.0, 0.0, 340740.0,\n",
       "        0.0, -10.0, 6742130.0)}"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "example_chip = example_chip.iloc[0]\n",
    "with rasterio.open(example_chip[\"B04_path\"]) as img:\n",
    "    chip_metadata = img.meta\n",
    "    img_array = img.read(1)\n",
    "\n",
    "chip_metadata"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "absolute-tolerance",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:29.861833Z",
     "start_time": "2021-12-30T19:08:29.813665Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Image array shape: (512, 512)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[3724, 3532, 3216, ..., 1148, 1102, 1032],\n",
       "       [3630, 3362, 2928, ..., 1322, 1164, 1062],\n",
       "       [3496, 3000, 2702, ..., 1460, 1278, 1138],\n",
       "       ...,\n",
       "       [5092, 5748, 5956, ..., 1198, 1226, 1224],\n",
       "       [4912, 5724, 5988, ..., 1212, 1226, 1208],\n",
       "       [4792, 5556, 5816, ..., 1176, 1216, 1246]], dtype=uint16)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# what does the image array look like?\n",
    "print(\"Image array shape:\", img_array.shape)\n",
    "img_array"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "coupled-declaration",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:30.217183Z",
     "start_time": "2021-12-30T19:08:30.171796Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.isnan(img_array).sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "hydraulic-nigeria",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:31.280298Z",
     "start_time": "2021-12-30T19:08:31.119999Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(img_array)\n",
    "plt.title(f\"B04 band for chip id {example_chip.chip_id}\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "southwest-horizon",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:31.603677Z",
     "start_time": "2021-12-30T19:08:31.558486Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Longitude: -52.615982199160364, latitude: -29.46353304963184\n"
     ]
    }
   ],
   "source": [
    "# longitude/latitude of image's center\n",
    "with rasterio.open(example_chip[\"B04_path\"]) as img:\n",
    "    lon, lat = img.lnglat()\n",
    "    bounds = img.bounds\n",
    "print(f\"Longitude: {lon}, latitude: {lat}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "attended-mystery",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:31.994404Z",
     "start_time": "2021-12-30T19:08:31.950390Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "BoundingBox(left=340740.0, bottom=6737010.0, right=345860.0, top=6742130.0)"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bounds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "perfect-fields",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:32.427671Z",
     "start_time": "2021-12-30T19:08:32.308573Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Image coordinates (lat, long):\n",
      "Start: (-29.486946830796455, -52.6427458966144)\n",
      "End: (-29.44011455494337, -52.589230450963804)\n"
     ]
    }
   ],
   "source": [
    "def lat_long_bounds(filepath):\n",
    "    \"\"\"Given the path to a GeoTIFF, returns the image bounds in latitude and\n",
    "    longitude coordinates.\n",
    "\n",
    "    Returns points as a tuple of (left, bottom, right, top)\n",
    "    \"\"\"\n",
    "    with rasterio.open(filepath) as im:\n",
    "        bounds = im.bounds\n",
    "        meta = im.meta\n",
    "    # create a converter starting with the current projection\n",
    "    current_crs = pyproj.CRS(meta[\"crs\"])\n",
    "    crs_transform = pyproj.Transformer.from_crs(current_crs, current_crs.geodetic_crs)\n",
    "\n",
    "    # returns left, bottom, right, top\n",
    "    return crs_transform.transform_bounds(*bounds)\n",
    "left, bottom, right, top = lat_long_bounds(example_chip[\"B04_path\"])\n",
    "print(\n",
    "    f\"Image coordinates (lat, long):\\nStart: ({left}, {bottom})\"\n",
    "    f\"\\nEnd: ({right}, {top})\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "desperate-chorus",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:32.954025Z",
     "start_time": "2021-12-30T19:08:32.904681Z"
    }
   },
   "outputs": [],
   "source": [
    "def get_xarray(filepath):\n",
    "    \"\"\"Put images in xarray.DataArray format\"\"\"\n",
    "    im_arr = np.array(Image.open(filepath))\n",
    "    return xarray.DataArray(im_arr, dims=[\"y\", \"x\"])\n",
    "\n",
    "\n",
    "def true_color_img(chip_id, data_dir=\"data/train_features\"):\n",
    "    \"\"\"Given the path to the directory of Sentinel-2 chip feature images,\n",
    "    plots the true color image\"\"\"\n",
    "    chip_dir = f\"{data_dir}/{chip_id}\"\n",
    "    red = get_xarray(f\"{chip_dir}/B04.tif\")\n",
    "    green = get_xarray(f\"{chip_dir}/B03.tif\")\n",
    "    blue = get_xarray(f\"{chip_dir}/B02.tif\")\n",
    "\n",
    "    return ms.true_color(r=red, g=green, b=blue)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "behavioral-offset",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:33.866421Z",
     "start_time": "2021-12-30T19:08:33.427441Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=(5, 5))\n",
    "im = true_color_img(example_chip.chip_id)\n",
    "ax.imshow(im)\n",
    "plt.title(f\"True color image for chip id {example_chip.chip_id}\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "chinese-forty",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:34.094611Z",
     "start_time": "2021-12-30T19:08:34.049632Z"
    }
   },
   "outputs": [],
   "source": [
    "def display_random_chip(random_state):\n",
    "    fig, ax = plt.subplots(1, 2, figsize=(8, 4))\n",
    "    random_chip = train_meta.sample(random_state=random_state).iloc[0]\n",
    "    ax[0].imshow(true_color_img(random_chip.chip_id))\n",
    "    ax[0].set_title(f\"Chip {random_chip.chip_id}\\n(Location: {random_chip.location})\")\n",
    "    label_im = np.array(Image.open(random_chip.label_path))\n",
    "    ax[1].imshow(label_im, cmap=\"gray\")\n",
    "    ax[1].set_title(f\"Chip {random_chip.chip_id} label, {100 * (label_im.sum()/512**2):.1f}% clouds\")\n",
    "\n",
    "    plt.tight_layout()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "acoustic-wyoming",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:34.796263Z",
     "start_time": "2021-12-30T19:08:34.752266Z"
    },
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "# for k in range(100):\n",
    "#     display_random_chip(k)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "authentic-southwest",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:37.493222Z",
     "start_time": "2021-12-30T19:08:37.425998Z"
    }
   },
   "outputs": [],
   "source": [
    "train_meta = pd.read_csv(\"data/train_metadata.csv\")\n",
    "bands = [\"B02\", \"B03\", \"B04\", \"B08\"]\n",
    "feature_dir = \"data/train_features\"\n",
    "label_dir = \"data/train_labels\"\n",
    "\n",
    "for band in bands:\n",
    "    train_meta[f\"{band}_path\"] = train_meta[\"chip_id\"].apply(lambda x: f\"{feature_dir}/{x}/{band}.tif\")\n",
    "if label_dir is not None:\n",
    "    train_meta[\"label_path\"] = train_meta[\"chip_id\"].apply(lambda x: f\"{label_dir}/{x}.tif\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "honest-budget",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:40.289599Z",
     "start_time": "2021-12-30T19:08:40.243987Z"
    }
   },
   "outputs": [],
   "source": [
    "folds = [0, 1, 2, 3, 4] * (len(train_meta) // 5 + 1)\n",
    "folds = folds[:len(train_meta)]\n",
    "\n",
    "seed = 1337\n",
    "np.random.seed(seed)\n",
    "np.random.shuffle(folds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "automotive-consciousness",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:42.434405Z",
     "start_time": "2021-12-30T19:08:42.379667Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>chip_id</th>\n",
       "      <th>location</th>\n",
       "      <th>datetime</th>\n",
       "      <th>cloudpath</th>\n",
       "      <th>B02_path</th>\n",
       "      <th>B03_path</th>\n",
       "      <th>B04_path</th>\n",
       "      <th>B08_path</th>\n",
       "      <th>label_path</th>\n",
       "      <th>fold</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>4428</th>\n",
       "      <td>jxti</td>\n",
       "      <td>Australia - Central West</td>\n",
       "      <td>2020-09-02T02:06:04Z</td>\n",
       "      <td>az://./train_features/jxti</td>\n",
       "      <td>data/train_features/jxti/B02.tif</td>\n",
       "      <td>data/train_features/jxti/B03.tif</td>\n",
       "      <td>data/train_features/jxti/B04.tif</td>\n",
       "      <td>data/train_features/jxti/B08.tif</td>\n",
       "      <td>data/train_labels/jxti.tif</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2760</th>\n",
       "      <td>ewyz</td>\n",
       "      <td>Ethiopia</td>\n",
       "      <td>2020-09-12T08:39:12Z</td>\n",
       "      <td>az://./train_features/ewyz</td>\n",
       "      <td>data/train_features/ewyz/B02.tif</td>\n",
       "      <td>data/train_features/ewyz/B03.tif</td>\n",
       "      <td>data/train_features/ewyz/B04.tif</td>\n",
       "      <td>data/train_features/ewyz/B08.tif</td>\n",
       "      <td>data/train_labels/ewyz.tif</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10558</th>\n",
       "      <td>xrft</td>\n",
       "      <td>Malemba Nkulu</td>\n",
       "      <td>2020-09-07T08:42:26Z</td>\n",
       "      <td>az://./train_features/xrft</td>\n",
       "      <td>data/train_features/xrft/B02.tif</td>\n",
       "      <td>data/train_features/xrft/B03.tif</td>\n",
       "      <td>data/train_features/xrft/B04.tif</td>\n",
       "      <td>data/train_features/xrft/B08.tif</td>\n",
       "      <td>data/train_labels/xrft.tif</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7317</th>\n",
       "      <td>rcbj</td>\n",
       "      <td>Lusaka</td>\n",
       "      <td>2020-08-22T08:24:41Z</td>\n",
       "      <td>az://./train_features/rcbj</td>\n",
       "      <td>data/train_features/rcbj/B02.tif</td>\n",
       "      <td>data/train_features/rcbj/B03.tif</td>\n",
       "      <td>data/train_features/rcbj/B04.tif</td>\n",
       "      <td>data/train_features/rcbj/B08.tif</td>\n",
       "      <td>data/train_labels/rcbj.tif</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6783</th>\n",
       "      <td>pibw</td>\n",
       "      <td>Canary Islands</td>\n",
       "      <td>2020-05-15T11:53:46Z</td>\n",
       "      <td>az://./train_features/pibw</td>\n",
       "      <td>data/train_features/pibw/B02.tif</td>\n",
       "      <td>data/train_features/pibw/B03.tif</td>\n",
       "      <td>data/train_features/pibw/B04.tif</td>\n",
       "      <td>data/train_features/pibw/B08.tif</td>\n",
       "      <td>data/train_labels/pibw.tif</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      chip_id                  location              datetime  \\\n",
       "4428     jxti  Australia - Central West  2020-09-02T02:06:04Z   \n",
       "2760     ewyz                  Ethiopia  2020-09-12T08:39:12Z   \n",
       "10558    xrft             Malemba Nkulu  2020-09-07T08:42:26Z   \n",
       "7317     rcbj                    Lusaka  2020-08-22T08:24:41Z   \n",
       "6783     pibw            Canary Islands  2020-05-15T11:53:46Z   \n",
       "\n",
       "                        cloudpath                          B02_path  \\\n",
       "4428   az://./train_features/jxti  data/train_features/jxti/B02.tif   \n",
       "2760   az://./train_features/ewyz  data/train_features/ewyz/B02.tif   \n",
       "10558  az://./train_features/xrft  data/train_features/xrft/B02.tif   \n",
       "7317   az://./train_features/rcbj  data/train_features/rcbj/B02.tif   \n",
       "6783   az://./train_features/pibw  data/train_features/pibw/B02.tif   \n",
       "\n",
       "                               B03_path                          B04_path  \\\n",
       "4428   data/train_features/jxti/B03.tif  data/train_features/jxti/B04.tif   \n",
       "2760   data/train_features/ewyz/B03.tif  data/train_features/ewyz/B04.tif   \n",
       "10558  data/train_features/xrft/B03.tif  data/train_features/xrft/B04.tif   \n",
       "7317   data/train_features/rcbj/B03.tif  data/train_features/rcbj/B04.tif   \n",
       "6783   data/train_features/pibw/B03.tif  data/train_features/pibw/B04.tif   \n",
       "\n",
       "                               B08_path                  label_path  fold  \n",
       "4428   data/train_features/jxti/B08.tif  data/train_labels/jxti.tif     1  \n",
       "2760   data/train_features/ewyz/B08.tif  data/train_labels/ewyz.tif     1  \n",
       "10558  data/train_features/xrft/B08.tif  data/train_labels/xrft.tif     3  \n",
       "7317   data/train_features/rcbj/B08.tif  data/train_labels/rcbj.tif     1  \n",
       "6783   data/train_features/pibw/B08.tif  data/train_labels/pibw.tif     1  "
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_meta[\"fold\"] = folds\n",
    "train_meta.sample(5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "touched-cookbook",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:43.099518Z",
     "start_time": "2021-12-30T19:08:43.022015Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(11748, 11)"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_meta[\"path_exists\"] = train_meta[\"B02_path\"].apply(lambda x: os.path.exists(x))\n",
    "train_meta = train_meta[train_meta[\"path_exists\"]==True]\n",
    "train_meta.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "union-artist",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2021-12-30T19:08:48.263984Z",
     "start_time": "2021-12-30T19:08:48.210938Z"
    }
   },
   "outputs": [],
   "source": [
    "train_meta.to_pickle(\"data/trainval_2021_12_30.pkl\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "unavailable-milan",
   "metadata": {},
   "source": [
    "# Resize additional bands"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "center-example",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-18T14:18:46.807793Z",
     "start_time": "2022-01-18T14:18:46.555071Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True    11748\n",
      "Name: B01_path_exists, dtype: int64\n",
      "True    11748\n",
      "Name: B09_path_exists, dtype: int64\n",
      "True    11748\n",
      "Name: B11_path_exists, dtype: int64\n",
      "True    11748\n",
      "Name: B12_path_exists, dtype: int64\n"
     ]
    }
   ],
   "source": [
    "df = pd.read_csv(\"data/train_metadata.csv\")\n",
    "new_bands = ['B01', 'B09', 'B11', 'B12']\n",
    "feature_dir = \"data/train_features\"\n",
    "label_dir = \"data/train_labels\"\n",
    "\n",
    "for band in new_bands:\n",
    "    df[f\"{band}_path\"] = df[\"chip_id\"].apply(lambda x: f\"{feature_dir}/{x}/{band}.tif\")\n",
    "    df[f\"{band}_path_exists\"] = df[f\"{band}_path\"].apply(lambda x: os.path.exists(x))\n",
    "    print(df[f\"{band}_path_exists\"].value_counts())\n",
    "    \n",
    "paths_b1 = df[\"B01_path\"].tolist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "spectacular-lover",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-18T14:17:47.087424Z",
     "start_time": "2022-01-18T14:17:46.993324Z"
    }
   },
   "outputs": [],
   "source": [
    "visualize = False\n",
    "\n",
    "def resize_and_save(idx):\n",
    "    if idx%1000==999: print(idx)\n",
    "        \n",
    "    path_b1 = paths_b1[idx]\n",
    "    resized_b1_output_path = path_b1.replace(\".tif\", \"_resized512.tif\")\n",
    "    for band in new_bands:\n",
    "        output_path = resized_b1_output_path.replace(\"B01_resized512.tif\", f\"{band}_resized512.tif\")\n",
    "#         if os.path.exists(output_path):\n",
    "#             continue\n",
    "        input_path = path_b1.replace(\"B01.tif\", f\"{band}.tif\")\n",
    "        img = tifffile.imread(input_path)\n",
    "        if visualize:\n",
    "            plt.imshow(img)\n",
    "            plt.show()\n",
    "            print(img.shape)\n",
    "        img = cv2.resize(img, (512, 512), interpolation=cv2.INTER_CUBIC)\n",
    "        if visualize:\n",
    "            print(img.shape)\n",
    "            plt.imshow(img)\n",
    "            plt.show()\n",
    "        \n",
    "        if input_path != output_path:\n",
    "            tifffile.imwrite(output_path, img)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "continuous-render",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-18T14:14:29.633843Z",
     "start_time": "2022-01-18T14:14:29.548215Z"
    },
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "resize_and_save(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "thrown-wrapping",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-18T14:17:59.874982Z",
     "start_time": "2022-01-18T14:17:47.824839Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "999\n",
      "1999\n",
      "2999\n",
      "3999\n",
      "4999\n",
      "5999\n",
      "6999\n",
      "7999\n",
      "8999\n",
      "9999\n",
      "10999\n"
     ]
    }
   ],
   "source": [
    "with ProcessPoolExecutor(cpus) as e: \n",
    "    e.map(resize_and_save, list(range(len(paths_b1))))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "textile-brazil",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-18T14:18:40.611132Z",
     "start_time": "2022-01-18T14:18:40.359667Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True    11748\n",
      "Name: B01_path_exists, dtype: int64\n",
      "True    11748\n",
      "Name: B09_path_exists, dtype: int64\n",
      "True    11748\n",
      "Name: B11_path_exists, dtype: int64\n",
      "True    11748\n",
      "Name: B12_path_exists, dtype: int64\n"
     ]
    }
   ],
   "source": [
    "df = pd.read_csv(\"data/train_metadata.csv\")\n",
    "new_bands = ['B01', 'B09', 'B11', 'B12']\n",
    "feature_dir = \"data/train_features\"\n",
    "label_dir = \"data/train_labels\"\n",
    "\n",
    "for band in new_bands:\n",
    "    df[f\"{band}_path\"] = df[\"chip_id\"].apply(lambda x: f\"{feature_dir}/{x}/{band}_resized512.tif\")\n",
    "    df[f\"{band}_path_exists\"] = df[f\"{band}_path\"].apply(lambda x: os.path.exists(x))\n",
    "    print(df[f\"{band}_path_exists\"].value_counts())"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "worldwide-tobacco",
   "metadata": {},
   "source": [
    "# Split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "id": "conceptual-vault",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-27T21:13:32.381263Z",
     "start_time": "2022-01-27T21:13:31.909641Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True    11748\n",
      "Name: B02_path_exists, dtype: int64\n",
      "True    11748\n",
      "Name: B03_path_exists, dtype: int64\n",
      "True    11748\n",
      "Name: B04_path_exists, dtype: int64\n",
      "True    11748\n",
      "Name: B08_path_exists, dtype: int64\n",
      "True    11748\n",
      "Name: B01_path_exists, dtype: int64\n",
      "True    11748\n",
      "Name: B09_path_exists, dtype: int64\n",
      "True    11748\n",
      "Name: B11_path_exists, dtype: int64\n",
      "True    11748\n",
      "Name: B12_path_exists, dtype: int64\n"
     ]
    }
   ],
   "source": [
    "df = pd.read_csv(\"data/train_metadata.csv\")\n",
    "bands = [\"B02\", \"B03\", \"B04\", \"B08\"]\n",
    "new_bands = ['B01', 'B09', 'B11', 'B12']\n",
    "feature_dir = \"data/train_features\"\n",
    "label_dir = \"data/train_labels\"\n",
    "\n",
    "for band in bands:\n",
    "    df[f\"{band}_path\"] = df[\"chip_id\"].apply(lambda x: f\"{feature_dir}/{x}/{band}.tif\")\n",
    "    df[f\"{band}_path_exists\"] = df[f\"{band}_path\"].apply(lambda x: os.path.exists(x))\n",
    "    print(df[f\"{band}_path_exists\"].value_counts())\n",
    "    \n",
    "for band in new_bands:\n",
    "    df[f\"{band}_path\"] = df[\"chip_id\"].apply(lambda x: f\"{feature_dir}/{x}/{band}_resized512.tif\")\n",
    "    df[f\"{band}_path_exists\"] = df[f\"{band}_path\"].apply(lambda x: os.path.exists(x))\n",
    "    print(df[f\"{band}_path_exists\"].value_counts())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "id": "fourth-reducing",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-27T21:13:38.885862Z",
     "start_time": "2022-01-27T21:13:38.718958Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "11748"
      ]
     },
     "execution_count": 84,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[\"label_path\"] = df[\"chip_id\"].apply(lambda x: f\"{label_dir}/{x}.tif\")\n",
    "\n",
    "folds = [0, 1, 2, 3, 4] * (len(df) // 5 + 1)\n",
    "folds = folds[:len(df)]\n",
    "\n",
    "seed = 11111\n",
    "np.random.seed(seed)\n",
    "np.random.shuffle(folds)\n",
    "df[\"fold\"] = folds\n",
    "\n",
    "df[\"label_path\"] = df[\"label_path\"].apply(lambda x: x.replace(\".tif\", \"_improved_weakly.tif\") if os.path.exists(x.replace(\".tif\", \"_improved_weakly.tif\")) else x)\n",
    "df[\"label_path\"] = df[\"label_path\"].apply(lambda x: x.replace(\".tif\", \"_improved.tif\") if os.path.exists(x.replace(\".tif\", \"_improved.tif\")) else x)\n",
    "df[\"label_path\"].apply(lambda x: os.path.exists(x)).sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "id": "attempted-december",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-27T21:13:49.068628Z",
     "start_time": "2022-01-27T21:13:49.005913Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5322     data/train_labels/mbhf_improved_weakly.tif\n",
       "7283     data/train_labels/razr_improved_weakly.tif\n",
       "443      data/train_labels/aqek_improved_weakly.tif\n",
       "7988            data/train_labels/snal_improved.tif\n",
       "11318    data/train_labels/yuig_improved_weakly.tif\n",
       "2790     data/train_labels/exvn_improved_weakly.tif\n",
       "11074    data/train_labels/yknh_improved_weakly.tif\n",
       "11535    data/train_labels/zaub_improved_weakly.tif\n",
       "4968                     data/train_labels/lpvc.tif\n",
       "3093     data/train_labels/ghcp_improved_weakly.tif\n",
       "Name: label_path, dtype: object"
      ]
     },
     "execution_count": 92,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[\"label_path\"].sample(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "id": "burning-profile",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-27T21:13:52.380125Z",
     "start_time": "2022-01-27T21:13:52.317614Z"
    }
   },
   "outputs": [],
   "source": [
    "counts = df[\"location\"].value_counts().reset_index()\n",
    "# counts"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "id": "liked-there",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-27T21:13:52.644922Z",
     "start_time": "2022-01-27T21:13:52.480682Z"
    },
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Setting Chifunfu                       to 1.0\n",
      "Setting Timbuktu                       to 1.0\n",
      "Setting Chingola                       to 1.0\n",
      "Setting Lusaka                         to 1.0\n",
      "Setting Bambesa                        to 1.0\n",
      "Setting South America - Brazil         to 1.0\n",
      "Setting Canary Islands                 to 1.0\n",
      "Setting Port Gentil                    to 1.0\n",
      "Setting Uganda                         to 1.0\n",
      "Setting Australia - Central            to 1.0\n",
      "Setting Malabo                         to 1.0\n",
      "Setting Jimma                          to 1.0\n",
      "Setting Pibor                          to 1.0\n",
      "Setting South America - Suriname       to 1.0\n",
      "Setting Isiro                          to 1.0\n",
      "Setting San Lorenzo                    to 1.0\n",
      "Setting Eswatini                       to 1.0\n",
      "Setting Riobamba                       to 1.0\n",
      "Setting South America - Peru           to 1.0\n",
      "Setting Porto Velho                    to 1.0\n",
      "Setting Australia - North West         to 1.0\n",
      "Setting Matadi                         to 1.0\n",
      "Setting Paijan                         to 1.0\n",
      "Setting Lodwar                         to 1.0\n",
      "Setting Georgetown                     to 1.0\n",
      "Setting Macapa                         to 1.0\n",
      "Setting Para                           to 1.0\n",
      "Setting Manaus                         to 1.0\n",
      "Setting Bahir Dar                      to 1.0\n",
      "Setting Ethiopia                       to 1.0\n",
      "Setting Eritrea                        to 1.0\n",
      "Setting Nairobi                        to 1.0\n",
      "Setting Malemba Nkulu                  to 1.0\n",
      "Setting Bor                            to 1.0\n",
      "Setting Kolwezi                        to 1.0\n",
      "Setting Harare                         to 1.0\n",
      "Setting Ghana                          to 1.0\n",
      "Setting Mahajanga                      to 1.0\n",
      "Setting Madagascar                     to 1.0\n",
      "Setting Senanga                        to 1.0\n",
      "Setting Juba                           to 1.0\n",
      "Setting Monongue                       to 1.0\n",
      "Setting Santa Cruz de la Sierra        to 1.0\n",
      "Setting Kalamie                        to 1.0\n",
      "Setting Bechar                         to 1.0\n",
      "Setting Vitoria                        to 1.0\n",
      "Setting Cabo Verdo                     to 1.0\n",
      "Setting Beira                          to 1.0\n",
      "Setting Kimberley                      to 1.0\n",
      "Setting Chibemba                       to 1.0\n",
      "Setting Alexander Bay                  to 1.0\n",
      "Setting Addis Ababa                    to 1.0\n",
      "Setting Tunis                          to 1.0\n",
      "Setting Australia - Central West       to 1.0\n",
      "Setting Rio Branco                     to 1.0\n",
      "Setting Santa Fe                       to 1.0\n",
      "Setting Port Augusta                   to 1.0\n",
      "Setting Santa Cruz do Sul              to 1.0\n",
      "Setting Rustenburg                     to 1.0\n",
      "Setting Bunbury                        to 1.0\n",
      "Setting Asmara nort                    to 1.0\n",
      "Setting Adelaide                       to 1.0\n",
      "Setting San Sabastian de La Gomera     to 1.0\n",
      "Setting Cordoba                        to 1.0\n",
      "Setting Adi Ramets                     to 1.0\n",
      "Setting Launceston                     to 1.0\n",
      "Setting South America - Argentina      to 1.0\n",
      "Setting South Africa                   to 1.0\n",
      "Setting Angola                         to 1.0\n",
      "Setting DRC                            to 1.0\n",
      "Setting Paraguay                       to 1.0\n",
      "Setting Malawi                         to 1.0\n",
      "Setting Sierra Leone                   to 1.0\n",
      "Setting Kenya                          to 1.0\n",
      "Setting Ecuador                        to 1.0\n",
      "Setting Alto Hama                      to 1.0\n",
      "Setting Sudan                          to 1.0\n",
      "Setting Gabon                          to 1.0\n",
      "Setting Sfax                           to 1.0\n",
      "Setting Morocco                        to 1.0\n",
      "Setting Angkor Wat                     to 1.0\n"
     ]
    }
   ],
   "source": [
    "df[\"weight\"] = 1\n",
    "power = 0.0\n",
    "max_ = counts[\"location\"].max()\n",
    "for idx, row in counts.iterrows():\n",
    "    df.loc[df[\"location\"]==row[\"index\"], \"weight\"] *= (max_ / row[\"location\"]) ** power\n",
    "    print(f\"Setting {row['index']:30} to {(max_ / row['location']) ** power:.1f}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "id": "bottom-butterfly",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-27T21:13:58.367260Z",
     "start_time": "2022-01-27T21:13:58.287872Z"
    },
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "df.to_pickle(\"data/trainval_2022_01_27.pkl\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "neural-sociology",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-02-05T22:12:51.981152Z",
     "start_time": "2022-02-05T22:12:51.899845Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "11735"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[\"label_path\"].apply(lambda x: os.path.exists(x)).sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "endless-toilet",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-02-05T22:12:54.054796Z",
     "start_time": "2022-02-05T22:12:53.967341Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "11748"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[\"label_path\"] = df[\"label_path\"].apply(lambda x: x if os.path.exists(x) else x.replace(\"_improved.tif\", \".tif\"))\n",
    "df[\"label_path\"].apply(lambda x: os.path.exists(x)).sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "complete-nickname",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-02-05T22:13:09.213312Z",
     "start_time": "2022-02-05T22:13:09.178354Z"
    }
   },
   "outputs": [],
   "source": [
    "df.to_pickle(\"data/trainval_2022_02_05.pkl\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "vulnerable-clinic",
   "metadata": {},
   "source": [
    "# Train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "contrary-forward",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-02-06T05:44:16.942884Z",
     "start_time": "2022-02-05T23:24:01.734848Z"
    },
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "05-Feb-22 23:24:01 - Training with 1 GPUS\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Data Augmentations: [RandomCrop(always_apply=False, p=1.0, height=384, width=384)]\n",
      "Train Image Data Augmentations: []\n",
      "####################################################################################################\n",
      "####################################################################################################\n",
      "Fold 0 --> Train: 9398, Val: 2350\n",
      "####################################################################################################\n",
      "Last 5 sample weights: [1.0, 1.0, 1.0, 1.0, 1.0]\n",
      "unet timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "05-Feb-22 23:24:03 - Training is done with mixed precision\n",
      "05-Feb-22 23:25:16 - Ep: [1] TotalT: 1.2 min, BatchT: 0.292s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.1407\n",
      "05-Feb-22 23:25:30 - Ep: [1]  ValT: 0.23 min, BatchT: 0.183s, DataT: 0.015s, Loss: 0.0553, IoU: 0.9376 (val)\n",
      "05-Feb-22 23:26:38 - Ep: [2] TotalT: 2.6 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0498\n",
      "05-Feb-22 23:26:48 - Ep: [2]  ValT: 0.17 min, BatchT: 0.141s, DataT: 0.015s, Loss: 0.0373, IoU: 0.9478 (val)\n",
      "05-Feb-22 23:27:57 - Ep: [3] TotalT: 3.9 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0373\n",
      "05-Feb-22 23:28:06 - Ep: [3]  ValT: 0.16 min, BatchT: 0.133s, DataT: 0.012s, Loss: 0.0298, IoU: 0.9542 (val)\n",
      "05-Feb-22 23:29:15 - Ep: [4] TotalT: 5.2 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0321\n",
      "05-Feb-22 23:29:25 - Ep: [4]  ValT: 0.16 min, BatchT: 0.133s, DataT: 0.013s, Loss: 0.0268, IoU: 0.9572 (val)\n",
      "05-Feb-22 23:30:32 - Ep: [5] TotalT: 6.5 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0286\n",
      "05-Feb-22 23:30:43 - Ep: [5]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.015s, Loss: 0.0251, IoU: 0.9583 (val)\n",
      "05-Feb-22 23:31:51 - Ep: [6] TotalT: 7.8 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0249\n",
      "05-Feb-22 23:32:01 - Ep: [6]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0233, IoU: 0.9609 (val)\n",
      "05-Feb-22 23:33:09 - Ep: [7] TotalT: 9.1 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0238\n",
      "05-Feb-22 23:33:19 - Ep: [7]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.016s, Loss: 0.0225, IoU: 0.9617 (val)\n",
      "05-Feb-22 23:34:27 - Ep: [8] TotalT: 10.4 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0222\n",
      "05-Feb-22 23:34:38 - Ep: [8]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.014s, Loss: 0.0213, IoU: 0.9630 (val)\n",
      "05-Feb-22 23:35:47 - Ep: [9] TotalT: 11.7 min, BatchT: 0.273s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0208\n",
      "05-Feb-22 23:35:57 - Ep: [9]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0199, IoU: 0.9652 (val)\n",
      "05-Feb-22 23:37:05 - Ep: [10] TotalT: 13.0 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0202\n",
      "05-Feb-22 23:37:16 - Ep: [10]  ValT: 0.18 min, BatchT: 0.146s, DataT: 0.016s, Loss: 0.0197, IoU: 0.9655 (val)\n",
      "05-Feb-22 23:38:25 - Ep: [11] TotalT: 14.4 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0194\n",
      "05-Feb-22 23:38:35 - Ep: [11]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.015s, Loss: 0.0194, IoU: 0.9659 (val)\n",
      "05-Feb-22 23:39:44 - Ep: [12] TotalT: 15.7 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0186\n",
      "05-Feb-22 23:39:54 - Ep: [12]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.013s, Loss: 0.0188, IoU: 0.9667 (val)\n",
      "05-Feb-22 23:41:03 - Ep: [13] TotalT: 17.0 min, BatchT: 0.274s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0179\n",
      "05-Feb-22 23:41:14 - Ep: [13]  ValT: 0.18 min, BatchT: 0.143s, DataT: 0.015s, Loss: 0.0187, IoU: 0.9669 (val)\n",
      "05-Feb-22 23:42:22 - Ep: [14] TotalT: 18.3 min, BatchT: 0.273s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0178\n",
      "05-Feb-22 23:42:33 - Ep: [14]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.015s, Loss: 0.0174, IoU: 0.9689 (val)\n",
      "05-Feb-22 23:43:41 - Ep: [15] TotalT: 19.6 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0180\n",
      "05-Feb-22 23:43:51 - Ep: [15]  ValT: 0.17 min, BatchT: 0.134s, DataT: 0.012s, Loss: 0.0186, IoU: 0.9665 (val)\n",
      "05-Feb-22 23:44:59 - Ep: [16] TotalT: 20.9 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0173\n",
      "05-Feb-22 23:45:10 - Ep: [16]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.014s, Loss: 0.0179, IoU: 0.9678 (val)\n",
      "05-Feb-22 23:46:18 - Ep: [17] TotalT: 22.3 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0176\n",
      "05-Feb-22 23:46:28 - Ep: [17]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.015s, Loss: 0.0195, IoU: 0.9649 (val)\n",
      "05-Feb-22 23:47:36 - Ep: [18] TotalT: 23.6 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0171\n",
      "05-Feb-22 23:47:46 - Ep: [18]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.013s, Loss: 0.0164, IoU: 0.9706 (val)\n",
      "05-Feb-22 23:48:55 - Ep: [19] TotalT: 24.9 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0158\n",
      "05-Feb-22 23:49:05 - Ep: [19]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.014s, Loss: 0.0162, IoU: 0.9708 (val)\n",
      "05-Feb-22 23:50:13 - Ep: [20] TotalT: 26.2 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0165\n",
      "05-Feb-22 23:50:23 - Ep: [20]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.015s, Loss: 0.0162, IoU: 0.9708 (val)\n",
      "05-Feb-22 23:51:31 - Ep: [21] TotalT: 27.5 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0153\n",
      "05-Feb-22 23:51:41 - Ep: [21]  ValT: 0.16 min, BatchT: 0.132s, DataT: 0.012s, Loss: 0.0173, IoU: 0.9689 (val)\n",
      "05-Feb-22 23:52:49 - Ep: [22] TotalT: 28.8 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0149\n",
      "05-Feb-22 23:53:00 - Ep: [22]  ValT: 0.18 min, BatchT: 0.146s, DataT: 0.017s, Loss: 0.0151, IoU: 0.9727 (val)\n",
      "05-Feb-22 23:54:09 - Ep: [23] TotalT: 30.1 min, BatchT: 0.274s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0149\n",
      "05-Feb-22 23:54:19 - Ep: [23]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.012s, Loss: 0.0156, IoU: 0.9715 (val)\n",
      "05-Feb-22 23:55:27 - Ep: [24] TotalT: 31.4 min, BatchT: 0.272s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0145\n",
      "05-Feb-22 23:55:37 - Ep: [24]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0154, IoU: 0.9720 (val)\n",
      "05-Feb-22 23:56:46 - Ep: [25] TotalT: 32.7 min, BatchT: 0.272s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0148\n",
      "05-Feb-22 23:56:56 - Ep: [25]  ValT: 0.17 min, BatchT: 0.141s, DataT: 0.015s, Loss: 0.0150, IoU: 0.9727 (val)\n",
      "05-Feb-22 23:58:05 - Ep: [26] TotalT: 34.0 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0146\n",
      "05-Feb-22 23:58:16 - Ep: [26]  ValT: 0.18 min, BatchT: 0.144s, DataT: 0.017s, Loss: 0.0158, IoU: 0.9712 (val)\n",
      "05-Feb-22 23:59:24 - Ep: [27] TotalT: 35.4 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0140\n",
      "05-Feb-22 23:59:36 - Ep: [27]  ValT: 0.19 min, BatchT: 0.153s, DataT: 0.026s, Loss: 0.0143, IoU: 0.9739 (val)\n",
      "06-Feb-22 00:00:44 - Ep: [28] TotalT: 36.7 min, BatchT: 0.272s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0139\n",
      "06-Feb-22 00:00:55 - Ep: [28]  ValT: 0.18 min, BatchT: 0.146s, DataT: 0.018s, Loss: 0.0152, IoU: 0.9720 (val)\n",
      "06-Feb-22 00:00:55 - LR was reduced to 5.0000e-05\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    22: reducing learning rate of group 0 to 5.0000e-05.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 00:02:04 - Ep: [29] TotalT: 38.0 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0131\n",
      "06-Feb-22 00:02:14 - Ep: [29]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.015s, Loss: 0.0150, IoU: 0.9724 (val)\n",
      "06-Feb-22 00:03:22 - Ep: [30] TotalT: 39.3 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0128\n",
      "06-Feb-22 00:03:32 - Ep: [30]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0145, IoU: 0.9735 (val)\n",
      "06-Feb-22 00:04:41 - Ep: [31] TotalT: 40.6 min, BatchT: 0.273s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0125\n",
      "06-Feb-22 00:04:51 - Ep: [31]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0150, IoU: 0.9724 (val)\n",
      "06-Feb-22 00:05:59 - Ep: [32] TotalT: 41.9 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0127\n",
      "06-Feb-22 00:06:09 - Ep: [32]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.014s, Loss: 0.0145, IoU: 0.9735 (val)\n",
      "06-Feb-22 00:07:17 - Ep: [33] TotalT: 43.2 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0127\n",
      "06-Feb-22 00:07:27 - Ep: [33]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.013s, Loss: 0.0148, IoU: 0.9729 (val)\n",
      "06-Feb-22 00:08:35 - Ep: [34] TotalT: 44.5 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0126\n",
      "06-Feb-22 00:08:45 - Ep: [34]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.014s, Loss: 0.0145, IoU: 0.9735 (val)\n",
      "06-Feb-22 00:08:45 - LR was reduced to 2.5000e-05\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    28: reducing learning rate of group 0 to 2.5000e-05.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 00:09:54 - Ep: [35] TotalT: 45.9 min, BatchT: 0.273s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0123\n",
      "06-Feb-22 00:10:05 - Ep: [35]  ValT: 0.18 min, BatchT: 0.146s, DataT: 0.017s, Loss: 0.0141, IoU: 0.9741 (val)\n",
      "06-Feb-22 00:11:14 - Ep: [36] TotalT: 47.2 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.003s, Loss: 0.0116\n",
      "06-Feb-22 00:11:24 - Ep: [36]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.014s, Loss: 0.0139, IoU: 0.9746 (val)\n",
      "06-Feb-22 00:12:32 - Ep: [37] TotalT: 48.5 min, BatchT: 0.272s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0124\n",
      "06-Feb-22 00:12:43 - Ep: [37]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.012s, Loss: 0.0140, IoU: 0.9742 (val)\n",
      "06-Feb-22 00:13:51 - Ep: [38] TotalT: 49.8 min, BatchT: 0.272s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0117\n",
      "06-Feb-22 00:14:02 - Ep: [38]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.012s, Loss: 0.0138, IoU: 0.9747 (val)\n",
      "06-Feb-22 00:15:11 - Ep: [39] TotalT: 51.1 min, BatchT: 0.274s, DataT: 0.004s, GpuDaT: 0.003s, Loss: 0.0116\n",
      "06-Feb-22 00:15:21 - Ep: [39]  ValT: 0.18 min, BatchT: 0.143s, DataT: 0.017s, Loss: 0.0136, IoU: 0.9750 (val)\n",
      "06-Feb-22 00:16:30 - Ep: [40] TotalT: 52.5 min, BatchT: 0.272s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0120\n",
      "06-Feb-22 00:16:40 - Ep: [40]  ValT: 0.16 min, BatchT: 0.133s, DataT: 0.011s, Loss: 0.0138, IoU: 0.9746 (val)\n",
      "06-Feb-22 00:17:48 - Ep: [41] TotalT: 53.8 min, BatchT: 0.272s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0119\n",
      "06-Feb-22 00:17:58 - Ep: [41]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.011s, Loss: 0.0139, IoU: 0.9743 (val)\n",
      "06-Feb-22 00:19:07 - Ep: [42] TotalT: 55.1 min, BatchT: 0.274s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0118\n",
      "06-Feb-22 00:19:18 - Ep: [42]  ValT: 0.18 min, BatchT: 0.145s, DataT: 0.018s, Loss: 0.0139, IoU: 0.9745 (val)\n",
      "06-Feb-22 00:19:18 - LR was reduced to 1.2500e-05\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    36: reducing learning rate of group 0 to 1.2500e-05.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 00:20:26 - Ep: [43] TotalT: 56.4 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0119\n",
      "06-Feb-22 00:20:36 - Ep: [43]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.013s, Loss: 0.0140, IoU: 0.9743 (val)\n",
      "06-Feb-22 00:21:45 - Ep: [44] TotalT: 57.7 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0115\n",
      "06-Feb-22 00:21:55 - Ep: [44]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.018s, Loss: 0.0139, IoU: 0.9745 (val)\n",
      "06-Feb-22 00:23:03 - Ep: [45] TotalT: 59.0 min, BatchT: 0.272s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0116\n",
      "06-Feb-22 00:23:14 - Ep: [45]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.013s, Loss: 0.0138, IoU: 0.9745 (val)\n",
      "06-Feb-22 00:24:22 - Ep: [46] TotalT: 60.3 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0113\n",
      "06-Feb-22 00:24:33 - Ep: [46]  ValT: 0.18 min, BatchT: 0.142s, DataT: 0.015s, Loss: 0.0136, IoU: 0.9749 (val)\n",
      "06-Feb-22 00:25:41 - Ep: [47] TotalT: 61.6 min, BatchT: 0.274s, DataT: 0.004s, GpuDaT: 0.003s, Loss: 0.0117\n",
      "06-Feb-22 00:25:52 - Ep: [47]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.014s, Loss: 0.0137, IoU: 0.9747 (val)\n",
      "06-Feb-22 00:27:01 - Ep: [48] TotalT: 63.0 min, BatchT: 0.275s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0116\n",
      "06-Feb-22 00:27:11 - Ep: [48]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.015s, Loss: 0.0138, IoU: 0.9745 (val)\n",
      "06-Feb-22 00:27:11 - LR was reduced to 6.2500e-06\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    42: reducing learning rate of group 0 to 6.2500e-06.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 00:28:20 - Ep: [49] TotalT: 64.3 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0112\n",
      "06-Feb-22 00:28:31 - Ep: [49]  ValT: 0.17 min, BatchT: 0.141s, DataT: 0.013s, Loss: 0.0137, IoU: 0.9748 (val)\n",
      "06-Feb-22 00:29:39 - Ep: [50] TotalT: 65.6 min, BatchT: 0.274s, DataT: 0.004s, GpuDaT: 0.003s, Loss: 0.0115\n",
      "06-Feb-22 00:29:50 - Ep: [50]  ValT: 0.18 min, BatchT: 0.142s, DataT: 0.017s, Loss: 0.0137, IoU: 0.9747 (val)\n",
      "06-Feb-22 00:30:58 - Ep: [51] TotalT: 66.9 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0114\n",
      "06-Feb-22 00:31:08 - Ep: [51]  ValT: 0.16 min, BatchT: 0.132s, DataT: 0.012s, Loss: 0.0137, IoU: 0.9747 (val)\n",
      "06-Feb-22 00:32:17 - Ep: [52] TotalT: 68.2 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0110\n",
      "06-Feb-22 00:32:27 - Ep: [52]  ValT: 0.18 min, BatchT: 0.142s, DataT: 0.018s, Loss: 0.0137, IoU: 0.9748 (val)\n",
      "06-Feb-22 00:32:27 - Early Stopping\n",
      "06-Feb-22 00:32:27 - Best validation combination metric of 0.97501 in epoch 39.\n",
      "06-Feb-22 00:32:28 - Training with 1 GPUS\n",
      "06-Feb-22 00:32:28 - Training is done with mixed precision\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Data Augmentations: [RandomCrop(always_apply=False, p=1.0, height=384, width=384)]\n",
      "Train Image Data Augmentations: []\n",
      "####################################################################################################\n",
      "####################################################################################################\n",
      "Fold 1 --> Train: 9398, Val: 2350\n",
      "####################################################################################################\n",
      "Last 5 sample weights: [1.0, 1.0, 1.0, 1.0, 1.0]\n",
      "unet timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 00:33:35 - Ep: [1] TotalT: 1.1 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.1252\n",
      "06-Feb-22 00:33:45 - Ep: [1]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.012s, Loss: 0.0540, IoU: 0.9406 (val)\n",
      "06-Feb-22 00:34:53 - Ep: [2] TotalT: 2.4 min, BatchT: 0.269s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0476\n",
      "06-Feb-22 00:35:04 - Ep: [2]  ValT: 0.18 min, BatchT: 0.143s, DataT: 0.018s, Loss: 0.0357, IoU: 0.9519 (val)\n",
      "06-Feb-22 00:36:11 - Ep: [3] TotalT: 3.7 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0370\n",
      "06-Feb-22 00:36:21 - Ep: [3]  ValT: 0.16 min, BatchT: 0.133s, DataT: 0.013s, Loss: 0.0300, IoU: 0.9556 (val)\n",
      "06-Feb-22 00:37:29 - Ep: [4] TotalT: 5.0 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0322\n",
      "06-Feb-22 00:37:39 - Ep: [4]  ValT: 0.17 min, BatchT: 0.134s, DataT: 0.014s, Loss: 0.0267, IoU: 0.9585 (val)\n",
      "06-Feb-22 00:38:46 - Ep: [5] TotalT: 6.3 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0283\n",
      "06-Feb-22 00:38:57 - Ep: [5]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.015s, Loss: 0.0240, IoU: 0.9616 (val)\n",
      "06-Feb-22 00:40:05 - Ep: [6] TotalT: 7.6 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0253\n",
      "06-Feb-22 00:40:16 - Ep: [6]  ValT: 0.17 min, BatchT: 0.142s, DataT: 0.015s, Loss: 0.0227, IoU: 0.9631 (val)\n",
      "06-Feb-22 00:41:24 - Ep: [7] TotalT: 8.9 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0241\n",
      "06-Feb-22 00:41:34 - Ep: [7]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.013s, Loss: 0.0212, IoU: 0.9650 (val)\n",
      "06-Feb-22 00:42:42 - Ep: [8] TotalT: 10.2 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0223\n",
      "06-Feb-22 00:42:52 - Ep: [8]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.013s, Loss: 0.0206, IoU: 0.9657 (val)\n",
      "06-Feb-22 00:44:00 - Ep: [9] TotalT: 11.5 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0208\n",
      "06-Feb-22 00:44:10 - Ep: [9]  ValT: 0.17 min, BatchT: 0.142s, DataT: 0.016s, Loss: 0.0202, IoU: 0.9663 (val)\n",
      "06-Feb-22 00:45:18 - Ep: [10] TotalT: 12.8 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0203\n",
      "06-Feb-22 00:45:29 - Ep: [10]  ValT: 0.18 min, BatchT: 0.142s, DataT: 0.018s, Loss: 0.0192, IoU: 0.9679 (val)\n",
      "06-Feb-22 00:46:37 - Ep: [11] TotalT: 14.2 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0199\n",
      "06-Feb-22 00:46:48 - Ep: [11]  ValT: 0.18 min, BatchT: 0.146s, DataT: 0.016s, Loss: 0.0179, IoU: 0.9696 (val)\n",
      "06-Feb-22 00:47:56 - Ep: [12] TotalT: 15.5 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0189\n",
      "06-Feb-22 00:48:06 - Ep: [12]  ValT: 0.17 min, BatchT: 0.134s, DataT: 0.014s, Loss: 0.0172, IoU: 0.9709 (val)\n",
      "06-Feb-22 00:49:14 - Ep: [13] TotalT: 16.8 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0187\n",
      "06-Feb-22 00:49:24 - Ep: [13]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.014s, Loss: 0.0180, IoU: 0.9695 (val)\n",
      "06-Feb-22 00:50:31 - Ep: [14] TotalT: 18.1 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0183\n",
      "06-Feb-22 00:50:41 - Ep: [14]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0176, IoU: 0.9700 (val)\n",
      "06-Feb-22 00:51:49 - Ep: [15] TotalT: 19.4 min, BatchT: 0.269s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0177\n",
      "06-Feb-22 00:51:59 - Ep: [15]  ValT: 0.16 min, BatchT: 0.132s, DataT: 0.011s, Loss: 0.0164, IoU: 0.9719 (val)\n",
      "06-Feb-22 00:53:06 - Ep: [16] TotalT: 20.6 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0177\n",
      "06-Feb-22 00:53:16 - Ep: [16]  ValT: 0.16 min, BatchT: 0.134s, DataT: 0.012s, Loss: 0.0170, IoU: 0.9710 (val)\n",
      "06-Feb-22 00:54:23 - Ep: [17] TotalT: 21.9 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0180\n",
      "06-Feb-22 00:54:34 - Ep: [17]  ValT: 0.17 min, BatchT: 0.141s, DataT: 0.016s, Loss: 0.0177, IoU: 0.9700 (val)\n",
      "06-Feb-22 00:55:41 - Ep: [18] TotalT: 23.2 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0162\n",
      "06-Feb-22 00:55:51 - Ep: [18]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.013s, Loss: 0.0163, IoU: 0.9720 (val)\n",
      "06-Feb-22 00:56:59 - Ep: [19] TotalT: 24.5 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0163\n",
      "06-Feb-22 00:57:10 - Ep: [19]  ValT: 0.18 min, BatchT: 0.146s, DataT: 0.018s, Loss: 0.0155, IoU: 0.9734 (val)\n",
      "06-Feb-22 00:58:19 - Ep: [20] TotalT: 25.9 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0154\n",
      "06-Feb-22 00:58:29 - Ep: [20]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.014s, Loss: 0.0154, IoU: 0.9734 (val)\n",
      "06-Feb-22 00:59:37 - Ep: [21] TotalT: 27.2 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0156\n",
      "06-Feb-22 00:59:48 - Ep: [21]  ValT: 0.18 min, BatchT: 0.146s, DataT: 0.018s, Loss: 0.0152, IoU: 0.9737 (val)\n",
      "06-Feb-22 01:00:56 - Ep: [22] TotalT: 28.5 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0154\n",
      "06-Feb-22 01:01:06 - Ep: [22]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.013s, Loss: 0.0151, IoU: 0.9741 (val)\n",
      "06-Feb-22 01:02:14 - Ep: [23] TotalT: 29.8 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0150\n",
      "06-Feb-22 01:02:24 - Ep: [23]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0151, IoU: 0.9739 (val)\n",
      "06-Feb-22 01:03:32 - Ep: [24] TotalT: 31.1 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0149\n",
      "06-Feb-22 01:03:43 - Ep: [24]  ValT: 0.17 min, BatchT: 0.142s, DataT: 0.016s, Loss: 0.0154, IoU: 0.9732 (val)\n",
      "06-Feb-22 01:04:50 - Ep: [25] TotalT: 32.4 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0144\n",
      "06-Feb-22 01:05:01 - Ep: [25]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.012s, Loss: 0.0145, IoU: 0.9747 (val)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    19: reducing learning rate of group 0 to 5.0000e-05.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 01:06:08 - Ep: [26] TotalT: 33.7 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0142\n",
      "06-Feb-22 01:06:19 - Ep: [26]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.018s, Loss: 0.0139, IoU: 0.9759 (val)\n",
      "06-Feb-22 01:07:26 - Ep: [27] TotalT: 35.0 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0141\n",
      "06-Feb-22 01:07:37 - Ep: [27]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.014s, Loss: 0.0138, IoU: 0.9759 (val)\n",
      "06-Feb-22 01:08:45 - Ep: [28] TotalT: 36.3 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0136\n",
      "06-Feb-22 01:08:55 - Ep: [28]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.013s, Loss: 0.0137, IoU: 0.9761 (val)\n",
      "06-Feb-22 01:10:03 - Ep: [29] TotalT: 37.6 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0134\n",
      "06-Feb-22 01:10:13 - Ep: [29]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.014s, Loss: 0.0145, IoU: 0.9748 (val)\n",
      "06-Feb-22 01:11:20 - Ep: [30] TotalT: 38.9 min, BatchT: 0.268s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0132\n",
      "06-Feb-22 01:11:31 - Ep: [30]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.011s, Loss: 0.0137, IoU: 0.9762 (val)\n",
      "06-Feb-22 01:12:38 - Ep: [31] TotalT: 40.2 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0128\n",
      "06-Feb-22 01:12:48 - Ep: [31]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.014s, Loss: 0.0136, IoU: 0.9764 (val)\n",
      "06-Feb-22 01:13:56 - Ep: [32] TotalT: 41.5 min, BatchT: 0.268s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0131\n",
      "06-Feb-22 01:14:06 - Ep: [32]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.014s, Loss: 0.0142, IoU: 0.9750 (val)\n",
      "06-Feb-22 01:14:06 - LR was reduced to 2.5000e-05\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    26: reducing learning rate of group 0 to 2.5000e-05.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 01:15:14 - Ep: [33] TotalT: 42.8 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0131\n",
      "06-Feb-22 01:15:25 - Ep: [33]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.015s, Loss: 0.0133, IoU: 0.9767 (val)\n",
      "06-Feb-22 01:16:33 - Ep: [34] TotalT: 44.1 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0127\n",
      "06-Feb-22 01:16:43 - Ep: [34]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.014s, Loss: 0.0131, IoU: 0.9769 (val)\n",
      "06-Feb-22 01:17:51 - Ep: [35] TotalT: 45.4 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0122\n",
      "06-Feb-22 01:18:01 - Ep: [35]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.013s, Loss: 0.0132, IoU: 0.9768 (val)\n",
      "06-Feb-22 01:19:09 - Ep: [36] TotalT: 46.7 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0126\n",
      "06-Feb-22 01:19:20 - Ep: [36]  ValT: 0.18 min, BatchT: 0.144s, DataT: 0.017s, Loss: 0.0128, IoU: 0.9776 (val)\n",
      "06-Feb-22 01:20:28 - Ep: [37] TotalT: 48.0 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0121\n",
      "06-Feb-22 01:20:39 - Ep: [37]  ValT: 0.17 min, BatchT: 0.141s, DataT: 0.016s, Loss: 0.0127, IoU: 0.9776 (val)\n",
      "06-Feb-22 01:21:47 - Ep: [38] TotalT: 49.3 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0121\n",
      "06-Feb-22 01:21:57 - Ep: [38]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.014s, Loss: 0.0129, IoU: 0.9772 (val)\n",
      "06-Feb-22 01:23:05 - Ep: [39] TotalT: 50.6 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0123\n",
      "06-Feb-22 01:23:15 - Ep: [39]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.012s, Loss: 0.0128, IoU: 0.9775 (val)\n",
      "06-Feb-22 01:24:23 - Ep: [40] TotalT: 51.9 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0118\n",
      "06-Feb-22 01:24:34 - Ep: [40]  ValT: 0.17 min, BatchT: 0.141s, DataT: 0.015s, Loss: 0.0133, IoU: 0.9766 (val)\n",
      "06-Feb-22 01:25:42 - Ep: [41] TotalT: 53.2 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0119\n",
      "06-Feb-22 01:25:52 - Ep: [41]  ValT: 0.17 min, BatchT: 0.142s, DataT: 0.016s, Loss: 0.0132, IoU: 0.9769 (val)\n",
      "06-Feb-22 01:27:00 - Ep: [42] TotalT: 54.5 min, BatchT: 0.269s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0121\n",
      "06-Feb-22 01:27:10 - Ep: [42]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.014s, Loss: 0.0126, IoU: 0.9778 (val)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    36: reducing learning rate of group 0 to 1.2500e-05.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 01:28:18 - Ep: [43] TotalT: 55.8 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0117\n",
      "06-Feb-22 01:28:29 - Ep: [43]  ValT: 0.17 min, BatchT: 0.141s, DataT: 0.015s, Loss: 0.0128, IoU: 0.9775 (val)\n",
      "06-Feb-22 01:29:36 - Ep: [44] TotalT: 57.1 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0119\n",
      "06-Feb-22 01:29:46 - Ep: [44]  ValT: 0.16 min, BatchT: 0.133s, DataT: 0.011s, Loss: 0.0128, IoU: 0.9776 (val)\n",
      "06-Feb-22 01:30:53 - Ep: [45] TotalT: 58.4 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0120\n",
      "06-Feb-22 01:31:04 - Ep: [45]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.016s, Loss: 0.0128, IoU: 0.9776 (val)\n",
      "06-Feb-22 01:32:11 - Ep: [46] TotalT: 59.7 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0118\n",
      "06-Feb-22 01:32:22 - Ep: [46]  ValT: 0.18 min, BatchT: 0.147s, DataT: 0.020s, Loss: 0.0128, IoU: 0.9776 (val)\n",
      "06-Feb-22 01:33:29 - Ep: [47] TotalT: 61.0 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0119\n",
      "06-Feb-22 01:33:40 - Ep: [47]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.012s, Loss: 0.0127, IoU: 0.9778 (val)\n",
      "06-Feb-22 01:34:48 - Ep: [48] TotalT: 62.3 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0117\n",
      "06-Feb-22 01:34:58 - Ep: [48]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.012s, Loss: 0.0130, IoU: 0.9772 (val)\n",
      "06-Feb-22 01:34:58 - LR was reduced to 6.2500e-06\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    42: reducing learning rate of group 0 to 6.2500e-06.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 01:36:07 - Ep: [49] TotalT: 63.7 min, BatchT: 0.272s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0117\n",
      "06-Feb-22 01:36:17 - Ep: [49]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.013s, Loss: 0.0130, IoU: 0.9772 (val)\n",
      "06-Feb-22 01:37:25 - Ep: [50] TotalT: 65.0 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0118\n",
      "06-Feb-22 01:37:35 - Ep: [50]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.015s, Loss: 0.0127, IoU: 0.9778 (val)\n",
      "06-Feb-22 01:38:43 - Ep: [51] TotalT: 66.3 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0117\n",
      "06-Feb-22 01:38:54 - Ep: [51]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.014s, Loss: 0.0126, IoU: 0.9779 (val)\n",
      "06-Feb-22 01:40:02 - Ep: [52] TotalT: 67.6 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0115\n",
      "06-Feb-22 01:40:12 - Ep: [52]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.013s, Loss: 0.0125, IoU: 0.9780 (val)\n",
      "06-Feb-22 01:41:20 - Ep: [53] TotalT: 68.9 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0117\n",
      "06-Feb-22 01:41:31 - Ep: [53]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0124, IoU: 0.9781 (val)\n",
      "06-Feb-22 01:42:38 - Ep: [54] TotalT: 70.2 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0115\n",
      "06-Feb-22 01:42:49 - Ep: [54]  ValT: 0.17 min, BatchT: 0.141s, DataT: 0.014s, Loss: 0.0124, IoU: 0.9781 (val)\n",
      "06-Feb-22 01:42:49 - LR was reduced to 3.1250e-06\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    48: reducing learning rate of group 0 to 3.1250e-06.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 01:43:56 - Ep: [55] TotalT: 71.5 min, BatchT: 0.269s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0117\n",
      "06-Feb-22 01:44:07 - Ep: [55]  ValT: 0.18 min, BatchT: 0.145s, DataT: 0.019s, Loss: 0.0125, IoU: 0.9780 (val)\n",
      "06-Feb-22 01:45:15 - Ep: [56] TotalT: 72.8 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0115\n",
      "06-Feb-22 01:45:25 - Ep: [56]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.015s, Loss: 0.0124, IoU: 0.9782 (val)\n",
      "06-Feb-22 01:46:33 - Ep: [57] TotalT: 74.1 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0112\n",
      "06-Feb-22 01:46:43 - Ep: [57]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.014s, Loss: 0.0124, IoU: 0.9782 (val)\n",
      "06-Feb-22 01:47:51 - Ep: [58] TotalT: 75.4 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0117\n",
      "06-Feb-22 01:48:01 - Ep: [58]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.015s, Loss: 0.0125, IoU: 0.9780 (val)\n",
      "06-Feb-22 01:49:09 - Ep: [59] TotalT: 76.7 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0115\n",
      "06-Feb-22 01:49:20 - Ep: [59]  ValT: 0.18 min, BatchT: 0.145s, DataT: 0.016s, Loss: 0.0127, IoU: 0.9777 (val)\n",
      "06-Feb-22 01:50:28 - Ep: [60] TotalT: 78.0 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0114\n",
      "06-Feb-22 01:50:38 - Ep: [60]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.013s, Loss: 0.0125, IoU: 0.9780 (val)\n",
      "06-Feb-22 01:50:38 - LR was reduced to 1.5625e-06\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    54: reducing learning rate of group 0 to 1.5625e-06.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 01:51:47 - Ep: [61] TotalT: 79.3 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0114\n",
      "06-Feb-22 01:51:57 - Ep: [61]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.013s, Loss: 0.0125, IoU: 0.9780 (val)\n",
      "06-Feb-22 01:53:04 - Ep: [62] TotalT: 80.6 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0115\n",
      "06-Feb-22 01:53:15 - Ep: [62]  ValT: 0.18 min, BatchT: 0.142s, DataT: 0.016s, Loss: 0.0125, IoU: 0.9780 (val)\n",
      "06-Feb-22 01:54:23 - Ep: [63] TotalT: 81.9 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0114\n",
      "06-Feb-22 01:54:34 - Ep: [63]  ValT: 0.18 min, BatchT: 0.144s, DataT: 0.016s, Loss: 0.0126, IoU: 0.9780 (val)\n",
      "06-Feb-22 01:55:41 - Ep: [64] TotalT: 83.2 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0115\n",
      "06-Feb-22 01:55:52 - Ep: [64]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.013s, Loss: 0.0125, IoU: 0.9781 (val)\n",
      "06-Feb-22 01:57:00 - Ep: [65] TotalT: 84.5 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0115\n",
      "06-Feb-22 01:57:10 - Ep: [65]  ValT: 0.17 min, BatchT: 0.142s, DataT: 0.013s, Loss: 0.0125, IoU: 0.9781 (val)\n",
      "06-Feb-22 01:58:18 - Ep: [66] TotalT: 85.8 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0113\n",
      "06-Feb-22 01:58:28 - Ep: [66]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.012s, Loss: 0.0125, IoU: 0.9781 (val)\n",
      "06-Feb-22 01:58:28 - Early Stopping\n",
      "06-Feb-22 01:58:28 - Best validation combination metric of 0.97820 in epoch 56.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    60: reducing learning rate of group 0 to 7.8125e-07.\n",
      "Train Data Augmentations: [RandomCrop(always_apply=False, p=1.0, height=384, width=384)]\n",
      "Train Image Data Augmentations: []\n",
      "####################################################################################################\n",
      "####################################################################################################\n",
      "Fold 2 --> Train: 9398, Val: 2350\n",
      "####################################################################################################\n",
      "Last 5 sample weights: [1.0, 1.0, 1.0, 1.0, 1.0]\n",
      "unet timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 01:58:29 - Training with 1 GPUS\n",
      "06-Feb-22 01:58:29 - Training is done with mixed precision\n",
      "06-Feb-22 01:59:36 - Ep: [1] TotalT: 1.1 min, BatchT: 0.269s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.1110\n",
      "06-Feb-22 01:59:47 - Ep: [1]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.015s, Loss: 0.0459, IoU: 0.9442 (val)\n",
      "06-Feb-22 02:00:54 - Ep: [2] TotalT: 2.4 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0447\n",
      "06-Feb-22 02:01:04 - Ep: [2]  ValT: 0.16 min, BatchT: 0.129s, DataT: 0.012s, Loss: 0.0335, IoU: 0.9497 (val)\n",
      "06-Feb-22 02:02:12 - Ep: [3] TotalT: 3.7 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0363\n",
      "06-Feb-22 02:02:22 - Ep: [3]  ValT: 0.16 min, BatchT: 0.130s, DataT: 0.012s, Loss: 0.0300, IoU: 0.9532 (val)\n",
      "06-Feb-22 02:03:29 - Ep: [4] TotalT: 5.0 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0309\n",
      "06-Feb-22 02:03:39 - Ep: [4]  ValT: 0.16 min, BatchT: 0.131s, DataT: 0.012s, Loss: 0.0277, IoU: 0.9561 (val)\n",
      "06-Feb-22 02:04:46 - Ep: [5] TotalT: 6.3 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0278\n",
      "06-Feb-22 02:04:57 - Ep: [5]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.014s, Loss: 0.0245, IoU: 0.9599 (val)\n",
      "06-Feb-22 02:06:05 - Ep: [6] TotalT: 7.6 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0258\n",
      "06-Feb-22 02:06:15 - Ep: [6]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.015s, Loss: 0.0227, IoU: 0.9628 (val)\n",
      "06-Feb-22 02:07:23 - Ep: [7] TotalT: 8.9 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0237\n",
      "06-Feb-22 02:07:33 - Ep: [7]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.015s, Loss: 0.0238, IoU: 0.9607 (val)\n",
      "06-Feb-22 02:08:41 - Ep: [8] TotalT: 10.2 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0229\n",
      "06-Feb-22 02:08:52 - Ep: [8]  ValT: 0.18 min, BatchT: 0.142s, DataT: 0.014s, Loss: 0.0203, IoU: 0.9661 (val)\n",
      "06-Feb-22 02:10:00 - Ep: [9] TotalT: 11.5 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0215\n",
      "06-Feb-22 02:10:10 - Ep: [9]  ValT: 0.16 min, BatchT: 0.131s, DataT: 0.011s, Loss: 0.0194, IoU: 0.9672 (val)\n",
      "06-Feb-22 02:11:17 - Ep: [10] TotalT: 12.8 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0206\n",
      "06-Feb-22 02:11:28 - Ep: [10]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.012s, Loss: 0.0218, IoU: 0.9622 (val)\n",
      "06-Feb-22 02:12:35 - Ep: [11] TotalT: 14.1 min, BatchT: 0.269s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0197\n",
      "06-Feb-22 02:12:45 - Ep: [11]  ValT: 0.16 min, BatchT: 0.130s, DataT: 0.013s, Loss: 0.0187, IoU: 0.9678 (val)\n",
      "06-Feb-22 02:13:53 - Ep: [12] TotalT: 15.4 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0186\n",
      "06-Feb-22 02:14:03 - Ep: [12]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.017s, Loss: 0.0178, IoU: 0.9695 (val)\n",
      "06-Feb-22 02:15:10 - Ep: [13] TotalT: 16.7 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0181\n",
      "06-Feb-22 02:15:21 - Ep: [13]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0182, IoU: 0.9687 (val)\n",
      "06-Feb-22 02:16:28 - Ep: [14] TotalT: 18.0 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0182\n",
      "06-Feb-22 02:16:38 - Ep: [14]  ValT: 0.16 min, BatchT: 0.133s, DataT: 0.013s, Loss: 0.0176, IoU: 0.9697 (val)\n",
      "06-Feb-22 02:17:45 - Ep: [15] TotalT: 19.3 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0187\n",
      "06-Feb-22 02:17:56 - Ep: [15]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.020s, Loss: 0.0175, IoU: 0.9696 (val)\n",
      "06-Feb-22 02:19:03 - Ep: [16] TotalT: 20.6 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0168\n",
      "06-Feb-22 02:19:13 - Ep: [16]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.015s, Loss: 0.0177, IoU: 0.9686 (val)\n",
      "06-Feb-22 02:20:21 - Ep: [17] TotalT: 21.9 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0178\n",
      "06-Feb-22 02:20:31 - Ep: [17]  ValT: 0.16 min, BatchT: 0.133s, DataT: 0.013s, Loss: 0.0171, IoU: 0.9701 (val)\n",
      "06-Feb-22 02:21:38 - Ep: [18] TotalT: 23.2 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0169\n",
      "06-Feb-22 02:21:48 - Ep: [18]  ValT: 0.16 min, BatchT: 0.134s, DataT: 0.014s, Loss: 0.0171, IoU: 0.9701 (val)\n",
      "06-Feb-22 02:21:48 - LR was reduced to 5.0000e-05\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    12: reducing learning rate of group 0 to 5.0000e-05.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 02:22:55 - Ep: [19] TotalT: 24.4 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0160\n",
      "06-Feb-22 02:23:05 - Ep: [19]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.015s, Loss: 0.0156, IoU: 0.9728 (val)\n",
      "06-Feb-22 02:24:13 - Ep: [20] TotalT: 25.7 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0153\n",
      "06-Feb-22 02:24:23 - Ep: [20]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.015s, Loss: 0.0156, IoU: 0.9728 (val)\n",
      "06-Feb-22 02:25:31 - Ep: [21] TotalT: 27.0 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0148\n",
      "06-Feb-22 02:25:41 - Ep: [21]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0150, IoU: 0.9738 (val)\n",
      "06-Feb-22 02:26:49 - Ep: [22] TotalT: 28.3 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0153\n",
      "06-Feb-22 02:26:59 - Ep: [22]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.013s, Loss: 0.0149, IoU: 0.9740 (val)\n",
      "06-Feb-22 02:28:07 - Ep: [23] TotalT: 29.6 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0146\n",
      "06-Feb-22 02:28:17 - Ep: [23]  ValT: 0.16 min, BatchT: 0.134s, DataT: 0.012s, Loss: 0.0152, IoU: 0.9735 (val)\n",
      "06-Feb-22 02:29:24 - Ep: [24] TotalT: 30.9 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0146\n",
      "06-Feb-22 02:29:35 - Ep: [24]  ValT: 0.18 min, BatchT: 0.144s, DataT: 0.016s, Loss: 0.0152, IoU: 0.9734 (val)\n",
      "06-Feb-22 02:30:43 - Ep: [25] TotalT: 32.2 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0150\n",
      "06-Feb-22 02:30:54 - Ep: [25]  ValT: 0.18 min, BatchT: 0.145s, DataT: 0.016s, Loss: 0.0149, IoU: 0.9740 (val)\n",
      "06-Feb-22 02:30:54 - LR was reduced to 2.5000e-05\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    19: reducing learning rate of group 0 to 2.5000e-05.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 02:32:01 - Ep: [26] TotalT: 33.5 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0142\n",
      "06-Feb-22 02:32:11 - Ep: [26]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.014s, Loss: 0.0147, IoU: 0.9745 (val)\n",
      "06-Feb-22 02:33:19 - Ep: [27] TotalT: 34.8 min, BatchT: 0.269s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0144\n",
      "06-Feb-22 02:33:29 - Ep: [27]  ValT: 0.16 min, BatchT: 0.134s, DataT: 0.013s, Loss: 0.0146, IoU: 0.9744 (val)\n",
      "06-Feb-22 02:34:37 - Ep: [28] TotalT: 36.1 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0139\n",
      "06-Feb-22 02:34:47 - Ep: [28]  ValT: 0.17 min, BatchT: 0.134s, DataT: 0.013s, Loss: 0.0145, IoU: 0.9746 (val)\n",
      "06-Feb-22 02:35:55 - Ep: [29] TotalT: 37.4 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0136\n",
      "06-Feb-22 02:36:05 - Ep: [29]  ValT: 0.16 min, BatchT: 0.133s, DataT: 0.014s, Loss: 0.0144, IoU: 0.9748 (val)\n",
      "06-Feb-22 02:37:13 - Ep: [30] TotalT: 38.7 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0139\n",
      "06-Feb-22 02:37:23 - Ep: [30]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.016s, Loss: 0.0143, IoU: 0.9750 (val)\n",
      "06-Feb-22 02:38:30 - Ep: [31] TotalT: 40.0 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0138\n",
      "06-Feb-22 02:38:40 - Ep: [31]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.013s, Loss: 0.0145, IoU: 0.9745 (val)\n",
      "06-Feb-22 02:39:47 - Ep: [32] TotalT: 41.3 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0133\n",
      "06-Feb-22 02:39:58 - Ep: [32]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.015s, Loss: 0.0145, IoU: 0.9746 (val)\n",
      "06-Feb-22 02:39:58 - LR was reduced to 1.2500e-05\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    26: reducing learning rate of group 0 to 1.2500e-05.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 02:41:05 - Ep: [33] TotalT: 42.6 min, BatchT: 0.269s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0133\n",
      "06-Feb-22 02:41:15 - Ep: [33]  ValT: 0.17 min, BatchT: 0.134s, DataT: 0.014s, Loss: 0.0144, IoU: 0.9747 (val)\n",
      "06-Feb-22 02:42:23 - Ep: [34] TotalT: 43.9 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0134\n",
      "06-Feb-22 02:42:33 - Ep: [34]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.012s, Loss: 0.0142, IoU: 0.9750 (val)\n",
      "06-Feb-22 02:43:41 - Ep: [35] TotalT: 45.2 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0133\n",
      "06-Feb-22 02:43:51 - Ep: [35]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.014s, Loss: 0.0143, IoU: 0.9750 (val)\n",
      "06-Feb-22 02:44:59 - Ep: [36] TotalT: 46.5 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0135\n",
      "06-Feb-22 02:45:09 - Ep: [36]  ValT: 0.18 min, BatchT: 0.142s, DataT: 0.016s, Loss: 0.0142, IoU: 0.9751 (val)\n",
      "06-Feb-22 02:46:18 - Ep: [37] TotalT: 47.8 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0133\n",
      "06-Feb-22 02:46:28 - Ep: [37]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.017s, Loss: 0.0140, IoU: 0.9754 (val)\n",
      "06-Feb-22 02:47:36 - Ep: [38] TotalT: 49.1 min, BatchT: 0.269s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0130\n",
      "06-Feb-22 02:47:46 - Ep: [38]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.013s, Loss: 0.0140, IoU: 0.9754 (val)\n",
      "06-Feb-22 02:47:46 - LR was reduced to 6.2500e-06\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    32: reducing learning rate of group 0 to 6.2500e-06.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 02:48:54 - Ep: [39] TotalT: 50.4 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0131\n",
      "06-Feb-22 02:49:05 - Ep: [39]  ValT: 0.18 min, BatchT: 0.144s, DataT: 0.018s, Loss: 0.0139, IoU: 0.9757 (val)\n",
      "06-Feb-22 02:50:13 - Ep: [40] TotalT: 51.7 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0128\n",
      "06-Feb-22 02:50:23 - Ep: [40]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0140, IoU: 0.9754 (val)\n",
      "06-Feb-22 02:51:30 - Ep: [41] TotalT: 53.0 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0129\n",
      "06-Feb-22 02:51:41 - Ep: [41]  ValT: 0.17 min, BatchT: 0.142s, DataT: 0.016s, Loss: 0.0141, IoU: 0.9753 (val)\n",
      "06-Feb-22 02:52:49 - Ep: [42] TotalT: 54.3 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0130\n",
      "06-Feb-22 02:52:59 - Ep: [42]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.012s, Loss: 0.0139, IoU: 0.9757 (val)\n",
      "06-Feb-22 02:54:07 - Ep: [43] TotalT: 55.6 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0129\n",
      "06-Feb-22 02:54:18 - Ep: [43]  ValT: 0.18 min, BatchT: 0.146s, DataT: 0.018s, Loss: 0.0140, IoU: 0.9754 (val)\n",
      "06-Feb-22 02:55:25 - Ep: [44] TotalT: 56.9 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0127\n",
      "06-Feb-22 02:55:35 - Ep: [44]  ValT: 0.16 min, BatchT: 0.133s, DataT: 0.012s, Loss: 0.0139, IoU: 0.9757 (val)\n",
      "06-Feb-22 02:55:35 - LR was reduced to 3.1250e-06\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    38: reducing learning rate of group 0 to 3.1250e-06.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 02:56:43 - Ep: [45] TotalT: 58.2 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0131\n",
      "06-Feb-22 02:56:53 - Ep: [45]  ValT: 0.16 min, BatchT: 0.133s, DataT: 0.013s, Loss: 0.0139, IoU: 0.9756 (val)\n",
      "06-Feb-22 02:58:00 - Ep: [46] TotalT: 59.5 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0130\n",
      "06-Feb-22 02:58:11 - Ep: [46]  ValT: 0.17 min, BatchT: 0.141s, DataT: 0.014s, Loss: 0.0139, IoU: 0.9755 (val)\n",
      "06-Feb-22 02:59:19 - Ep: [47] TotalT: 60.8 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0131\n",
      "06-Feb-22 02:59:29 - Ep: [47]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.015s, Loss: 0.0139, IoU: 0.9756 (val)\n",
      "06-Feb-22 03:00:37 - Ep: [48] TotalT: 62.1 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0132\n",
      "06-Feb-22 03:00:47 - Ep: [48]  ValT: 0.17 min, BatchT: 0.134s, DataT: 0.012s, Loss: 0.0139, IoU: 0.9756 (val)\n",
      "06-Feb-22 03:01:55 - Ep: [49] TotalT: 63.4 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0132\n",
      "06-Feb-22 03:02:05 - Ep: [49]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.013s, Loss: 0.0140, IoU: 0.9755 (val)\n",
      "06-Feb-22 03:03:13 - Ep: [50] TotalT: 64.7 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0125\n",
      "06-Feb-22 03:03:24 - Ep: [50]  ValT: 0.18 min, BatchT: 0.143s, DataT: 0.016s, Loss: 0.0139, IoU: 0.9756 (val)\n",
      "06-Feb-22 03:03:24 - LR was reduced to 1.5625e-06\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    44: reducing learning rate of group 0 to 1.5625e-06.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 03:04:31 - Ep: [51] TotalT: 66.0 min, BatchT: 0.269s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0128\n",
      "06-Feb-22 03:04:42 - Ep: [51]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.014s, Loss: 0.0138, IoU: 0.9758 (val)\n",
      "06-Feb-22 03:05:49 - Ep: [52] TotalT: 67.3 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0129\n",
      "06-Feb-22 03:06:00 - Ep: [52]  ValT: 0.17 min, BatchT: 0.141s, DataT: 0.013s, Loss: 0.0138, IoU: 0.9758 (val)\n",
      "06-Feb-22 03:06:00 - Early Stopping\n",
      "06-Feb-22 03:06:00 - Best validation combination metric of 0.97580 in epoch 52.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Data Augmentations: [RandomCrop(always_apply=False, p=1.0, height=384, width=384)]\n",
      "Train Image Data Augmentations: []\n",
      "####################################################################################################\n",
      "####################################################################################################\n",
      "Fold 3 --> Train: 9399, Val: 2349\n",
      "####################################################################################################\n",
      "Last 5 sample weights: [1.0, 1.0, 1.0, 1.0, 1.0]\n",
      "unet timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 03:06:01 - Training with 1 GPUS\n",
      "06-Feb-22 03:06:01 - Training is done with mixed precision\n",
      "06-Feb-22 03:07:09 - Ep: [1] TotalT: 1.1 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.1314\n",
      "06-Feb-22 03:07:21 - Ep: [1]  ValT: 0.20 min, BatchT: 0.159s, DataT: 0.014s, Loss: 0.0550, IoU: 0.9340 (val)\n",
      "06-Feb-22 03:08:30 - Ep: [2] TotalT: 2.5 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0498\n",
      "06-Feb-22 03:08:41 - Ep: [2]  ValT: 0.17 min, BatchT: 0.142s, DataT: 0.015s, Loss: 0.0385, IoU: 0.9453 (val)\n",
      "06-Feb-22 03:09:49 - Ep: [3] TotalT: 3.8 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0376\n",
      "06-Feb-22 03:09:59 - Ep: [3]  ValT: 0.17 min, BatchT: 0.134s, DataT: 0.012s, Loss: 0.0301, IoU: 0.9525 (val)\n",
      "06-Feb-22 03:11:07 - Ep: [4] TotalT: 5.1 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0322\n",
      "06-Feb-22 03:11:18 - Ep: [4]  ValT: 0.17 min, BatchT: 0.142s, DataT: 0.016s, Loss: 0.0277, IoU: 0.9547 (val)\n",
      "06-Feb-22 03:12:26 - Ep: [5] TotalT: 6.4 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0284\n",
      "06-Feb-22 03:12:36 - Ep: [5]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0248, IoU: 0.9583 (val)\n",
      "06-Feb-22 03:13:44 - Ep: [6] TotalT: 7.7 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0263\n",
      "06-Feb-22 03:13:54 - Ep: [6]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.012s, Loss: 0.0231, IoU: 0.9609 (val)\n",
      "06-Feb-22 03:15:03 - Ep: [7] TotalT: 9.0 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0238\n",
      "06-Feb-22 03:15:13 - Ep: [7]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.012s, Loss: 0.0209, IoU: 0.9644 (val)\n",
      "06-Feb-22 03:16:21 - Ep: [8] TotalT: 10.3 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0221\n",
      "06-Feb-22 03:16:31 - Ep: [8]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.015s, Loss: 0.0204, IoU: 0.9648 (val)\n",
      "06-Feb-22 03:17:40 - Ep: [9] TotalT: 11.7 min, BatchT: 0.273s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0213\n",
      "06-Feb-22 03:17:51 - Ep: [9]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.015s, Loss: 0.0198, IoU: 0.9653 (val)\n",
      "06-Feb-22 03:18:59 - Ep: [10] TotalT: 13.0 min, BatchT: 0.272s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0205\n",
      "06-Feb-22 03:19:10 - Ep: [10]  ValT: 0.17 min, BatchT: 0.142s, DataT: 0.015s, Loss: 0.0201, IoU: 0.9645 (val)\n",
      "06-Feb-22 03:20:19 - Ep: [11] TotalT: 14.3 min, BatchT: 0.275s, DataT: 0.004s, GpuDaT: 0.003s, Loss: 0.0193\n",
      "06-Feb-22 03:20:29 - Ep: [11]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.013s, Loss: 0.0200, IoU: 0.9650 (val)\n",
      "06-Feb-22 03:21:38 - Ep: [12] TotalT: 15.6 min, BatchT: 0.275s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0190\n",
      "06-Feb-22 03:21:48 - Ep: [12]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.013s, Loss: 0.0183, IoU: 0.9677 (val)\n",
      "06-Feb-22 03:22:57 - Ep: [13] TotalT: 16.9 min, BatchT: 0.272s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0188\n",
      "06-Feb-22 03:23:08 - Ep: [13]  ValT: 0.18 min, BatchT: 0.144s, DataT: 0.015s, Loss: 0.0181, IoU: 0.9675 (val)\n",
      "06-Feb-22 03:24:17 - Ep: [14] TotalT: 18.3 min, BatchT: 0.274s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0180\n",
      "06-Feb-22 03:24:27 - Ep: [14]  ValT: 0.17 min, BatchT: 0.142s, DataT: 0.014s, Loss: 0.0179, IoU: 0.9678 (val)\n",
      "06-Feb-22 03:25:35 - Ep: [15] TotalT: 19.6 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0173\n",
      "06-Feb-22 03:25:46 - Ep: [15]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.013s, Loss: 0.0168, IoU: 0.9699 (val)\n",
      "06-Feb-22 03:26:54 - Ep: [16] TotalT: 20.9 min, BatchT: 0.272s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0170\n",
      "06-Feb-22 03:27:04 - Ep: [16]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.015s, Loss: 0.0193, IoU: 0.9650 (val)\n",
      "06-Feb-22 03:28:12 - Ep: [17] TotalT: 22.2 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0170\n",
      "06-Feb-22 03:28:23 - Ep: [17]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.012s, Loss: 0.0162, IoU: 0.9709 (val)\n",
      "06-Feb-22 03:29:32 - Ep: [18] TotalT: 23.5 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0165\n",
      "06-Feb-22 03:29:42 - Ep: [18]  ValT: 0.17 min, BatchT: 0.141s, DataT: 0.016s, Loss: 0.0166, IoU: 0.9698 (val)\n",
      "06-Feb-22 03:30:50 - Ep: [19] TotalT: 24.8 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0165\n",
      "06-Feb-22 03:31:00 - Ep: [19]  ValT: 0.18 min, BatchT: 0.143s, DataT: 0.014s, Loss: 0.0159, IoU: 0.9711 (val)\n",
      "06-Feb-22 03:32:09 - Ep: [20] TotalT: 26.1 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0158\n",
      "06-Feb-22 03:32:19 - Ep: [20]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0169, IoU: 0.9694 (val)\n",
      "06-Feb-22 03:33:27 - Ep: [21] TotalT: 27.4 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0160\n",
      "06-Feb-22 03:33:38 - Ep: [21]  ValT: 0.18 min, BatchT: 0.146s, DataT: 0.019s, Loss: 0.0165, IoU: 0.9698 (val)\n",
      "06-Feb-22 03:33:38 - LR was reduced to 5.0000e-05\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    15: reducing learning rate of group 0 to 5.0000e-05.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 03:34:46 - Ep: [22] TotalT: 28.8 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0145\n",
      "06-Feb-22 03:34:56 - Ep: [22]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.014s, Loss: 0.0154, IoU: 0.9720 (val)\n",
      "06-Feb-22 03:36:05 - Ep: [23] TotalT: 30.1 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0147\n",
      "06-Feb-22 03:36:15 - Ep: [23]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.012s, Loss: 0.0149, IoU: 0.9730 (val)\n",
      "06-Feb-22 03:37:24 - Ep: [24] TotalT: 31.4 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0144\n",
      "06-Feb-22 03:37:35 - Ep: [24]  ValT: 0.17 min, BatchT: 0.142s, DataT: 0.018s, Loss: 0.0148, IoU: 0.9731 (val)\n",
      "06-Feb-22 03:38:43 - Ep: [25] TotalT: 32.7 min, BatchT: 0.272s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0140\n",
      "06-Feb-22 03:38:53 - Ep: [25]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.013s, Loss: 0.0144, IoU: 0.9740 (val)\n",
      "06-Feb-22 03:40:02 - Ep: [26] TotalT: 34.0 min, BatchT: 0.272s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0140\n",
      "06-Feb-22 03:40:13 - Ep: [26]  ValT: 0.18 min, BatchT: 0.147s, DataT: 0.017s, Loss: 0.0148, IoU: 0.9731 (val)\n",
      "06-Feb-22 03:41:22 - Ep: [27] TotalT: 35.4 min, BatchT: 0.274s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0140\n",
      "06-Feb-22 03:41:32 - Ep: [27]  ValT: 0.18 min, BatchT: 0.143s, DataT: 0.014s, Loss: 0.0148, IoU: 0.9730 (val)\n",
      "06-Feb-22 03:42:41 - Ep: [28] TotalT: 36.7 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0138\n",
      "06-Feb-22 03:42:52 - Ep: [28]  ValT: 0.18 min, BatchT: 0.144s, DataT: 0.015s, Loss: 0.0150, IoU: 0.9725 (val)\n",
      "06-Feb-22 03:44:00 - Ep: [29] TotalT: 38.0 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0133\n",
      "06-Feb-22 03:44:10 - Ep: [29]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.014s, Loss: 0.0146, IoU: 0.9733 (val)\n",
      "06-Feb-22 03:45:19 - Ep: [30] TotalT: 39.3 min, BatchT: 0.272s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0134\n",
      "06-Feb-22 03:45:29 - Ep: [30]  ValT: 0.18 min, BatchT: 0.145s, DataT: 0.018s, Loss: 0.0139, IoU: 0.9747 (val)\n",
      "06-Feb-22 03:46:37 - Ep: [31] TotalT: 40.6 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0134\n",
      "06-Feb-22 03:46:48 - Ep: [31]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.012s, Loss: 0.0140, IoU: 0.9744 (val)\n",
      "06-Feb-22 03:46:48 - LR was reduced to 2.5000e-05\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    25: reducing learning rate of group 0 to 2.5000e-05.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 03:47:56 - Ep: [32] TotalT: 41.9 min, BatchT: 0.273s, DataT: 0.005s, GpuDaT: 0.002s, Loss: 0.0128\n",
      "06-Feb-22 03:48:07 - Ep: [32]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.011s, Loss: 0.0138, IoU: 0.9749 (val)\n",
      "06-Feb-22 03:49:15 - Ep: [33] TotalT: 43.2 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0134\n",
      "06-Feb-22 03:49:25 - Ep: [33]  ValT: 0.18 min, BatchT: 0.142s, DataT: 0.017s, Loss: 0.0135, IoU: 0.9755 (val)\n",
      "06-Feb-22 03:50:33 - Ep: [34] TotalT: 44.5 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0126\n",
      "06-Feb-22 03:50:44 - Ep: [34]  ValT: 0.18 min, BatchT: 0.144s, DataT: 0.017s, Loss: 0.0134, IoU: 0.9756 (val)\n",
      "06-Feb-22 03:51:52 - Ep: [35] TotalT: 45.9 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0127\n",
      "06-Feb-22 03:52:02 - Ep: [35]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.012s, Loss: 0.0131, IoU: 0.9761 (val)\n",
      "06-Feb-22 03:53:11 - Ep: [36] TotalT: 47.2 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0126\n",
      "06-Feb-22 03:53:21 - Ep: [36]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.012s, Loss: 0.0132, IoU: 0.9761 (val)\n",
      "06-Feb-22 03:54:29 - Ep: [37] TotalT: 48.5 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0126\n",
      "06-Feb-22 03:54:40 - Ep: [37]  ValT: 0.18 min, BatchT: 0.143s, DataT: 0.014s, Loss: 0.0135, IoU: 0.9754 (val)\n",
      "06-Feb-22 03:55:49 - Ep: [38] TotalT: 49.8 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0124\n",
      "06-Feb-22 03:55:59 - Ep: [38]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.012s, Loss: 0.0133, IoU: 0.9758 (val)\n",
      "06-Feb-22 03:57:07 - Ep: [39] TotalT: 51.1 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0121\n",
      "06-Feb-22 03:57:18 - Ep: [39]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.013s, Loss: 0.0136, IoU: 0.9751 (val)\n",
      "06-Feb-22 03:57:18 - LR was reduced to 1.2500e-05\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    33: reducing learning rate of group 0 to 1.2500e-05.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 03:58:26 - Ep: [40] TotalT: 52.4 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0124\n",
      "06-Feb-22 03:58:37 - Ep: [40]  ValT: 0.18 min, BatchT: 0.147s, DataT: 0.019s, Loss: 0.0131, IoU: 0.9761 (val)\n",
      "06-Feb-22 03:59:45 - Ep: [41] TotalT: 53.7 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0121\n",
      "06-Feb-22 03:59:56 - Ep: [41]  ValT: 0.18 min, BatchT: 0.143s, DataT: 0.016s, Loss: 0.0132, IoU: 0.9761 (val)\n",
      "06-Feb-22 04:01:04 - Ep: [42] TotalT: 55.1 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0123\n",
      "06-Feb-22 04:01:15 - Ep: [42]  ValT: 0.18 min, BatchT: 0.149s, DataT: 0.021s, Loss: 0.0132, IoU: 0.9760 (val)\n",
      "06-Feb-22 04:02:23 - Ep: [43] TotalT: 56.4 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0121\n",
      "06-Feb-22 04:02:34 - Ep: [43]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.013s, Loss: 0.0137, IoU: 0.9751 (val)\n",
      "06-Feb-22 04:03:42 - Ep: [44] TotalT: 57.7 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0120\n",
      "06-Feb-22 04:03:52 - Ep: [44]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.012s, Loss: 0.0133, IoU: 0.9758 (val)\n",
      "06-Feb-22 04:05:00 - Ep: [45] TotalT: 59.0 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0121\n",
      "06-Feb-22 04:05:10 - Ep: [45]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.012s, Loss: 0.0131, IoU: 0.9761 (val)\n",
      "06-Feb-22 04:05:10 - LR was reduced to 6.2500e-06\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    39: reducing learning rate of group 0 to 6.2500e-06.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 04:06:18 - Ep: [46] TotalT: 60.3 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0119\n",
      "06-Feb-22 04:06:29 - Ep: [46]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.014s, Loss: 0.0129, IoU: 0.9766 (val)\n",
      "06-Feb-22 04:07:37 - Ep: [47] TotalT: 61.6 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0122\n",
      "06-Feb-22 04:07:47 - Ep: [47]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.012s, Loss: 0.0129, IoU: 0.9766 (val)\n",
      "06-Feb-22 04:08:55 - Ep: [48] TotalT: 62.9 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0120\n",
      "06-Feb-22 04:09:05 - Ep: [48]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.015s, Loss: 0.0131, IoU: 0.9762 (val)\n",
      "06-Feb-22 04:10:13 - Ep: [49] TotalT: 64.2 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0119\n",
      "06-Feb-22 04:10:24 - Ep: [49]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.012s, Loss: 0.0130, IoU: 0.9763 (val)\n",
      "06-Feb-22 04:11:32 - Ep: [50] TotalT: 65.5 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0122\n",
      "06-Feb-22 04:11:43 - Ep: [50]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.014s, Loss: 0.0130, IoU: 0.9763 (val)\n",
      "06-Feb-22 04:12:51 - Ep: [51] TotalT: 66.8 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0117\n",
      "06-Feb-22 04:13:02 - Ep: [51]  ValT: 0.17 min, BatchT: 0.141s, DataT: 0.016s, Loss: 0.0130, IoU: 0.9762 (val)\n",
      "06-Feb-22 04:13:02 - LR was reduced to 3.1250e-06\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    45: reducing learning rate of group 0 to 3.1250e-06.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 04:14:10 - Ep: [52] TotalT: 68.2 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0116\n",
      "06-Feb-22 04:14:21 - Ep: [52]  ValT: 0.18 min, BatchT: 0.149s, DataT: 0.020s, Loss: 0.0129, IoU: 0.9765 (val)\n",
      "06-Feb-22 04:15:30 - Ep: [53] TotalT: 69.5 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0117\n",
      "06-Feb-22 04:15:40 - Ep: [53]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.014s, Loss: 0.0129, IoU: 0.9766 (val)\n",
      "06-Feb-22 04:16:48 - Ep: [54] TotalT: 70.8 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0116\n",
      "06-Feb-22 04:16:59 - Ep: [54]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.012s, Loss: 0.0130, IoU: 0.9764 (val)\n",
      "06-Feb-22 04:18:07 - Ep: [55] TotalT: 72.1 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0121\n",
      "06-Feb-22 04:18:18 - Ep: [55]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.013s, Loss: 0.0130, IoU: 0.9763 (val)\n",
      "06-Feb-22 04:19:25 - Ep: [56] TotalT: 73.4 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0120\n",
      "06-Feb-22 04:19:36 - Ep: [56]  ValT: 0.18 min, BatchT: 0.143s, DataT: 0.015s, Loss: 0.0130, IoU: 0.9762 (val)\n",
      "06-Feb-22 04:20:45 - Ep: [57] TotalT: 74.7 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0119\n",
      "06-Feb-22 04:20:55 - Ep: [57]  ValT: 0.18 min, BatchT: 0.142s, DataT: 0.018s, Loss: 0.0130, IoU: 0.9763 (val)\n",
      "06-Feb-22 04:20:55 - LR was reduced to 1.5625e-06\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    51: reducing learning rate of group 0 to 1.5625e-06.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 04:22:03 - Ep: [58] TotalT: 76.0 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0120\n",
      "06-Feb-22 04:22:14 - Ep: [58]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.014s, Loss: 0.0131, IoU: 0.9760 (val)\n",
      "06-Feb-22 04:23:22 - Ep: [59] TotalT: 77.4 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0119\n",
      "06-Feb-22 04:23:32 - Ep: [59]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.014s, Loss: 0.0131, IoU: 0.9761 (val)\n",
      "06-Feb-22 04:23:32 - Early Stopping\n",
      "06-Feb-22 04:23:32 - Best validation combination metric of 0.97660 in epoch 53.\n",
      "06-Feb-22 04:23:32 - Training with 1 GPUS\n",
      "06-Feb-22 04:23:32 - Training is done with mixed precision\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Data Augmentations: [RandomCrop(always_apply=False, p=1.0, height=384, width=384)]\n",
      "Train Image Data Augmentations: []\n",
      "####################################################################################################\n",
      "####################################################################################################\n",
      "Fold 4 --> Train: 9399, Val: 2349\n",
      "####################################################################################################\n",
      "Last 5 sample weights: [1.0, 1.0, 1.0, 1.0, 1.0]\n",
      "unet timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 04:24:40 - Ep: [1] TotalT: 1.1 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.1128\n",
      "06-Feb-22 04:24:51 - Ep: [1]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.012s, Loss: 0.0490, IoU: 0.9389 (val)\n",
      "06-Feb-22 04:25:59 - Ep: [2] TotalT: 2.4 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0451\n",
      "06-Feb-22 04:26:09 - Ep: [2]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.014s, Loss: 0.0346, IoU: 0.9478 (val)\n",
      "06-Feb-22 04:27:18 - Ep: [3] TotalT: 3.8 min, BatchT: 0.273s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0360\n",
      "06-Feb-22 04:27:28 - Ep: [3]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.015s, Loss: 0.0311, IoU: 0.9506 (val)\n",
      "06-Feb-22 04:28:37 - Ep: [4] TotalT: 5.1 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0304\n",
      "06-Feb-22 04:28:47 - Ep: [4]  ValT: 0.17 min, BatchT: 0.141s, DataT: 0.013s, Loss: 0.0274, IoU: 0.9549 (val)\n",
      "06-Feb-22 04:29:56 - Ep: [5] TotalT: 6.4 min, BatchT: 0.273s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0260\n",
      "06-Feb-22 04:30:07 - Ep: [5]  ValT: 0.18 min, BatchT: 0.144s, DataT: 0.016s, Loss: 0.0263, IoU: 0.9568 (val)\n",
      "06-Feb-22 04:31:15 - Ep: [6] TotalT: 7.7 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0255\n",
      "06-Feb-22 04:31:26 - Ep: [6]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.015s, Loss: 0.0234, IoU: 0.9605 (val)\n",
      "06-Feb-22 04:32:34 - Ep: [7] TotalT: 9.0 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0222\n",
      "06-Feb-22 04:32:45 - Ep: [7]  ValT: 0.18 min, BatchT: 0.146s, DataT: 0.017s, Loss: 0.0215, IoU: 0.9630 (val)\n",
      "06-Feb-22 04:33:54 - Ep: [8] TotalT: 10.4 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0219\n",
      "06-Feb-22 04:34:04 - Ep: [8]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.011s, Loss: 0.0231, IoU: 0.9600 (val)\n",
      "06-Feb-22 04:35:13 - Ep: [9] TotalT: 11.7 min, BatchT: 0.273s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0210\n",
      "06-Feb-22 04:35:23 - Ep: [9]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.012s, Loss: 0.0209, IoU: 0.9638 (val)\n",
      "06-Feb-22 04:36:31 - Ep: [10] TotalT: 13.0 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0196\n",
      "06-Feb-22 04:36:42 - Ep: [10]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.012s, Loss: 0.0201, IoU: 0.9650 (val)\n",
      "06-Feb-22 04:37:50 - Ep: [11] TotalT: 14.3 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0187\n",
      "06-Feb-22 04:38:00 - Ep: [11]  ValT: 0.18 min, BatchT: 0.142s, DataT: 0.014s, Loss: 0.0205, IoU: 0.9644 (val)\n",
      "06-Feb-22 04:39:09 - Ep: [12] TotalT: 15.6 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0185\n",
      "06-Feb-22 04:39:19 - Ep: [12]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.012s, Loss: 0.0196, IoU: 0.9655 (val)\n",
      "06-Feb-22 04:40:27 - Ep: [13] TotalT: 16.9 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0180\n",
      "06-Feb-22 04:40:37 - Ep: [13]  ValT: 0.17 min, BatchT: 0.141s, DataT: 0.014s, Loss: 0.0183, IoU: 0.9677 (val)\n",
      "06-Feb-22 04:41:46 - Ep: [14] TotalT: 18.2 min, BatchT: 0.270s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0179\n",
      "06-Feb-22 04:41:56 - Ep: [14]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.013s, Loss: 0.0180, IoU: 0.9681 (val)\n",
      "06-Feb-22 04:43:04 - Ep: [15] TotalT: 19.5 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0176\n",
      "06-Feb-22 04:43:15 - Ep: [15]  ValT: 0.18 min, BatchT: 0.143s, DataT: 0.017s, Loss: 0.0181, IoU: 0.9680 (val)\n",
      "06-Feb-22 04:44:23 - Ep: [16] TotalT: 20.8 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0176\n",
      "06-Feb-22 04:44:33 - Ep: [16]  ValT: 0.17 min, BatchT: 0.140s, DataT: 0.016s, Loss: 0.0192, IoU: 0.9662 (val)\n",
      "06-Feb-22 04:45:41 - Ep: [17] TotalT: 22.1 min, BatchT: 0.271s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0167\n",
      "06-Feb-22 04:45:52 - Ep: [17]  ValT: 0.17 min, BatchT: 0.141s, DataT: 0.015s, Loss: 0.0185, IoU: 0.9669 (val)\n",
      "06-Feb-22 04:47:01 - Ep: [18] TotalT: 23.5 min, BatchT: 0.274s, DataT: 0.005s, GpuDaT: 0.002s, Loss: 0.0160\n",
      "06-Feb-22 04:47:11 - Ep: [18]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.013s, Loss: 0.0173, IoU: 0.9688 (val)\n",
      "06-Feb-22 04:48:20 - Ep: [19] TotalT: 24.8 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0161\n",
      "06-Feb-22 04:48:30 - Ep: [19]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.013s, Loss: 0.0194, IoU: 0.9651 (val)\n",
      "06-Feb-22 04:48:30 - LR was reduced to 5.0000e-05\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    13: reducing learning rate of group 0 to 5.0000e-05.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 04:49:38 - Ep: [20] TotalT: 26.1 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0154\n",
      "06-Feb-22 04:49:49 - Ep: [20]  ValT: 0.18 min, BatchT: 0.145s, DataT: 0.014s, Loss: 0.0172, IoU: 0.9694 (val)\n",
      "06-Feb-22 04:50:58 - Ep: [21] TotalT: 27.4 min, BatchT: 0.272s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0150\n",
      "06-Feb-22 04:51:08 - Ep: [21]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.015s, Loss: 0.0157, IoU: 0.9719 (val)\n",
      "06-Feb-22 04:52:16 - Ep: [22] TotalT: 28.7 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0143\n",
      "06-Feb-22 04:52:27 - Ep: [22]  ValT: 0.18 min, BatchT: 0.143s, DataT: 0.015s, Loss: 0.0156, IoU: 0.9721 (val)\n",
      "06-Feb-22 04:53:37 - Ep: [23] TotalT: 30.1 min, BatchT: 0.276s, DataT: 0.005s, GpuDaT: 0.002s, Loss: 0.0150\n",
      "06-Feb-22 04:53:47 - Ep: [23]  ValT: 0.17 min, BatchT: 0.139s, DataT: 0.016s, Loss: 0.0154, IoU: 0.9725 (val)\n",
      "06-Feb-22 04:54:55 - Ep: [24] TotalT: 31.4 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0148\n",
      "06-Feb-22 04:55:06 - Ep: [24]  ValT: 0.18 min, BatchT: 0.144s, DataT: 0.019s, Loss: 0.0152, IoU: 0.9728 (val)\n",
      "06-Feb-22 04:56:14 - Ep: [25] TotalT: 32.7 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0141\n",
      "06-Feb-22 04:56:25 - Ep: [25]  ValT: 0.18 min, BatchT: 0.142s, DataT: 0.014s, Loss: 0.0152, IoU: 0.9730 (val)\n",
      "06-Feb-22 04:57:34 - Ep: [26] TotalT: 34.0 min, BatchT: 0.274s, DataT: 0.005s, GpuDaT: 0.003s, Loss: 0.0141\n",
      "06-Feb-22 04:57:44 - Ep: [26]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.014s, Loss: 0.0159, IoU: 0.9717 (val)\n",
      "06-Feb-22 04:58:52 - Ep: [27] TotalT: 35.3 min, BatchT: 0.272s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0137\n",
      "06-Feb-22 04:59:03 - Ep: [27]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.013s, Loss: 0.0154, IoU: 0.9725 (val)\n",
      "06-Feb-22 04:59:03 - LR was reduced to 2.5000e-05\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    21: reducing learning rate of group 0 to 2.5000e-05.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 05:00:11 - Ep: [28] TotalT: 36.6 min, BatchT: 0.271s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0138\n",
      "06-Feb-22 05:00:21 - Ep: [28]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.015s, Loss: 0.0148, IoU: 0.9736 (val)\n",
      "06-Feb-22 05:01:29 - Ep: [29] TotalT: 37.9 min, BatchT: 0.270s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0134\n",
      "06-Feb-22 05:01:39 - Ep: [29]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0151, IoU: 0.9728 (val)\n",
      "06-Feb-22 05:02:47 - Ep: [30] TotalT: 39.2 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0132\n",
      "06-Feb-22 05:02:57 - Ep: [30]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.015s, Loss: 0.0152, IoU: 0.9729 (val)\n",
      "06-Feb-22 05:04:05 - Ep: [31] TotalT: 40.5 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0131\n",
      "06-Feb-22 05:04:15 - Ep: [31]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.012s, Loss: 0.0149, IoU: 0.9734 (val)\n",
      "06-Feb-22 05:05:22 - Ep: [32] TotalT: 41.8 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0129\n",
      "06-Feb-22 05:05:32 - Ep: [32]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.012s, Loss: 0.0151, IoU: 0.9730 (val)\n",
      "06-Feb-22 05:06:39 - Ep: [33] TotalT: 43.1 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0131\n",
      "06-Feb-22 05:06:50 - Ep: [33]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.012s, Loss: 0.0150, IoU: 0.9731 (val)\n",
      "06-Feb-22 05:07:57 - Ep: [34] TotalT: 44.4 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0129\n",
      "06-Feb-22 05:08:07 - Ep: [34]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0149, IoU: 0.9735 (val)\n",
      "06-Feb-22 05:08:07 - LR was reduced to 1.2500e-05\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    28: reducing learning rate of group 0 to 1.2500e-05.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 05:09:15 - Ep: [35] TotalT: 45.7 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0128\n",
      "06-Feb-22 05:09:25 - Ep: [35]  ValT: 0.17 min, BatchT: 0.134s, DataT: 0.011s, Loss: 0.0145, IoU: 0.9738 (val)\n",
      "06-Feb-22 05:10:32 - Ep: [36] TotalT: 47.0 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0129\n",
      "06-Feb-22 05:10:42 - Ep: [36]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0147, IoU: 0.9736 (val)\n",
      "06-Feb-22 05:11:50 - Ep: [37] TotalT: 48.3 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0123\n",
      "06-Feb-22 05:12:00 - Ep: [37]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.014s, Loss: 0.0148, IoU: 0.9734 (val)\n",
      "06-Feb-22 05:13:07 - Ep: [38] TotalT: 49.6 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0121\n",
      "06-Feb-22 05:13:18 - Ep: [38]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.013s, Loss: 0.0146, IoU: 0.9739 (val)\n",
      "06-Feb-22 05:14:25 - Ep: [39] TotalT: 50.9 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0126\n",
      "06-Feb-22 05:14:35 - Ep: [39]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.012s, Loss: 0.0146, IoU: 0.9737 (val)\n",
      "06-Feb-22 05:15:42 - Ep: [40] TotalT: 52.2 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0124\n",
      "06-Feb-22 05:15:52 - Ep: [40]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0147, IoU: 0.9736 (val)\n",
      "06-Feb-22 05:15:52 - LR was reduced to 6.2500e-06\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    34: reducing learning rate of group 0 to 6.2500e-06.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 05:17:00 - Ep: [41] TotalT: 53.5 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0120\n",
      "06-Feb-22 05:17:10 - Ep: [41]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.014s, Loss: 0.0146, IoU: 0.9739 (val)\n",
      "06-Feb-22 05:18:18 - Ep: [42] TotalT: 54.8 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0122\n",
      "06-Feb-22 05:18:28 - Ep: [42]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.014s, Loss: 0.0146, IoU: 0.9739 (val)\n",
      "06-Feb-22 05:19:36 - Ep: [43] TotalT: 56.1 min, BatchT: 0.269s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0126\n",
      "06-Feb-22 05:19:46 - Ep: [43]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.014s, Loss: 0.0147, IoU: 0.9735 (val)\n",
      "06-Feb-22 05:20:54 - Ep: [44] TotalT: 57.4 min, BatchT: 0.269s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0122\n",
      "06-Feb-22 05:21:04 - Ep: [44]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.013s, Loss: 0.0145, IoU: 0.9739 (val)\n",
      "06-Feb-22 05:22:11 - Ep: [45] TotalT: 58.6 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0125\n",
      "06-Feb-22 05:22:21 - Ep: [45]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0146, IoU: 0.9738 (val)\n",
      "06-Feb-22 05:23:29 - Ep: [46] TotalT: 59.9 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0122\n",
      "06-Feb-22 05:23:39 - Ep: [46]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.013s, Loss: 0.0146, IoU: 0.9737 (val)\n",
      "06-Feb-22 05:23:39 - LR was reduced to 3.1250e-06\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    40: reducing learning rate of group 0 to 3.1250e-06.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 05:24:46 - Ep: [47] TotalT: 61.2 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0123\n",
      "06-Feb-22 05:24:56 - Ep: [47]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.012s, Loss: 0.0144, IoU: 0.9742 (val)\n",
      "06-Feb-22 05:26:03 - Ep: [48] TotalT: 62.5 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0123\n",
      "06-Feb-22 05:26:14 - Ep: [48]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0145, IoU: 0.9740 (val)\n",
      "06-Feb-22 05:27:21 - Ep: [49] TotalT: 63.8 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0120\n",
      "06-Feb-22 05:27:31 - Ep: [49]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.014s, Loss: 0.0142, IoU: 0.9744 (val)\n",
      "06-Feb-22 05:28:38 - Ep: [50] TotalT: 65.1 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0122\n",
      "06-Feb-22 05:28:48 - Ep: [50]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0143, IoU: 0.9743 (val)\n",
      "06-Feb-22 05:29:55 - Ep: [51] TotalT: 66.4 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0123\n",
      "06-Feb-22 05:30:06 - Ep: [51]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.012s, Loss: 0.0146, IoU: 0.9737 (val)\n",
      "06-Feb-22 05:31:13 - Ep: [52] TotalT: 67.7 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0119\n",
      "06-Feb-22 05:31:23 - Ep: [52]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.014s, Loss: 0.0145, IoU: 0.9739 (val)\n",
      "06-Feb-22 05:31:23 - LR was reduced to 1.5625e-06\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    46: reducing learning rate of group 0 to 1.5625e-06.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 05:32:30 - Ep: [53] TotalT: 69.0 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0122\n",
      "06-Feb-22 05:32:40 - Ep: [53]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.012s, Loss: 0.0144, IoU: 0.9741 (val)\n",
      "06-Feb-22 05:33:47 - Ep: [54] TotalT: 70.2 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0122\n",
      "06-Feb-22 05:33:58 - Ep: [54]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.014s, Loss: 0.0145, IoU: 0.9740 (val)\n",
      "06-Feb-22 05:35:05 - Ep: [55] TotalT: 71.5 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0119\n",
      "06-Feb-22 05:35:15 - Ep: [55]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0147, IoU: 0.9737 (val)\n",
      "06-Feb-22 05:36:22 - Ep: [56] TotalT: 72.8 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0122\n",
      "06-Feb-22 05:36:32 - Ep: [56]  ValT: 0.17 min, BatchT: 0.135s, DataT: 0.012s, Loss: 0.0145, IoU: 0.9740 (val)\n",
      "06-Feb-22 05:37:39 - Ep: [57] TotalT: 74.1 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0119\n",
      "06-Feb-22 05:37:49 - Ep: [57]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.013s, Loss: 0.0145, IoU: 0.9740 (val)\n",
      "06-Feb-22 05:38:57 - Ep: [58] TotalT: 75.4 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0121\n",
      "06-Feb-22 05:39:07 - Ep: [58]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0144, IoU: 0.9742 (val)\n",
      "06-Feb-22 05:39:07 - LR was reduced to 7.8125e-07\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch    52: reducing learning rate of group 0 to 7.8125e-07.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "06-Feb-22 05:40:14 - Ep: [59] TotalT: 76.7 min, BatchT: 0.266s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0122\n",
      "06-Feb-22 05:40:24 - Ep: [59]  ValT: 0.17 min, BatchT: 0.137s, DataT: 0.014s, Loss: 0.0147, IoU: 0.9738 (val)\n",
      "06-Feb-22 05:41:31 - Ep: [60] TotalT: 78.0 min, BatchT: 0.267s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0120\n",
      "06-Feb-22 05:41:41 - Ep: [60]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.012s, Loss: 0.0144, IoU: 0.9741 (val)\n",
      "06-Feb-22 05:42:49 - Ep: [61] TotalT: 79.3 min, BatchT: 0.267s, DataT: 0.004s, GpuDaT: 0.002s, Loss: 0.0120\n",
      "06-Feb-22 05:42:59 - Ep: [61]  ValT: 0.17 min, BatchT: 0.138s, DataT: 0.015s, Loss: 0.0151, IoU: 0.9731 (val)\n",
      "06-Feb-22 05:44:06 - Ep: [62] TotalT: 80.6 min, BatchT: 0.268s, DataT: 0.003s, GpuDaT: 0.002s, Loss: 0.0119\n",
      "06-Feb-22 05:44:16 - Ep: [62]  ValT: 0.17 min, BatchT: 0.136s, DataT: 0.013s, Loss: 0.0146, IoU: 0.9738 (val)\n",
      "06-Feb-22 05:44:16 - Early Stopping\n",
      "06-Feb-22 05:44:16 - Best validation combination metric of 0.97441 in epoch 49.\n"
     ]
    }
   ],
   "source": [
    "exp_nb = \"7-4\"\n",
    "path = 'data/trainval_2022_02_05.pkl' #'data/trainval_2021_12_31.pkl'\n",
    "\n",
    "hps_dict = {\n",
    "    ############\n",
    "    # Data\n",
    "    ############\n",
    "    \"df_path\": path,\n",
    "\n",
    "    ############\n",
    "    # Training\n",
    "    ############\n",
    "\n",
    "    ## Experiment Setup\n",
    "    \"name\": f\"Exp{exp_nb}\",\n",
    "\n",
    "    ## Model\n",
    "    \"num_classes\": 1,\n",
    "    \"input_channel\": 4,\n",
    "    \"backbone\": \"timm_efficientnet_b1\",\n",
    "    \"pretrained\": 1,\n",
    "    \"model\": \"unet\",\n",
    "\n",
    "    # Training Setup\n",
    "#     \"resume\": \"trained_models/Exp7-4/fold_0/2022-02-05_23-13-25/best_metric_18_0.9768.pt\",\n",
    "    \"print_freq\": 1000,\n",
    "    \"use_fp16\": 1,\n",
    "    \"patience\": 5,\n",
    "\n",
    "    # Optimizer\n",
    "    \"lr\": 0.0001,\n",
    "    \"weight_decay\": 0.0,\n",
    "\n",
    "    ## Data Augmentation on CPU\n",
    "    \"train_crop_size\": 384,\n",
    "    \"train_batch_size\": 32,\n",
    "    \"cutmix_alpha\": 0,\n",
    "    \"da_brightness_magnitude\": 0.0,\n",
    "    \"da_contrast_magnitude\": 0.0,\n",
    "\n",
    "    # Data Augmentation on GPU\n",
    "    \"gpu_da_params\": [0.25],\n",
    "    \n",
    "    ### Loss, Metric\n",
    "    \"alpha\": 0,\n",
    "#     \"loss\": \"lovasz\",\n",
    "           }\n",
    "\n",
    "for fold_nb in range(5):\n",
    "    hps = HyperParams(**hps_dict)\n",
    "    hps.fold_nb = fold_nb\n",
    "    num_batches = 8000 // (32 * torch.cuda.device_count())\n",
    "    hps.num_batches = num_batches\n",
    "    train_dataset, train_loader, val_dataset, val_loader = get_dataloaders(hps)\n",
    "#     continue\n",
    "\n",
    "    best_metric, best_metric_epoch = train(hps, train_loader, val_loader)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "marked-dinner",
   "metadata": {},
   "source": [
    "# Validate"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "indirect-adventure",
   "metadata": {},
   "source": [
    "## Get best threshold"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "reliable-toronto",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-02-05T22:09:55.204155Z",
     "start_time": "2022-02-05T22:09:55.172435Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "HyperParams(df_path='data/trainval_2022_01_03_2.pkl', fold_nb=0, only_val=False, train_batch_size=32, val_batch_size=32, train_crop_size=384, randomsizecrop_interpolation=1, da_p_cutout=0.0, cutmix_alpha=0.0, random_sized_params=[0.0, 0.0], da_brightness_magnitude=0.0, da_contrast_magnitude=0.0, name='Exp7-3', num_classes=1, input_channel=4, extra_bands=[], smp_decoder_use_batchnorm=True, smp_decoder_channels_mult=1.0, smp_decoder_use_attention=False, backbone='timm_efficientnet_b1', pretrained=1, model='unetplusplus', n_epochs=1000, use_fp16=1, patience=4, num_batches=250, print_freq=1000, weight_decay=0.0, lr=0.001, gpu_da_params=[0.25], alpha=0.5)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "hps = HyperParams(**open_from_yaml(\"trained_models/Exp7-3\"))\n",
    "hps"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "inclusive-return",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-02-06T09:18:35.103472Z",
     "start_time": "2022-02-06T09:17:10.233803Z"
    },
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "data/trainval_2021_12_20.pkl\n",
      "unet timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp12-0/fold_0/2022-02-06_00-24-24/best_metric_31_0.9111.pt at epoch 31\n",
      "unet timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp12-0/fold_1/2022-02-06_01-30-34/best_metric_29_0.9221.pt at epoch 29\n",
      "unet timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp12-0/fold_2/2022-02-06_02-36-46/best_metric_31_0.9129.pt at epoch 31\n",
      "unet timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp12-0/fold_3/2022-02-06_03-42-36/best_metric_31_0.9122.pt at epoch 31\n",
      "unet timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp12-0/fold_4/2022-02-06_04-46-48/best_metric_29_0.9155.pt at epoch 29\n",
      "####################################################################################################\n",
      "Performance @ t = 0.30 (['Exp12-0'])\n",
      "====================================================================================================\n",
      "recall       0.977226\n",
      "precision    0.915227\n",
      "IoU          0.896113\n",
      "dtype: float64\n",
      "####################################################################################################\n",
      "Performance @ t = 0.35 (['Exp12-0'])\n",
      "====================================================================================================\n",
      "recall       0.973477\n",
      "precision    0.926139\n",
      "IoU          0.903345\n",
      "dtype: float64\n",
      "####################################################################################################\n",
      "Performance @ t = 0.40 (['Exp12-0'])\n",
      "====================================================================================================\n",
      "recall       0.969452\n",
      "precision    0.935893\n",
      "IoU          0.909084\n",
      "dtype: float64\n",
      "####################################################################################################\n",
      "Performance @ t = 0.45 (['Exp12-0'])\n",
      "====================================================================================================\n",
      "recall       0.964641\n",
      "precision    0.944524\n",
      "IoU          0.912917\n",
      "dtype: float64\n",
      "####################################################################################################\n",
      "Performance @ t = 0.50 (['Exp12-0'])\n",
      "====================================================================================================\n",
      "recall       0.959407\n",
      "precision    0.951589\n",
      "IoU          0.914758\n",
      "dtype: float64\n",
      "####################################################################################################\n",
      "Performance @ t = 0.55 (['Exp12-0'])\n",
      "====================================================================================================\n",
      "recall       0.953437\n",
      "precision    0.957105\n",
      "IoU          0.914365\n",
      "dtype: float64\n",
      "####################################################################################################\n",
      "Performance @ t = 0.60 (['Exp12-0'])\n",
      "====================================================================================================\n",
      "recall       0.947130\n",
      "precision    0.961160\n",
      "IoU          0.912217\n",
      "dtype: float64\n",
      "####################################################################################################\n",
      "Performance @ t = 0.50 (['Exp12-0'])\n",
      "====================================================================================================\n",
      "recall       0.959407\n",
      "precision    0.951589\n",
      "IoU          0.914758\n",
      "dtype: float64\n",
      "====================================================================================================\n",
      "Best IoU of @0.50: 0.9148\n"
     ]
    }
   ],
   "source": [
    "hps = HyperParams(**open_from_yaml(\"trained_models/Exp12-0\"))\n",
    "print(hps.df_path)\n",
    "df, stats, best_threshold = cross_validation([hps], thresholds=[0.3, 0.35, 0.4, 0.45, 0.5, 0.55, 0.6], debug=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "incomplete-letters",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-21T08:08:42.781449Z",
     "start_time": "2022-01-21T08:05:52.360284Z"
    },
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp9-0/fold_0/2022-01-20_19-01-20/best_metric_40_0.9677.pt at epoch 40\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp9-1/fold_0/2022-01-20_11-22-02/best_metric_46_0.9667.pt at epoch 46\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp9-0/fold_1/2022-01-20_20-28-12/best_metric_32_0.9669.pt at epoch 32\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp9-1/fold_1/2022-01-20_12-50-05/best_metric_43_0.9664.pt at epoch 43\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp9-0/fold_2/2022-01-20_21-41-48/best_metric_30_0.9679.pt at epoch 30\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp9-1/fold_2/2022-01-20_14-23-16/best_metric_43_0.9674.pt at epoch 43\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp9-0/fold_3/2022-01-20_22-51-56/best_metric_42_0.9677.pt at epoch 42\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp9-1/fold_3/2022-01-20_15-52-37/best_metric_47_0.9669.pt at epoch 47\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp9-0/fold_4/2022-01-21_00-07-16/best_metric_39_0.9687.pt at epoch 39\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp9-1/fold_4/2022-01-20_17-32-06/best_metric_51_0.9682.pt at epoch 51\n",
      "####################################################################################################\n",
      "Performance @ t = 0.5 (['Exp9-0', 'Exp9-1'])\n",
      "====================================================================================================\n",
      "recall       0.984082\n",
      "precision    0.983686\n",
      "IoU          0.968280\n",
      "dtype: float64\n",
      "====================================================================================================\n",
      "Best IoU of @0.5: 0.9683\n"
     ]
    }
   ],
   "source": [
    "hps_list = []\n",
    "hps_list.append(HyperParams(**open_from_yaml(\"trained_models/Exp9-0\")))\n",
    "hps_list.append(HyperParams(**open_from_yaml(\"trained_models/Exp9-1\")))\n",
    "df, stats, best_threshold = cross_validation(hps_list, debug=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "abroad-framework",
   "metadata": {},
   "source": [
    "## Visualize"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "numerous-mattress",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-27T20:28:48.228534Z",
     "start_time": "2022-01-27T20:28:48.183400Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Exp9-1\n"
     ]
    }
   ],
   "source": [
    "hps_list = []\n",
    "hps = HyperParams(**open_from_yaml(\"trained_models/Exp9-1\"))\n",
    "hps_list.append(hps)\n",
    "df = pd.read_pickle(hps.df_path)\n",
    "\n",
    "pred_name = \"_\".join([hps_.name for hps_ in hps_list])\n",
    "print(pred_name)\n",
    "threshold = 0.5 #0.8 #0.3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "typical-montana",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-27T20:30:04.485808Z",
     "start_time": "2022-01-27T20:28:48.229869Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp9-1/fold_0/2022-01-20_11-22-02/best_metric_46_0.9667.pt at epoch 46\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp9-1/fold_1/2022-01-20_12-50-05/best_metric_43_0.9664.pt at epoch 43\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp9-1/fold_2/2022-01-20_14-23-16/best_metric_43_0.9674.pt at epoch 43\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp9-1/fold_3/2022-01-20_15-52-37/best_metric_47_0.9669.pt at epoch 47\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Loaded trained_models/Exp9-1/fold_4/2022-01-20_17-32-06/best_metric_51_0.9682.pt at epoch 51\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.8/site-packages/numpy/lib/npyio.py:528: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray.\n",
      "  arr = np.asanyarray(arr)\n"
     ]
    }
   ],
   "source": [
    "also_save_preds = True\n",
    "\n",
    "subset = \"val_all_folds\"\n",
    "export = return_metrics_all_folds(hps_list, threshold=threshold, on_gpu=True, output_spatial_size=(512, 512), also_save_preds=also_save_preds,\n",
    "                                  save_path=f\"/data/C2S_CNN/predictions/{pred_name}_{subset}.bc\")\n",
    "export_path = f\"/data/C2S_CNN/predictions/metrics_{pred_name}_{subset}.npy\"\n",
    "np.save(export_path, export)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "precise-ethernet",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-27T20:47:02.272593Z",
     "start_time": "2022-01-27T20:47:02.227931Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Exp9-0_Exp9-1\n"
     ]
    }
   ],
   "source": [
    "hps_list = []\n",
    "hps = HyperParams(**open_from_yaml(\"trained_models/Exp9-1\"))\n",
    "hps_list.append(hps)\n",
    "df = pd.read_pickle(hps.df_path)\n",
    "\n",
    "pred_name = \"_\".join([hps_.name for hps_ in hps_list])\n",
    "print(pred_name)\n",
    "threshold = 0.5 #0.8 #0.3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "elder-october",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-27T20:47:02.482317Z",
     "start_time": "2022-01-27T20:47:02.379055Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(11748, 512, 512)\n",
      "11748 11748\n"
     ]
    }
   ],
   "source": [
    "subset = \"val_all_folds\"\n",
    "\n",
    "export_path = f\"/data/C2S_CNN/predictions/metrics_{pred_name}_{subset}.npy\"\n",
    "pred_path = f\"/data/C2S_CNN/predictions/{pred_name}_{subset}.bc\"\n",
    "preds = bcolz.open(pred_path)\n",
    "\n",
    "_, tps, fps, fns = np.load(export_path, allow_pickle=True)\n",
    "print(preds.shape)\n",
    "    \n",
    "img_paths, GT_paths, biomes = [], [], []\n",
    "hps.only_val = True\n",
    "for fold_nb in range(5):    \n",
    "    hps.fold_nb = fold_nb\n",
    "    val_dataset, val_loader = get_dataloaders(hps=hps)\n",
    "    img_paths.extend(val_dataset.img_paths)\n",
    "    GT_paths.extend(val_dataset.mask_paths)\n",
    "\n",
    "print(len(img_paths), len(GT_paths))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "worthy-student",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-27T21:00:30.854871Z",
     "start_time": "2022-01-27T21:00:30.827921Z"
    }
   },
   "outputs": [],
   "source": [
    "preds_list = [bcolz.open(f\"/data/C2S_CNN/predictions/Exp9-0_{subset}.bc\"), bcolz.open(f\"/data/C2S_CNN/predictions/Exp9-1_{subset}.bc\")]\n",
    "\n",
    "tp_col, fp_col, fn_col, pred_stds_015 = [], [], [], []\n",
    "for img_path in tqdm(df[\"B02_path\"].tolist()):\n",
    "    i = img_paths.index(img_path)\n",
    "    tp_col.append(tps[i])\n",
    "    fp_col.append(fps[i])\n",
    "    fn_col.append(fns[i])\n",
    "    pred = np.array([(preds[i].astype(np.float32) / 255.0) for preds in preds_list])\n",
    "    pred_std = np.std(pred, axis=0)\n",
    "    pred_stds_015.append((pred_std > 0.15).sum())\n",
    "\n",
    "df[\"tps\"] = tp_col\n",
    "df[\"fps\"] = fp_col\n",
    "df[\"fns\"] = fn_col\n",
    "df[\"pred_std_015_sum\"] = pred_stds_015\n",
    "\n",
    "calc_more_columns = True\n",
    "results = df.groupby(\"location\").apply(lambda grp: calc_ious(grp, calc_more_columns=calc_more_columns))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "growing-grammar",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-27T21:05:59.423143Z",
     "start_time": "2022-01-27T21:05:59.397892Z"
    },
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "selection_criteria = \"FP+FN\" #\"FP+FN\" #\"FP+FN ratio\" #\"(FP+FN)/TP ratio\" #\"FP+FN\" #\"random\" #\"FP+FN\" #\"random\" \n",
    "specific_location = 0 #\"Senanga\" #\"Matadi\" #\"South America - Argentina\" #0 #\"Santa Cruz do Sul\" #\"Timbuktu\" #0 #\"Eritrea\" #0\n",
    "\n",
    "looked_at_paths, looked_at_idx = visualize_single_model(\n",
    "    df, tps, fps, fns, preds_list, threshold, img_paths, GT_paths, export_how_many=150, skip_how_many=0, alpha=0.4, selection_criteria=selection_criteria, specific_location=specific_location)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "judicial-skirt",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-21T09:11:16.324168Z",
     "start_time": "2022-01-21T09:11:16.260274Z"
    },
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['data/train_features/dlmw/B02.tif',\n",
       " 'data/train_features/qjkg/B02.tif',\n",
       " 'data/train_features/yuji/B02.tif',\n",
       " 'data/train_features/iimu/B02.tif',\n",
       " 'data/train_features/tgdi/B02.tif',\n",
       " 'data/train_features/dfdi/B02.tif',\n",
       " 'data/train_features/dget/B02.tif',\n",
       " 'data/train_features/lcrt/B02.tif',\n",
       " 'data/train_features/rrdb/B02.tif',\n",
       " 'data/train_features/vdjb/B02.tif',\n",
       " 'data/train_features/agtf/B02.tif',\n",
       " 'data/train_features/pctp/B02.tif']"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "idx_to_match = [420, 8504, 10706, 7425, 7417, 2278, 7993, 9420, 1418, 8806, 6399, 5481, ]\n",
    "[path for idx, path in enumerate(img_paths) if idx in idx_to_match]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "postal-analyst",
   "metadata": {},
   "source": [
    "## Overwrite wrong labels with predictions"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "frequent-reunion",
   "metadata": {},
   "source": [
    "### v1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "pressed-education",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-02-05T22:10:06.654945Z",
     "start_time": "2022-02-05T22:10:06.557815Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Exp0\n",
      "(11745, 512, 512)\n",
      "11745 11745\n"
     ]
    }
   ],
   "source": [
    "hps_list = []\n",
    "hps_dict = open_from_yaml(\"trained_models/Exp0\")\n",
    "hps_dict.pop(\"gamma\")\n",
    "hps = HyperParams(**hps_dict)\n",
    "hps_list.append(hps)\n",
    "df = pd.read_pickle(hps.df_path)\n",
    "\n",
    "threshold = 0.5 #0.3 #0.8 #0.3\n",
    "df = pd.read_pickle(hps.df_path)\n",
    "df = df.reset_index(drop=True)\n",
    "# img_paths = df[\"B02_path\"].tolist()\n",
    "\n",
    "pred_name = \"_\".join([hps_.name for hps_ in hps_list])\n",
    "print(pred_name)\n",
    "subset = \"val_all_folds\"\n",
    "\n",
    "pred_path = f\"/data/C2S_CNN/predictions/{pred_name}_{subset}.bc\"\n",
    "preds = bcolz.open(pred_path)\n",
    "print(preds.shape)\n",
    "\n",
    "img_paths, GT_paths, biomes = [], [], []\n",
    "hps.only_val = True\n",
    "for fold_nb in range(5):    \n",
    "    hps.fold_nb = fold_nb\n",
    "    val_dataset, val_loader = get_dataloaders(hps=hps)\n",
    "    img_paths.extend(val_dataset.img_paths)\n",
    "    GT_paths.extend(val_dataset.mask_paths)\n",
    "\n",
    "print(len(img_paths), len(GT_paths))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "bibliographic-campus",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-02-05T22:10:08.250004Z",
     "start_time": "2022-02-05T22:10:07.169057Z"
    },
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "empty     : Replaced 18 labels\n",
      "full      : Replaced 0 labels\n",
      "use_pred  : Replaced 449 labels\n"
     ]
    }
   ],
   "source": [
    "relabel_based_on_preds(relabeling_v1, img_paths, GT_paths, preds, threshold=threshold, debug=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "biological-reunion",
   "metadata": {},
   "source": [
    "# Submit"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "otherwise-motor",
   "metadata": {},
   "source": [
    "## Export models to jit and copy jit paths"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "unique-category",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-28T16:37:14.579359Z",
     "start_time": "2022-01-28T16:37:14.549215Z"
    }
   },
   "outputs": [],
   "source": [
    "def export_to_jit(experiment_yaml_path,  model_export_root=\"benchmark_src/assets\"):\n",
    "    \"\"\"Traces and exports a PyTorch model to jit\"\"\"\n",
    "    hps = HyperParams(**open_from_yaml(experiment_yaml_path))\n",
    "    hps.pretrained = False\n",
    "    experiment_name = experiment_yaml_path.split(\"/\")[-4]\n",
    "    fold_nb = experiment_yaml_path.split(\"/\")[-3].split(\"_\")[-1]\n",
    "    example_tensor = torch.rand(2, hps.input_channel, 256, 256).cuda()\n",
    "    model = build_model(hps)\n",
    "    model = model.cuda()\n",
    "    model.eval()\n",
    "    model_weights_paths = glob.glob(f\"{'/'.join(experiment_yaml_path.split('/')[:-1])}/*.pt\")\n",
    "    if len(model_weights_paths) != 1:\n",
    "        raise ValueError(f\"No model weights or multiple found: {model_weights_paths}\")\n",
    "    model_checkpoint = torch.load(model_weights_paths[0])\n",
    "    model.load_state_dict(model_checkpoint[\"model_state_dict\"])\n",
    "    traced_script_module = torch.jit.trace(model, example_tensor, check_trace=True, check_tolerance=1e-05)\n",
    "    os.makedirs(model_export_root, exist_ok=True)\n",
    "    jit_export_path = f\"{model_export_root}/CNN-{experiment_name}_fold{fold_nb}_jit.pt\"\n",
    "    print(f\"Exporting jit model to {jit_export_path}\")\n",
    "    traced_script_module.save(jit_export_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "preceding-wildlife",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-05-02T09:00:43.922575Z",
     "start_time": "2022-05-02T09:00:43.919960Z"
    }
   },
   "outputs": [],
   "source": [
    "# import shutil\n",
    "# shutil.rmtree(\"benchmark_src/assets/\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "directed-determination",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-27T09:51:53.290686Z",
     "start_time": "2022-01-27T09:51:42.983503Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Exporting jit model to benchmark_src/assets/CNN-Exp4-1_fold4_jit.pt\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Exporting jit model to benchmark_src/assets/CNN-Exp4-1_fold3_jit.pt\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Exporting jit model to benchmark_src/assets/CNN-Exp4-1_fold2_jit.pt\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Exporting jit model to benchmark_src/assets/CNN-Exp4-1_fold0_jit.pt\n",
      "unetplusplus timm-efficientnet-b1 noisy-student BatchNorm: True Decoder Channel: [256, 128, 64, 32, 16], Decoder Attention Type: None\n",
      "Exporting jit model to benchmark_src/assets/CNN-Exp4-1_fold1_jit.pt\n"
     ]
    }
   ],
   "source": [
    "for path in glob.glob(\"trained_models/Exp4-1/*/*/*.yaml\")[:]:\n",
    "    export_to_jit(path)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "equivalent-breast",
   "metadata": {},
   "source": [
    "## Predict"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "temporal-thesis",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-28T16:37:57.057396Z",
     "start_time": "2022-01-28T16:37:57.026353Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Overwriting benchmark_src/main.py\n"
     ]
    }
   ],
   "source": [
    "%%file benchmark_src/main.py\n",
    "import os\n",
    "from pathlib import Path\n",
    "from typing import List\n",
    "\n",
    "from loguru import logger\n",
    "import pandas as pd\n",
    "from PIL import Image\n",
    "import torch\n",
    "import typer\n",
    "import numpy as np\n",
    "import tifffile\n",
    "import segmentation_models_pytorch as smp\n",
    "\n",
    "#### DRIVENDATA\n",
    "ROOT_DIRECTORY = Path(\"/codeexecution\")\n",
    "PREDICTIONS_DIRECTORY = ROOT_DIRECTORY / \"predictions\"\n",
    "ASSETS_DIRECTORY = ROOT_DIRECTORY / \"assets\"\n",
    "DATA_DIRECTORY = ROOT_DIRECTORY / \"data\"\n",
    "INPUT_IMAGES_DIRECTORY = DATA_DIRECTORY / \"test_features\"\n",
    "\n",
    "# ###### LOCAL TESTING\n",
    "# ROOT_DIRECTORY = Path(\"benchmark_src\")\n",
    "# DATA_DIRECTORY = Path(\"data\")\n",
    "# PREDICTIONS_DIRECTORY = ROOT_DIRECTORY / \"predictions\"\n",
    "# ASSETS_DIRECTORY = ROOT_DIRECTORY / \"assets\"\n",
    "# INPUT_IMAGES_DIRECTORY = DATA_DIRECTORY / \"test_features\"\n",
    "\n",
    "# Set the pytorch cache directory and include cached models in your submission.zip\n",
    "os.environ[\"TORCH_HOME\"] = str(ASSETS_DIRECTORY / \"assets/torch\")\n",
    "\n",
    "class TestDataset(torch.utils.data.Dataset):\n",
    "    def __init__(self, img_paths):\n",
    "        self.img_paths = img_paths\n",
    "\n",
    "    def __getitem__(self, idx):\n",
    "        arr_x = tifffile.imread([self.img_paths[idx].replace(\"B02.tif\", f\"B0{k}.tif\") for k in [2, 3, 4, 8]])\n",
    "        arr_x = np.nan_to_num(arr_x)\n",
    "        arr_x = (arr_x / 2 ** 16).astype(np.float32)\n",
    "\n",
    "        sample = {\"image\": arr_x}\n",
    "        sample[\"chip_id\"] = self.img_paths[idx].split(\"/\")[-2]\n",
    "        return sample\n",
    "\n",
    "    def __len__(self):\n",
    "        return len(self.img_paths)\n",
    "\n",
    "def get_metadata(features_dir, bands):\n",
    "    \"\"\"\n",
    "    Given a folder of feature data, return a dataframe where the index is the chip id\n",
    "    and there is a column for the path to each band's TIF image.\n",
    "\n",
    "    Args:\n",
    "        features_dir (os.PathLike): path to the directory of feature data, which should have\n",
    "            a folder for each chip\n",
    "        bands (list[str]): list of bands provided for each chip\n",
    "    \"\"\"\n",
    "    chip_metadata = pd.DataFrame(index=[f\"{band}_path\" for band in bands])\n",
    "    chip_ids = (\n",
    "        pth.name for pth in features_dir.iterdir() if not pth.name.startswith(\".\")\n",
    "    )\n",
    "\n",
    "    for chip_id in chip_ids:\n",
    "        chip_bands = [f\"{features_dir}/{chip_id}/{band}.tif\" for band in bands]\n",
    "        chip_metadata[chip_id] = chip_bands\n",
    "\n",
    "    return chip_metadata.transpose().reset_index().rename(columns={\"index\": \"chip_id\"})\n",
    "\n",
    "\n",
    "def make_predictions(models, x_paths, bands, predictions_dir):\n",
    "    \"\"\"Predicts cloud cover and saves results to the predictions directory.\n",
    "\n",
    "    Args:\n",
    "        model (CloudModel): an instantiated CloudModel based on pl.LightningModule\n",
    "        x_paths (pd.DataFrame): a dataframe with a row for each chip. There must be a column for chip_id,\n",
    "                and a column with the path to the TIF for each of bands provided\n",
    "        bands (list[str]): list of bands provided for each chip\n",
    "        predictions_dir (os.PathLike): Destination directory to save the predicted TIF masks\n",
    "    \"\"\"\n",
    "#     print(x_paths)\n",
    "    test_dataset = TestDataset(img_paths=x_paths)\n",
    "    test_dataloader = torch.utils.data.DataLoader(\n",
    "        test_dataset,\n",
    "        batch_size=14,\n",
    "        num_workers=4,\n",
    "        shuffle=False,\n",
    "        pin_memory=True,\n",
    "    )\n",
    "    torch.set_grad_enabled(False)\n",
    "    for batch_index, batch in enumerate(test_dataloader):\n",
    "        logger.debug(f\"Predicting batch {batch_index} of {len(test_dataloader)}\")\n",
    "        x = batch[\"image\"].cuda(non_blocking=True)\n",
    "        preds = torch.softmax(models[0](x), dim=1)[:, 1]\n",
    "        for model_nb in range(1, len(models)):\n",
    "            preds += torch.softmax(models[model_nb](x), dim=1)[:, 1]\n",
    "        preds /= len(models)\n",
    "        preds = (preds > 0.45).cpu().numpy().astype(\"uint8\")\n",
    "        for chip_id, pred in zip(batch[\"chip_id\"], preds):\n",
    "            chip_pred_path = predictions_dir / f\"{chip_id}.tif\"\n",
    "            chip_pred_im = Image.fromarray(pred)\n",
    "            chip_pred_im.save(chip_pred_path)\n",
    "\n",
    "\n",
    "def main(\n",
    "    experiment_names=[\"Exp4\"] * 5,\n",
    "    fold_list=[0, 1, 2, 3, 4],\n",
    "    trained_models_dir: Path = ASSETS_DIRECTORY,\n",
    "    test_features_dir: Path = DATA_DIRECTORY / \"test_features\",\n",
    "    predictions_dir: Path = PREDICTIONS_DIRECTORY,\n",
    "    bands: List[str] = [\"B02\", \"B03\", \"B04\", \"B08\"],\n",
    "):\n",
    "    \"\"\"\n",
    "    Generate predictions for the chips in test_features_dir using the model saved at\n",
    "    model_weights_path.\n",
    "\n",
    "    Predictions are saved in predictions_dir. The default paths to all three files are based on\n",
    "    the structure of the code execution runtime.\n",
    "\n",
    "    Args:\n",
    "        hps_paths: Path to the model hyperparameters.\n",
    "        test_features_dir (os.PathLike, optional): Path to the features for the test data. Defaults\n",
    "            to 'data/test_features' in the same directory as main.py\n",
    "        predictions_dir (os.PathLike, optional): Destination directory to save the predicted TIF masks\n",
    "            Defaults to 'predictions' in the same directory as main.py\n",
    "        bands (List[str], optional): List of bands provided for each chip\n",
    "    \"\"\"\n",
    "    if not os.path.exists(test_features_dir):\n",
    "        raise ValueError(\n",
    "            f\"The directory for test feature images must exist and {test_features_dir} does not exist\"\n",
    "        )\n",
    "    os.makedirs(predictions_dir, exist_ok=True)\n",
    "\n",
    "    models = []\n",
    "    for experiment_name, fold_nb in zip(experiment_names, fold_list):\n",
    "        jit_model_path = f\"{trained_models_dir}/CNN-{experiment_name}_fold{fold_nb}_jit.pt\"\n",
    "        if not os.path.exists(jit_model_path):\n",
    "            raise ValueError(f\"Model path {jit_model_path} not found.\")\n",
    "        model = torch.jit.load(jit_model_path)\n",
    "        model.eval()\n",
    "        model = model.cuda()\n",
    "        models.append(model)\n",
    "        logger.info(f\"Using {jit_model_path}\")\n",
    "    \n",
    "    logger.info(\"Loading test metadata\")\n",
    "    test_metadata = get_metadata(test_features_dir, bands=bands)\n",
    "    logger.info(f\"Found {len(test_metadata)} chips\")\n",
    "\n",
    "    logger.info(\"Generating predictions in batches\")\n",
    "    make_predictions(models, test_metadata[\"B02_path\"].tolist(), bands, predictions_dir)\n",
    "    logger.info(f\"\"\"Saved {len(list(predictions_dir.glob(\"*.tif\")))} predictions\"\"\")\n",
    "\n",
    "if __name__ == \"__main__\":\n",
    "    typer.run(main)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "inside-entrepreneur",
   "metadata": {
    "ExecuteTime": {
     "start_time": "2022-01-04T14:28:39.775Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-01-04 14:28:25.544 | INFO     | benchmark_src.main:main:142 - Using benchmark_src/assets/CNN-Exp7_fold0_jit.pt\n",
      "2022-01-04 14:28:25.953 | INFO     | benchmark_src.main:main:142 - Using benchmark_src/assets/CNN-Exp7_fold1_jit.pt\n",
      "2022-01-04 14:28:26.332 | INFO     | benchmark_src.main:main:142 - Using benchmark_src/assets/CNN-Exp7_fold2_jit.pt\n",
      "2022-01-04 14:28:26.621 | INFO     | benchmark_src.main:main:142 - Using benchmark_src/assets/CNN-Exp7_fold3_jit.pt\n",
      "2022-01-04 14:28:26.919 | INFO     | benchmark_src.main:main:142 - Using benchmark_src/assets/CNN-Exp7_fold4_jit.pt\n",
      "2022-01-04 14:28:27.239 | INFO     | benchmark_src.main:main:142 - Using benchmark_src/assets/CNN-Exp7_2_fold0_jit.pt\n",
      "2022-01-04 14:28:27.526 | INFO     | benchmark_src.main:main:142 - Using benchmark_src/assets/CNN-Exp7_2_fold1_jit.pt\n",
      "2022-01-04 14:28:27.816 | INFO     | benchmark_src.main:main:142 - Using benchmark_src/assets/CNN-Exp7_2_fold2_jit.pt\n",
      "2022-01-04 14:28:28.108 | INFO     | benchmark_src.main:main:142 - Using benchmark_src/assets/CNN-Exp7_2_fold3_jit.pt\n",
      "2022-01-04 14:28:28.403 | INFO     | benchmark_src.main:main:142 - Using benchmark_src/assets/CNN-Exp7_2_fold4_jit.pt\n",
      "2022-01-04 14:28:28.403 | INFO     | benchmark_src.main:main:144 - Loading test metadata\n",
      "/home/philip/Python/OnCloudN/benchmark_src/main.py:64: PerformanceWarning: DataFrame is highly fragmented.  This is usually the result of calling `frame.insert` many times, which has poor performance.  Consider joining all columns at once using pd.concat(axis=1) instead.  To get a de-fragmented frame, use `newframe = frame.copy()`\n",
      "  chip_metadata[chip_id] = chip_bands\n",
      "2022-01-04 14:28:28.506 | INFO     | benchmark_src.main:main:146 - Found 483 chips\n",
      "2022-01-04 14:28:28.507 | INFO     | benchmark_src.main:main:148 - Generating predictions in batches\n",
      "2022-01-04 14:28:28.825 | DEBUG    | benchmark_src.main:make_predictions:90 - Predicting batch 0 of 41\n",
      "2022-01-04 14:28:33.794 | DEBUG    | benchmark_src.main:make_predictions:90 - Predicting batch 1 of 41\n",
      "2022-01-04 14:29:16.379 | DEBUG    | benchmark_src.main:make_predictions:90 - Predicting batch 2 of 41\n",
      "2022-01-04 14:29:17.618 | DEBUG    | benchmark_src.main:make_predictions:90 - Predicting batch 3 of 41\n",
      "2022-01-04 14:29:18.857 | DEBUG    | benchmark_src.main:make_predictions:90 - Predicting batch 4 of 41\n",
      "2022-01-04 14:29:20.097 | DEBUG    | benchmark_src.main:make_predictions:90 - Predicting batch 5 of 41\n",
      "2022-01-04 14:29:21.338 | DEBUG    | benchmark_src.main:make_predictions:90 - Predicting batch 6 of 41\n",
      "2022-01-04 14:29:22.581 | DEBUG    | benchmark_src.main:make_predictions:90 - Predicting batch 7 of 41\n",
      "2022-01-04 14:29:23.825 | DEBUG    | benchmark_src.main:make_predictions:90 - Predicting batch 8 of 41\n",
      "2022-01-04 14:29:25.070 | DEBUG    | benchmark_src.main:make_predictions:90 - Predicting batch 9 of 41\n",
      "2022-01-04 14:29:26.316 | DEBUG    | benchmark_src.main:make_predictions:90 - Predicting batch 10 of 41\n",
      "2022-01-04 14:29:27.561 | DEBUG    | benchmark_src.main:make_predictions:90 - Predicting batch 11 of 41\n"
     ]
    }
   ],
   "source": [
    "from benchmark_src.main import main\n",
    "from pathlib import Path\n",
    "\n",
    "main()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "dramatic-jersey",
   "metadata": {
    "ExecuteTime": {
     "end_time": "2022-01-28T16:38:30.475214Z",
     "start_time": "2022-01-28T16:38:01.661411Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "updating: __pycache__/ (stored 0%)\n",
      "updating: __pycache__/main.cpython-38.pyc (deflated 47%)\n",
      "updating: assets/ (stored 0%)\n",
      "updating: assets/CNN-Exp11-0_fold4_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp11-0_fold0_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp4_fold3_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp8_3_fold4_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp8_3_fold2_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp11-0_fold2_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp11-0_fold1_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp8_3_fold1_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp11-0_fold3_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp4_fold0_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp8_3_fold3_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp8_3_fold0_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp7-3_fold4_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp7-3_fold0_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp7-3_fold1_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp4_fold4_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp7-3_fold2_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp7-3_fold3_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp4_fold1_jit.pt (deflated 9%)\n",
      "updating: assets/CNN-Exp4_fold2_jit.pt (deflated 9%)\n",
      "updating: main.py (deflated 63%)\n"
     ]
    }
   ],
   "source": [
    "!cd benchmark_src && zip -r ../submission17.zip *"
   ]
  }
 ],
 "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": "3.8.5"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}