playground.ipynb

{
 "cells": [
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# SFXLearner\n",
    "\n",
    "This notebook is used for demonstrating the workflow of the paper \"Automatic Recognition of Cascaded Guitar Effects\" at DAFX23."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# clone SFXLearner\n",
    "!git clone https://github.com/fisheggg/SFXlearner.git\n",
    "%cd SFXlearner\n",
    "\n",
    "# install requitements\n",
    "!apt install sox libsox-fmt-all\n",
    "%pip install -r requirements.txt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# library imports\n",
    "import os\n",
    "import tqdm\n",
    "import torch\n",
    "import sklearn\n",
    "import mirdata.datasets.guitarset\n",
    "import pytorch_lightning as pl\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# internal imports\n",
    "from src.model.pl_wrapper import LightningWrapper\n",
    "from src.model.resnet import resnet18\n",
    "from src.dataloader import MultiFXDataset\n",
    "from src.generate_data import (\n",
    "    slice_guitarset,\n",
    "    slice_idmt_smt_guitar,\n",
    "    generate_dataset_sox,\n",
    ")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Data Preprocessing\n",
    "You can skip this step and jump to `Effect Processing` if you only want to have a demo run over the provided `guitarset10` subset.\n",
    "\n",
    "Here we download the clean guitar datasets, and slice them into 5-second clips.\n",
    "#### GuitarSet\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO: Downloading ['audio_mic'] to ../raw_dataset/GuitarSet/\n",
      "INFO: [audio_mic] downloading audio_mono-mic.zip\n",
      "626MB [02:26, 4.49MB/s]                               \n"
     ]
    }
   ],
   "source": [
    "# Change to your desired path\n",
    "guitarset_path = \"../raw_dataset/GuitarSet/\"\n",
    "guitarset_path_sliced = \"../dataset/clean/GuitarSet\"\n",
    "\n",
    "guitarset = mirdata.datasets.guitarset.Dataset(data_home=guitarset_path)\n",
    "print(guitarset.readme)\n",
    "print(guitarset.cite())\n",
    "guitarset.download(partial_download=[\"audio_mic\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 360/360 [03:05<00:00,  1.94it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Generation complete! 2004 audio files is generated to ../dataset/GuitarSet/guitarset_5.0s_clean.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "# slice guitarset into 5s clips\n",
    "guitarset_path_sliced = slice_guitarset(\n",
    "    guitarset_path, save_dir=guitarset_path_sliced, duration=5\n",
    ")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### IDMT-SMT-GUITAR\n",
    "\n",
    "Please download `IDMT-SMT-Guitar` and unzip manually: https://zenodo.org/record/7544110\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 507/507 [03:52<00:00,  2.18it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Generation complete! 2382 audio files is generated to ../dataset/IDMT-SMT-GUITAR/IDMT-SMT-GUITAR_5s.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "# Change to your desired path\n",
    "idmt_path = \"../raw_dataset/IDMT-SMT-GUITAR_V2\"\n",
    "idmt_path_sliced = \"../dataset/clean/IDMT-SMT-GUITAR\"\n",
    "\n",
    "idmt_path_sliced = slice_idmt_smt_guitar(\n",
    "    data_home=idmt_path, save_dir=idmt_path_sliced, duration=5\n",
    ")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Effect Processing"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Kind reminder: generating the full dataset will take several hours for each dataset.\n",
    "\n",
    "Here we run the generation over a small subset with only 10 clean samples.\n",
    "\n",
    "A `settings.yml` file will be generated in the output folder, recording the generation parameters."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# change this to the paths of the full dataset to run the full experiment \n",
    "guitarset10_path = \"dataset/clean/guitarset10\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=> Generating FX dataset...\n",
      "=> Using default fx params and grouping\n",
      "=> Settings:\n",
      "{'fx_chain_type': 'multi',\n",
      " 'fx_grouping': [(0, 1), (2, 3, 4, 5), (6,), (7, 8), (9, 10), (11, 12)],\n",
      " 'fx_params': OrderedDict([('overdrive', {'gain_db': 5}),\n",
      "                           ('distortion', {'gain_db': 15}),\n",
      "                           ('chorus', {'n_voices': 5}),\n",
      "                           ('flanger', {'depth': 5, 'phase': 50}),\n",
      "                           ('phaser', {}),\n",
      "                           ('tremolo', {}),\n",
      "                           ('reverb', {'reverberance': 80}),\n",
      "                           ('feedback_delay',\n",
      "                            {'decays': [0.4, 0.2, 0.1],\n",
      "                             'delays': [200, 400, 600],\n",
      "                             'gain_out': 0.5,\n",
      "                             'n_echos': 3}),\n",
      "                           ('slapback_delay',\n",
      "                            {'decays': [0.4, 0.2, 0.1],\n",
      "                             'delays': [200, 400, 600],\n",
      "                             'gain_out': 0.5,\n",
      "                             'n_echos': 3}),\n",
      "                           ('low_boost', {'frequency': 200, 'gain_db': 10}),\n",
      "                           ('low_reduct', {'frequency': 200, 'gain_db': -10}),\n",
      "                           ('hi_boost', {'frequency': 8000, 'gain_db': 20}),\n",
      "                           ('hi_reduct', {'frequency': 8000, 'gain_db': -20})]),\n",
      " 'generate_date': 'May-25-2023 15:55:08',\n",
      " 'generation_type': [1, 5],\n",
      " 'n_classes': 13,\n",
      " 'n_fx_groups': 6,\n",
      " 'nomalized': False,\n",
      " 'origins': ['dataset/clean/guitarset10'],\n",
      " 'random_seed': 42,\n",
      " 'sample_rate': 44100,\n",
      " 'train_size': 0,\n",
      " 'valid_size': 0}\n",
      "=> Initializing sox transformers\n",
      "=> Rendering audio files\n",
      "=> Generating training set from dataset/clean/guitarset10\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "sample count: 100%|██████████| 6/6 [01:50<00:00, 18.47s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=> Generating valid set from dataset/clean/guitarset10\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "sample count: 100%|██████████| 4/4 [01:16<00:00, 19.01s/it]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=> Generated 1326 samples to dataset/generated/guitarset10/gen_multiFX_05252023/train\n",
      "=> Generated 884 samples to dataset/generated/guitarset10/gen_multiFX_05252023/valid\n",
      "=> Training set labels written to dataset/generated/guitarset10/gen_multiFX_05252023/train/label_tensor.pt\n",
      "=> Validation set labels written to dataset/generated/guitarset10/gen_multiFX_05252023/valid/label_tensor.pt\n",
      "=> Training link file written to dataset/generated/guitarset10/gen_multiFX_05252023/train/clean_link.csv\n",
      "=> Validation link file written to dataset/generated/guitarset10/gen_multiFX_05252023/valid/clean_link.csv\n",
      "=> settings written to dataset/generated/guitarset10/gen_multiFX_05252023/settings.yml\n",
      "=> Done!\n"
     ]
    }
   ],
   "source": [
    "guitarset_path_rendered = \"dataset/generated/guitarset10\"\n",
    "generate_dataset_sox([guitarset10_path], guitarset_path_rendered, methods=[1, 5], valid_split=0.2)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Global seed set to 42\n",
      "=> Start training\n",
      "=> Training with clean\n",
      "/home/jinyueg/miniconda3/envs/sfx_learner/lib/python3.10/site-packages/pytorch_lightning/trainer/connectors/accelerator_connector.py:467: LightningDeprecationWarning: Setting `Trainer(gpus=1)` is deprecated in v1.7 and will be removed in v2.0. Please use `Trainer(accelerator='gpu', devices=1)` instead.\n",
      "  rank_zero_deprecation(\n",
      "GPU available: True (cuda), used: True\n",
      "TPU available: False, using: 0 TPU cores\n",
      "IPU available: False, using: 0 IPUs\n",
      "HPU available: False, using: 0 HPUs\n",
      "Missing logger folder: /home/jinyueg/felles/Research/Users/jinyueg/DAFX23/SFXlearner/lightning_logs/multiFX_resnet18_[1,5]_with_clean\n",
      "LOCAL_RANK: 0 - CUDA_VISIBLE_DEVICES: [0]\n",
      "\n",
      "  | Name      | Type                      | Params\n",
      "--------------------------------------------------------\n",
      "0 | model     | resnet18                  | 714 K \n",
      "1 | transform | MelSpectrogramDBTransform | 0     \n",
      "--------------------------------------------------------\n",
      "714 K     Trainable params\n",
      "0         Non-trainable params\n",
      "714 K     Total params\n",
      "2.859     Total estimated model params size (MB)\n",
      "Epoch 0:  21%|██▌         | 41/195 [00:07<00:26,  5.74it/s, loss=0.463, v_num=0]\n",
      "Validation: 0it [00:00, ?it/s]\u001b[A\n",
      "Validation:   0%|                                        | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Validation DataLoader 0:   0%|                           | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Epoch 0:  22%|██▌         | 42/195 [00:09<00:32,  4.65it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  22%|██▋         | 43/195 [00:09<00:32,  4.73it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  23%|██▋         | 44/195 [00:09<00:31,  4.81it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  23%|██▊         | 45/195 [00:09<00:30,  4.89it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  24%|██▊         | 46/195 [00:09<00:30,  4.96it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  24%|██▉         | 47/195 [00:09<00:29,  5.06it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  25%|██▉         | 48/195 [00:09<00:28,  5.15it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  25%|███         | 49/195 [00:09<00:27,  5.25it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  26%|███         | 50/195 [00:10<00:29,  4.91it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  26%|███▏        | 51/195 [00:10<00:28,  5.00it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  27%|███▏        | 52/195 [00:10<00:28,  5.08it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  27%|███▎        | 53/195 [00:10<00:27,  5.17it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  28%|███▎        | 54/195 [00:10<00:26,  5.25it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  28%|███▍        | 55/195 [00:10<00:26,  5.34it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  29%|███▍        | 56/195 [00:10<00:25,  5.42it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  29%|███▌        | 57/195 [00:10<00:25,  5.51it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  30%|███▌        | 58/195 [00:11<00:26,  5.17it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  30%|███▋        | 59/195 [00:11<00:25,  5.23it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  31%|███▋        | 60/195 [00:11<00:25,  5.30it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  31%|███▊        | 61/195 [00:11<00:24,  5.36it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  32%|███▊        | 62/195 [00:11<00:24,  5.43it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  32%|███▉        | 63/195 [00:11<00:24,  5.49it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  33%|███▉        | 64/195 [00:11<00:23,  5.56it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  33%|████        | 65/195 [00:11<00:23,  5.64it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  34%|████        | 66/195 [00:12<00:23,  5.42it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  34%|████        | 67/195 [00:12<00:23,  5.49it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  35%|████▏       | 68/195 [00:12<00:22,  5.56it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  35%|████▏       | 69/195 [00:12<00:22,  5.63it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  36%|████▎       | 70/195 [00:12<00:21,  5.70it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  36%|████▎       | 71/195 [00:12<00:21,  5.77it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  37%|████▍       | 72/195 [00:12<00:21,  5.84it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  37%|████▍       | 73/195 [00:12<00:20,  5.91it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  38%|████▌       | 74/195 [00:13<00:21,  5.58it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  38%|████▌       | 75/195 [00:13<00:21,  5.63it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  39%|████▋       | 76/195 [00:13<00:20,  5.68it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  39%|████▋       | 77/195 [00:13<00:20,  5.73it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  40%|████▊       | 78/195 [00:13<00:20,  5.79it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  41%|████▊       | 79/195 [00:13<00:19,  5.85it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  41%|████▉       | 80/195 [00:13<00:19,  5.91it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  42%|████▉       | 81/195 [00:13<00:19,  5.98it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  42%|█████       | 82/195 [00:14<00:19,  5.71it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  43%|█████       | 83/195 [00:14<00:19,  5.72it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  43%|█████▏      | 84/195 [00:14<00:19,  5.76it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  44%|█████▏      | 85/195 [00:14<00:18,  5.80it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  44%|█████▎      | 86/195 [00:14<00:18,  5.86it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  45%|█████▎      | 87/195 [00:14<00:18,  5.92it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  45%|█████▍      | 88/195 [00:14<00:17,  5.98it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  46%|█████▍      | 89/195 [00:14<00:17,  6.04it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  46%|█████▌      | 90/195 [00:15<00:18,  5.83it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  47%|█████▌      | 91/195 [00:15<00:17,  5.85it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  47%|█████▋      | 92/195 [00:15<00:17,  5.89it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  48%|█████▋      | 93/195 [00:15<00:17,  5.94it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  48%|█████▊      | 94/195 [00:15<00:16,  5.99it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  49%|█████▊      | 95/195 [00:15<00:16,  6.04it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  49%|█████▉      | 96/195 [00:15<00:16,  6.10it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  50%|█████▉      | 97/195 [00:15<00:16,  6.08it/s, loss=0.463, v_num=0]\u001b[A\n",
      "Epoch 0:  71%|███████▊   | 138/195 [00:20<00:08,  6.81it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Validation: 0it [00:00, ?it/s]\u001b[A\n",
      "Validation:   0%|                                        | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Validation DataLoader 0:   0%|                           | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Epoch 0:  71%|███████▊   | 139/195 [00:21<00:08,  6.52it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  72%|███████▉   | 140/195 [00:21<00:08,  6.56it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  72%|███████▉   | 141/195 [00:21<00:08,  6.60it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  73%|████████   | 142/195 [00:21<00:07,  6.64it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  73%|████████   | 143/195 [00:21<00:07,  6.68it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  74%|████████   | 144/195 [00:21<00:07,  6.72it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  74%|████████▏  | 145/195 [00:21<00:07,  6.76it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  75%|████████▏  | 146/195 [00:21<00:07,  6.80it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  75%|████████▎  | 147/195 [00:22<00:07,  6.61it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  76%|████████▎  | 148/195 [00:22<00:07,  6.64it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  76%|████████▍  | 149/195 [00:22<00:06,  6.68it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  77%|████████▍  | 150/195 [00:22<00:06,  6.72it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  77%|████████▌  | 151/195 [00:22<00:06,  6.76it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  78%|████████▌  | 152/195 [00:22<00:06,  6.79it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  78%|████████▋  | 153/195 [00:22<00:06,  6.83it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  79%|████████▋  | 154/195 [00:22<00:05,  6.87it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  79%|████████▋  | 155/195 [00:23<00:05,  6.68it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  80%|████████▊  | 156/195 [00:23<00:05,  6.71it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  81%|████████▊  | 157/195 [00:23<00:05,  6.74it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  81%|████████▉  | 158/195 [00:23<00:05,  6.77it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  82%|████████▉  | 159/195 [00:23<00:05,  6.81it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  82%|█████████  | 160/195 [00:23<00:05,  6.84it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  83%|█████████  | 161/195 [00:23<00:04,  6.88it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  83%|█████████▏ | 162/195 [00:23<00:04,  6.91it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  84%|█████████▏ | 163/195 [00:24<00:04,  6.75it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  84%|█████████▎ | 164/195 [00:24<00:04,  6.79it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  85%|█████████▎ | 165/195 [00:24<00:04,  6.82it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  85%|█████████▎ | 166/195 [00:24<00:04,  6.85it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  86%|█████████▍ | 167/195 [00:24<00:04,  6.89it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  86%|█████████▍ | 168/195 [00:24<00:03,  6.92it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  87%|█████████▌ | 169/195 [00:24<00:03,  6.96it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  87%|█████████▌ | 170/195 [00:24<00:03,  6.99it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  88%|█████████▋ | 171/195 [00:25<00:03,  6.81it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  88%|█████████▋ | 172/195 [00:25<00:03,  6.84it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  89%|█████████▊ | 173/195 [00:25<00:03,  6.86it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  89%|█████████▊ | 174/195 [00:25<00:03,  6.89it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  90%|█████████▊ | 175/195 [00:25<00:02,  6.91it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  90%|█████████▉ | 176/195 [00:25<00:02,  6.94it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  91%|█████████▉ | 177/195 [00:25<00:02,  6.98it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  91%|██████████ | 178/195 [00:25<00:02,  7.01it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  92%|██████████ | 179/195 [00:26<00:02,  6.88it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  92%|██████████▏| 180/195 [00:26<00:02,  6.89it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  93%|██████████▏| 181/195 [00:26<00:02,  6.92it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  93%|██████████▎| 182/195 [00:26<00:01,  6.96it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  94%|██████████▎| 183/195 [00:26<00:01,  6.99it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  94%|██████████▍| 184/195 [00:26<00:01,  7.02it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  95%|██████████▍| 185/195 [00:26<00:01,  7.05it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  95%|██████████▍| 186/195 [00:26<00:01,  7.09it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  96%|██████████▌| 187/195 [00:26<00:01,  6.96it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  96%|██████████▌| 188/195 [00:26<00:01,  6.98it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  97%|██████████▋| 189/195 [00:26<00:00,  7.01it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  97%|██████████▋| 190/195 [00:26<00:00,  7.04it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  98%|██████████▊| 191/195 [00:26<00:00,  7.08it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  98%|██████████▊| 192/195 [00:27<00:00,  7.11it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  99%|██████████▉| 193/195 [00:27<00:00,  7.14it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 0:  99%|██████████▉| 194/195 [00:27<00:00,  7.16it/s, loss=0.419, v_num=0]\u001b[A\n",
      "Epoch 1:  21%|██▌         | 41/195 [00:07<00:29,  5.26it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Validation: 0it [00:00, ?it/s]\u001b[A\n",
      "Validation:   0%|                                        | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Validation DataLoader 0:   0%|                           | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Epoch 1:  22%|██▌         | 42/195 [00:09<00:35,  4.31it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  22%|██▋         | 43/195 [00:09<00:34,  4.38it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  23%|██▋         | 44/195 [00:09<00:33,  4.46it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  23%|██▊         | 45/195 [00:09<00:33,  4.53it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  24%|██▊         | 46/195 [00:09<00:32,  4.61it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  24%|██▉         | 47/195 [00:10<00:31,  4.69it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  25%|██▉         | 48/195 [00:10<00:30,  4.78it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  25%|███         | 49/195 [00:10<00:29,  4.87it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  26%|███         | 50/195 [00:10<00:31,  4.63it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  26%|███▏        | 51/195 [00:10<00:30,  4.72it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  27%|███▏        | 52/195 [00:10<00:29,  4.80it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  27%|███▎        | 53/195 [00:10<00:29,  4.88it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  28%|███▎        | 54/195 [00:10<00:28,  4.97it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  28%|███▍        | 55/195 [00:10<00:27,  5.05it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  29%|███▍        | 56/195 [00:10<00:27,  5.13it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  29%|███▌        | 57/195 [00:10<00:26,  5.21it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  30%|███▌        | 58/195 [00:11<00:27,  4.94it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  30%|███▋        | 59/195 [00:11<00:27,  5.02it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  31%|███▋        | 60/195 [00:11<00:26,  5.09it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  31%|███▊        | 61/195 [00:11<00:25,  5.16it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  32%|███▊        | 62/195 [00:11<00:25,  5.24it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  32%|███▉        | 63/195 [00:11<00:24,  5.31it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  33%|███▉        | 64/195 [00:11<00:24,  5.38it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  33%|████        | 65/195 [00:11<00:23,  5.45it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  34%|████        | 66/195 [00:12<00:24,  5.20it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  34%|████        | 67/195 [00:12<00:24,  5.25it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  35%|████▏       | 68/195 [00:12<00:23,  5.31it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  35%|████▏       | 69/195 [00:12<00:23,  5.37it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  36%|████▎       | 70/195 [00:12<00:22,  5.44it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  36%|████▎       | 71/195 [00:12<00:22,  5.51it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  37%|████▍       | 72/195 [00:12<00:22,  5.57it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  37%|████▍       | 73/195 [00:12<00:21,  5.64it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  38%|████▌       | 74/195 [00:13<00:22,  5.45it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  38%|████▌       | 75/195 [00:13<00:21,  5.51it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  39%|████▋       | 76/195 [00:13<00:21,  5.57it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  39%|████▋       | 77/195 [00:13<00:20,  5.63it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  40%|████▊       | 78/195 [00:13<00:20,  5.69it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  41%|████▊       | 79/195 [00:13<00:20,  5.75it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  41%|████▉       | 80/195 [00:13<00:19,  5.81it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  42%|████▉       | 81/195 [00:13<00:19,  5.86it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  42%|█████       | 82/195 [00:14<00:19,  5.69it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  43%|█████       | 83/195 [00:14<00:19,  5.71it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  43%|█████▏      | 84/195 [00:14<00:19,  5.76it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  44%|█████▏      | 85/195 [00:14<00:18,  5.81it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  44%|█████▎      | 86/195 [00:14<00:18,  5.85it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  45%|█████▎      | 87/195 [00:14<00:18,  5.89it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  45%|█████▍      | 88/195 [00:14<00:18,  5.94it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  46%|█████▍      | 89/195 [00:14<00:17,  6.00it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  46%|█████▌      | 90/195 [00:15<00:17,  5.90it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  47%|█████▌      | 91/195 [00:15<00:17,  5.93it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  47%|█████▋      | 92/195 [00:15<00:17,  5.99it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  48%|█████▋      | 93/195 [00:15<00:16,  6.04it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  48%|█████▊      | 94/195 [00:15<00:16,  6.10it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  49%|█████▊      | 95/195 [00:15<00:16,  6.16it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  49%|█████▉      | 96/195 [00:15<00:15,  6.21it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  50%|█████▉      | 97/195 [00:15<00:15,  6.25it/s, loss=0.379, v_num=0]\u001b[A\n",
      "Epoch 1:  71%|███████▊   | 138/195 [00:19<00:08,  7.12it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Validation: 0it [00:00, ?it/s]\u001b[A\n",
      "Validation:   0%|                                        | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Validation DataLoader 0:   0%|                           | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Epoch 1:  71%|███████▊   | 139/195 [00:20<00:08,  6.82it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  72%|███████▉   | 140/195 [00:20<00:08,  6.86it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  72%|███████▉   | 141/195 [00:20<00:07,  6.89it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  73%|████████   | 142/195 [00:20<00:07,  6.93it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  73%|████████   | 143/195 [00:20<00:07,  6.97it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  74%|████████   | 144/195 [00:20<00:07,  7.00it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  74%|████████▏  | 145/195 [00:20<00:07,  7.04it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  75%|████████▏  | 146/195 [00:20<00:06,  7.08it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  75%|████████▎  | 147/195 [00:21<00:06,  6.90it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  76%|████████▎  | 148/195 [00:21<00:06,  6.94it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  76%|████████▍  | 149/195 [00:21<00:06,  6.98it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  77%|████████▍  | 150/195 [00:21<00:06,  7.01it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  77%|████████▌  | 151/195 [00:21<00:06,  7.05it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  78%|████████▌  | 152/195 [00:21<00:06,  7.09it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  78%|████████▋  | 153/195 [00:21<00:05,  7.13it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  79%|████████▋  | 154/195 [00:21<00:05,  7.16it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  79%|████████▋  | 155/195 [00:22<00:05,  6.99it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  80%|████████▊  | 156/195 [00:22<00:05,  7.03it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  81%|████████▊  | 157/195 [00:22<00:05,  7.06it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  81%|████████▉  | 158/195 [00:22<00:05,  7.10it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  82%|████████▉  | 159/195 [00:22<00:05,  7.13it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  82%|█████████  | 160/195 [00:22<00:04,  7.17it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  83%|█████████  | 161/195 [00:22<00:04,  7.20it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  83%|█████████▏ | 162/195 [00:22<00:04,  7.24it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  84%|█████████▏ | 163/195 [00:23<00:04,  7.08it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  84%|█████████▎ | 164/195 [00:23<00:04,  7.11it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  85%|█████████▎ | 165/195 [00:23<00:04,  7.13it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  85%|█████████▎ | 166/195 [00:23<00:04,  7.15it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  86%|█████████▍ | 167/195 [00:23<00:03,  7.18it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  86%|█████████▍ | 168/195 [00:23<00:03,  7.22it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  87%|█████████▌ | 169/195 [00:23<00:03,  7.25it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  87%|█████████▌ | 170/195 [00:23<00:03,  7.29it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  88%|█████████▋ | 171/195 [00:23<00:03,  7.17it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  88%|█████████▋ | 172/195 [00:23<00:03,  7.21it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  89%|█████████▊ | 173/195 [00:23<00:03,  7.24it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  89%|█████████▊ | 174/195 [00:23<00:02,  7.28it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  90%|█████████▊ | 175/195 [00:23<00:02,  7.31it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  90%|█████████▉ | 176/195 [00:23<00:02,  7.35it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  91%|█████████▉ | 177/195 [00:23<00:02,  7.38it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  91%|██████████ | 178/195 [00:24<00:02,  7.41it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  92%|██████████ | 179/195 [00:24<00:02,  7.29it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  92%|██████████▏| 180/195 [00:24<00:02,  7.29it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  93%|██████████▏| 181/195 [00:24<00:01,  7.33it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  93%|██████████▎| 182/195 [00:24<00:01,  7.36it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  94%|██████████▎| 183/195 [00:24<00:01,  7.39it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  94%|██████████▍| 184/195 [00:24<00:01,  7.42it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  95%|██████████▍| 185/195 [00:24<00:01,  7.45it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  95%|██████████▍| 186/195 [00:24<00:01,  7.49it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  96%|██████████▌| 187/195 [00:25<00:01,  7.37it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  96%|██████████▌| 188/195 [00:25<00:00,  7.39it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  97%|██████████▋| 189/195 [00:25<00:00,  7.41it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  97%|██████████▋| 190/195 [00:25<00:00,  7.44it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  98%|██████████▊| 191/195 [00:25<00:00,  7.46it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  98%|██████████▊| 192/195 [00:25<00:00,  7.48it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  99%|██████████▉| 193/195 [00:25<00:00,  7.52it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 1:  99%|██████████▉| 194/195 [00:25<00:00,  7.53it/s, loss=0.353, v_num=0]\u001b[A\n",
      "Epoch 2:  21%|██▌         | 41/195 [00:06<00:24,  6.29it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Validation: 0it [00:00, ?it/s]\u001b[A\n",
      "Validation:   0%|                                        | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Validation DataLoader 0:   0%|                           | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Epoch 2:  22%|██▌         | 42/195 [00:08<00:30,  5.01it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  22%|██▋         | 43/195 [00:08<00:29,  5.08it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  23%|██▋         | 44/195 [00:08<00:29,  5.16it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  23%|██▊         | 45/195 [00:08<00:28,  5.24it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  24%|██▊         | 46/195 [00:08<00:27,  5.33it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  24%|██▉         | 47/195 [00:08<00:27,  5.41it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  25%|██▉         | 48/195 [00:08<00:26,  5.51it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  25%|███         | 49/195 [00:08<00:26,  5.61it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  26%|███         | 50/195 [00:09<00:27,  5.34it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  26%|███▏        | 51/195 [00:09<00:26,  5.43it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  27%|███▏        | 52/195 [00:09<00:25,  5.51it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  27%|███▎        | 53/195 [00:09<00:25,  5.60it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  28%|███▎        | 54/195 [00:09<00:24,  5.69it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  28%|███▍        | 55/195 [00:09<00:24,  5.77it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  29%|███▍        | 56/195 [00:09<00:23,  5.86it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  29%|███▌        | 57/195 [00:09<00:23,  5.94it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  30%|███▌        | 58/195 [00:10<00:24,  5.67it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  30%|███▋        | 59/195 [00:10<00:23,  5.75it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  31%|███▋        | 60/195 [00:10<00:23,  5.82it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  31%|███▊        | 61/195 [00:10<00:22,  5.90it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  32%|███▊        | 62/195 [00:10<00:22,  5.97it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  32%|███▉        | 63/195 [00:10<00:21,  6.06it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  33%|███▉        | 64/195 [00:10<00:21,  6.14it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  33%|████        | 65/195 [00:10<00:20,  6.22it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  34%|████        | 66/195 [00:11<00:21,  5.97it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  34%|████        | 67/195 [00:11<00:21,  6.05it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  35%|████▏       | 68/195 [00:11<00:20,  6.11it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  35%|████▏       | 69/195 [00:11<00:20,  6.17it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  36%|████▎       | 70/195 [00:11<00:20,  6.23it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  36%|████▎       | 71/195 [00:11<00:19,  6.28it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  37%|████▍       | 72/195 [00:11<00:19,  6.36it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  37%|████▍       | 73/195 [00:11<00:18,  6.43it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  38%|████▌       | 74/195 [00:11<00:19,  6.18it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  38%|████▌       | 75/195 [00:11<00:19,  6.25it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  39%|████▋       | 76/195 [00:12<00:18,  6.33it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  39%|████▋       | 77/195 [00:12<00:18,  6.40it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  40%|████▊       | 78/195 [00:12<00:18,  6.46it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  41%|████▊       | 79/195 [00:12<00:17,  6.53it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  41%|████▉       | 80/195 [00:12<00:17,  6.60it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  42%|████▉       | 81/195 [00:12<00:17,  6.67it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  42%|█████       | 82/195 [00:12<00:17,  6.41it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  43%|█████       | 83/195 [00:12<00:17,  6.45it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  43%|█████▏      | 84/195 [00:12<00:17,  6.50it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  44%|█████▏      | 85/195 [00:12<00:16,  6.57it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  44%|█████▎      | 86/195 [00:12<00:16,  6.63it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  45%|█████▎      | 87/195 [00:13<00:16,  6.69it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  45%|█████▍      | 88/195 [00:13<00:15,  6.76it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  46%|█████▍      | 89/195 [00:13<00:15,  6.82it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  46%|█████▌      | 90/195 [00:13<00:15,  6.59it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  47%|█████▌      | 91/195 [00:13<00:15,  6.64it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  47%|█████▋      | 92/195 [00:13<00:15,  6.69it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  48%|█████▋      | 93/195 [00:13<00:15,  6.75it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  48%|█████▊      | 94/195 [00:13<00:14,  6.81it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  49%|█████▊      | 95/195 [00:13<00:14,  6.87it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  49%|█████▉      | 96/195 [00:13<00:14,  6.93it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  50%|█████▉      | 97/195 [00:13<00:14,  6.97it/s, loss=0.342, v_num=0]\u001b[A\n",
      "Epoch 2:  71%|███████▊   | 138/195 [00:18<00:07,  7.62it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Validation: 0it [00:00, ?it/s]\u001b[A\n",
      "Validation:   0%|                                        | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Validation DataLoader 0:   0%|                           | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Epoch 2:  71%|███████▊   | 139/195 [00:19<00:07,  7.28it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  72%|███████▉   | 140/195 [00:19<00:07,  7.32it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  72%|███████▉   | 141/195 [00:19<00:07,  7.36it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  73%|████████   | 142/195 [00:19<00:07,  7.41it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  73%|████████   | 143/195 [00:19<00:06,  7.45it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  74%|████████   | 144/195 [00:19<00:06,  7.49it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  74%|████████▏  | 145/195 [00:19<00:06,  7.53it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  75%|████████▏  | 146/195 [00:19<00:06,  7.57it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  75%|████████▎  | 147/195 [00:20<00:06,  7.32it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  76%|████████▎  | 148/195 [00:20<00:06,  7.35it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  76%|████████▍  | 149/195 [00:20<00:06,  7.38it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  77%|████████▍  | 150/195 [00:20<00:06,  7.41it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  77%|████████▌  | 151/195 [00:20<00:05,  7.44it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  78%|████████▌  | 152/195 [00:20<00:05,  7.47it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  78%|████████▋  | 153/195 [00:20<00:05,  7.51it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  79%|████████▋  | 154/195 [00:20<00:05,  7.55it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  79%|████████▋  | 155/195 [00:20<00:05,  7.40it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  80%|████████▊  | 156/195 [00:20<00:05,  7.44it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  81%|████████▊  | 157/195 [00:21<00:05,  7.48it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  81%|████████▉  | 158/195 [00:21<00:04,  7.51it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  82%|████████▉  | 159/195 [00:21<00:04,  7.55it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  82%|█████████  | 160/195 [00:21<00:04,  7.59it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  83%|█████████  | 161/195 [00:21<00:04,  7.63it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  83%|█████████▏ | 162/195 [00:21<00:04,  7.66it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  84%|█████████▏ | 163/195 [00:21<00:04,  7.46it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  84%|█████████▎ | 164/195 [00:21<00:04,  7.50it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  85%|█████████▎ | 165/195 [00:21<00:03,  7.54it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  85%|█████████▎ | 166/195 [00:21<00:03,  7.57it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  86%|█████████▍ | 167/195 [00:21<00:03,  7.60it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  86%|█████████▍ | 168/195 [00:22<00:03,  7.64it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  87%|█████████▌ | 169/195 [00:22<00:03,  7.67it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  87%|█████████▌ | 170/195 [00:22<00:03,  7.70it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  88%|█████████▋ | 171/195 [00:22<00:03,  7.51it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  88%|█████████▋ | 172/195 [00:22<00:03,  7.53it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  89%|█████████▊ | 173/195 [00:22<00:02,  7.55it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  89%|█████████▊ | 174/195 [00:22<00:02,  7.58it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  90%|█████████▊ | 175/195 [00:22<00:02,  7.62it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  90%|█████████▉ | 176/195 [00:23<00:02,  7.65it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  91%|█████████▉ | 177/195 [00:23<00:02,  7.69it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  91%|██████████ | 178/195 [00:23<00:02,  7.72it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  92%|██████████ | 179/195 [00:23<00:02,  7.60it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  92%|██████████▏| 180/195 [00:23<00:01,  7.61it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  93%|██████████▏| 181/195 [00:23<00:01,  7.64it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  93%|██████████▎| 182/195 [00:23<00:01,  7.68it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  94%|██████████▎| 183/195 [00:23<00:01,  7.71it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  94%|██████████▍| 184/195 [00:23<00:01,  7.75it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  95%|██████████▍| 185/195 [00:23<00:01,  7.78it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  95%|██████████▍| 186/195 [00:23<00:01,  7.82it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  96%|██████████▌| 187/195 [00:24<00:01,  7.67it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  96%|██████████▌| 188/195 [00:24<00:00,  7.69it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  97%|██████████▋| 189/195 [00:24<00:00,  7.72it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  97%|██████████▋| 190/195 [00:24<00:00,  7.76it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  98%|██████████▊| 191/195 [00:24<00:00,  7.79it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  98%|██████████▊| 192/195 [00:24<00:00,  7.83it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  99%|██████████▉| 193/195 [00:24<00:00,  7.86it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 2:  99%|██████████▉| 194/195 [00:24<00:00,  7.88it/s, loss=0.321, v_num=0]\u001b[A\n",
      "Epoch 3:  21%|██▋          | 41/195 [00:06<00:23,  6.45it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Validation: 0it [00:00, ?it/s]\u001b[A\n",
      "Validation:   0%|                                        | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Validation DataLoader 0:   0%|                           | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Epoch 3:  22%|██▊          | 42/195 [00:08<00:30,  4.99it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  22%|██▊          | 43/195 [00:08<00:29,  5.08it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  23%|██▉          | 44/195 [00:08<00:29,  5.16it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  23%|███          | 45/195 [00:08<00:28,  5.24it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  24%|███          | 46/195 [00:08<00:28,  5.32it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  24%|███▏         | 47/195 [00:08<00:27,  5.41it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  25%|███▏         | 48/195 [00:08<00:26,  5.51it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  25%|███▎         | 49/195 [00:08<00:26,  5.61it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  26%|███▎         | 50/195 [00:09<00:27,  5.32it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  26%|███▍         | 51/195 [00:09<00:26,  5.41it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  27%|███▍         | 52/195 [00:09<00:25,  5.50it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  27%|███▌         | 53/195 [00:09<00:25,  5.60it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  28%|███▌         | 54/195 [00:09<00:24,  5.69it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  28%|███▋         | 55/195 [00:09<00:24,  5.78it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  29%|███▋         | 56/195 [00:09<00:23,  5.86it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  29%|███▊         | 57/195 [00:09<00:23,  5.95it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  30%|███▊         | 58/195 [00:10<00:24,  5.59it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  30%|███▉         | 59/195 [00:10<00:24,  5.66it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  31%|████         | 60/195 [00:10<00:23,  5.74it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  31%|████         | 61/195 [00:10<00:23,  5.82it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  32%|████▏        | 62/195 [00:10<00:22,  5.90it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  32%|████▏        | 63/195 [00:10<00:22,  5.98it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  33%|████▎        | 64/195 [00:10<00:21,  6.06it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  33%|████▎        | 65/195 [00:10<00:21,  6.14it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  34%|████▍        | 66/195 [00:11<00:22,  5.86it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  34%|████▍        | 67/195 [00:11<00:21,  5.93it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  35%|████▌        | 68/195 [00:11<00:21,  6.00it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  35%|████▌        | 69/195 [00:11<00:20,  6.07it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  36%|████▋        | 70/195 [00:11<00:20,  6.15it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  36%|████▋        | 71/195 [00:11<00:19,  6.22it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  37%|████▊        | 72/195 [00:11<00:19,  6.29it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  37%|████▊        | 73/195 [00:11<00:19,  6.37it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  38%|████▉        | 74/195 [00:12<00:20,  6.03it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  38%|█████        | 75/195 [00:12<00:19,  6.09it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  39%|█████        | 76/195 [00:12<00:19,  6.14it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  39%|█████▏       | 77/195 [00:12<00:19,  6.19it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  40%|█████▏       | 78/195 [00:12<00:18,  6.24it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  41%|█████▎       | 79/195 [00:12<00:18,  6.31it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  41%|█████▎       | 80/195 [00:12<00:18,  6.38it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  42%|█████▍       | 81/195 [00:12<00:17,  6.44it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  42%|█████▍       | 82/195 [00:12<00:17,  6.32it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  43%|█████▌       | 83/195 [00:13<00:17,  6.37it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  43%|█████▌       | 84/195 [00:13<00:17,  6.42it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  44%|█████▋       | 85/195 [00:13<00:16,  6.48it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  44%|█████▋       | 86/195 [00:13<00:16,  6.54it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  45%|█████▊       | 87/195 [00:13<00:16,  6.61it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  45%|█████▊       | 88/195 [00:13<00:16,  6.68it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  46%|█████▉       | 89/195 [00:13<00:15,  6.74it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  46%|██████       | 90/195 [00:13<00:16,  6.47it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  47%|██████       | 91/195 [00:13<00:15,  6.52it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  47%|██████▏      | 92/195 [00:13<00:15,  6.58it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  48%|██████▏      | 93/195 [00:14<00:15,  6.64it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  48%|██████▎      | 94/195 [00:14<00:15,  6.70it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  49%|██████▎      | 95/195 [00:14<00:14,  6.76it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  49%|██████▍      | 96/195 [00:14<00:14,  6.82it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  50%|██████▍      | 97/195 [00:14<00:14,  6.85it/s, loss=0.28, v_num=0]\u001b[A\n",
      "Epoch 3:  71%|███████▊   | 138/195 [00:18<00:07,  7.50it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Validation: 0it [00:00, ?it/s]\u001b[A\n",
      "Validation:   0%|                                        | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Validation DataLoader 0:   0%|                           | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Epoch 3:  71%|███████▊   | 139/195 [00:19<00:07,  7.11it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  72%|███████▉   | 140/195 [00:19<00:07,  7.15it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  72%|███████▉   | 141/195 [00:19<00:07,  7.20it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  73%|████████   | 142/195 [00:19<00:07,  7.23it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  73%|████████   | 143/195 [00:19<00:07,  7.28it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  74%|████████   | 144/195 [00:19<00:06,  7.32it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  74%|████████▏  | 145/195 [00:19<00:06,  7.36it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  75%|████████▏  | 146/195 [00:19<00:06,  7.39it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  75%|████████▎  | 147/195 [00:20<00:06,  7.17it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  76%|████████▎  | 148/195 [00:20<00:06,  7.20it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  76%|████████▍  | 149/195 [00:20<00:06,  7.23it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  77%|████████▍  | 150/195 [00:20<00:06,  7.26it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  77%|████████▌  | 151/195 [00:20<00:06,  7.29it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  78%|████████▌  | 152/195 [00:20<00:05,  7.33it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  78%|████████▋  | 153/195 [00:20<00:05,  7.37it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  79%|████████▋  | 154/195 [00:20<00:05,  7.41it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  79%|████████▋  | 155/195 [00:21<00:05,  7.25it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  80%|████████▊  | 156/195 [00:21<00:05,  7.28it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  81%|████████▊  | 157/195 [00:21<00:05,  7.32it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  81%|████████▉  | 158/195 [00:21<00:05,  7.36it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  82%|████████▉  | 159/195 [00:21<00:04,  7.40it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  82%|█████████  | 160/195 [00:21<00:04,  7.43it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  83%|█████████  | 161/195 [00:21<00:04,  7.47it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  83%|█████████▏ | 162/195 [00:21<00:04,  7.51it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  84%|█████████▏ | 163/195 [00:22<00:04,  7.33it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  84%|█████████▎ | 164/195 [00:22<00:04,  7.37it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  85%|█████████▎ | 165/195 [00:22<00:04,  7.40it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  85%|█████████▎ | 166/195 [00:22<00:03,  7.44it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  86%|█████████▍ | 167/195 [00:22<00:03,  7.47it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  86%|█████████▍ | 168/195 [00:22<00:03,  7.51it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  87%|█████████▌ | 169/195 [00:22<00:03,  7.55it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  87%|█████████▌ | 170/195 [00:22<00:03,  7.58it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  88%|█████████▋ | 171/195 [00:23<00:03,  7.39it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  88%|█████████▋ | 172/195 [00:23<00:03,  7.42it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  89%|█████████▊ | 173/195 [00:23<00:02,  7.45it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  89%|█████████▊ | 174/195 [00:23<00:02,  7.48it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  90%|█████████▊ | 175/195 [00:23<00:02,  7.51it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  90%|█████████▉ | 176/195 [00:23<00:02,  7.55it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  91%|█████████▉ | 177/195 [00:23<00:02,  7.58it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  91%|██████████ | 178/195 [00:23<00:02,  7.61it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  92%|██████████ | 179/195 [00:23<00:02,  7.47it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  92%|██████████▏| 180/195 [00:24<00:02,  7.48it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  93%|██████████▏| 181/195 [00:24<00:01,  7.50it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  93%|██████████▎| 182/195 [00:24<00:01,  7.52it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  94%|██████████▎| 183/195 [00:24<00:01,  7.55it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  94%|██████████▍| 184/195 [00:24<00:01,  7.58it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  95%|██████████▍| 185/195 [00:24<00:01,  7.61it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  95%|██████████▍| 186/195 [00:24<00:01,  7.64it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  96%|██████████▌| 187/195 [00:24<00:01,  7.56it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  96%|██████████▌| 188/195 [00:24<00:00,  7.58it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  97%|██████████▋| 189/195 [00:24<00:00,  7.61it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  97%|██████████▋| 190/195 [00:24<00:00,  7.64it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  98%|██████████▊| 191/195 [00:24<00:00,  7.67it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  98%|██████████▊| 192/195 [00:24<00:00,  7.70it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  99%|██████████▉| 193/195 [00:24<00:00,  7.74it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 3:  99%|██████████▉| 194/195 [00:25<00:00,  7.76it/s, loss=0.283, v_num=0]\u001b[A\n",
      "Epoch 4:  21%|██▌         | 41/195 [00:06<00:24,  6.34it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Validation: 0it [00:00, ?it/s]\u001b[A\n",
      "Validation:   0%|                                        | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Validation DataLoader 0:   0%|                           | 0/56 [00:00<?, ?it/s]\u001b[A\n",
      "Epoch 4:  22%|██▌         | 42/195 [00:08<00:30,  5.03it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  22%|██▋         | 43/195 [00:08<00:29,  5.10it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  23%|██▋         | 44/195 [00:08<00:29,  5.19it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  23%|██▊         | 45/195 [00:08<00:28,  5.26it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  24%|██▊         | 46/195 [00:08<00:27,  5.33it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  24%|██▉         | 47/195 [00:08<00:27,  5.42it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  25%|██▉         | 48/195 [00:08<00:26,  5.52it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  25%|███         | 49/195 [00:08<00:25,  5.62it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  26%|███         | 50/195 [00:09<00:26,  5.40it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  26%|███▏        | 51/195 [00:09<00:26,  5.48it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  27%|███▏        | 52/195 [00:09<00:25,  5.57it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  27%|███▎        | 53/195 [00:09<00:25,  5.67it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  28%|███▎        | 54/195 [00:09<00:24,  5.76it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  28%|███▍        | 55/195 [00:09<00:23,  5.85it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  29%|███▍        | 56/195 [00:09<00:23,  5.94it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  29%|███▌        | 57/195 [00:09<00:22,  6.03it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  30%|███▌        | 58/195 [00:10<00:23,  5.74it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  30%|███▋        | 59/195 [00:10<00:23,  5.82it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  31%|███▋        | 60/195 [00:10<00:22,  5.90it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  31%|███▊        | 61/195 [00:10<00:22,  5.99it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  32%|███▊        | 62/195 [00:10<00:21,  6.08it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  32%|███▉        | 63/195 [00:10<00:21,  6.16it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  33%|███▉        | 64/195 [00:10<00:20,  6.24it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  33%|████        | 65/195 [00:10<00:20,  6.32it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  34%|████        | 66/195 [00:10<00:21,  6.07it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  34%|████        | 67/195 [00:10<00:20,  6.14it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  35%|████▏       | 68/195 [00:10<00:20,  6.22it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  35%|████▏       | 69/195 [00:10<00:20,  6.29it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  36%|████▎       | 70/195 [00:10<00:19,  6.36it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  36%|████▎       | 71/195 [00:11<00:19,  6.44it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  37%|████▍       | 72/195 [00:11<00:18,  6.52it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  37%|████▍       | 73/195 [00:11<00:18,  6.59it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  38%|████▌       | 74/195 [00:11<00:19,  6.29it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  38%|████▌       | 75/195 [00:11<00:18,  6.34it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  39%|████▋       | 76/195 [00:11<00:18,  6.40it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  39%|████▋       | 77/195 [00:11<00:18,  6.46it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  40%|████▊       | 78/195 [00:11<00:17,  6.53it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  41%|████▊       | 79/195 [00:11<00:17,  6.60it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  41%|████▉       | 80/195 [00:11<00:17,  6.67it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  42%|████▉       | 81/195 [00:12<00:16,  6.75it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  42%|█████       | 82/195 [00:12<00:17,  6.57it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  43%|█████       | 83/195 [00:12<00:16,  6.61it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  43%|█████▏      | 84/195 [00:12<00:16,  6.66it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  44%|█████▏      | 85/195 [00:12<00:16,  6.71it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  44%|█████▎      | 86/195 [00:12<00:16,  6.78it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  45%|█████▎      | 87/195 [00:12<00:15,  6.84it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  45%|█████▍      | 88/195 [00:12<00:15,  6.91it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  46%|█████▍      | 89/195 [00:12<00:15,  6.97it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  46%|█████▌      | 90/195 [00:13<00:15,  6.77it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  47%|█████▌      | 91/195 [00:13<00:15,  6.83it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  47%|█████▋      | 92/195 [00:13<00:14,  6.88it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  48%|█████▋      | 93/195 [00:13<00:14,  6.94it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  48%|█████▊      | 94/195 [00:13<00:14,  7.00it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  49%|█████▊      | 95/195 [00:13<00:14,  7.07it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  49%|█████▉      | 96/195 [00:13<00:13,  7.13it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  50%|█████▉      | 97/195 [00:13<00:13,  7.16it/s, loss=0.227, v_num=0]\u001b[A\n",
      "Epoch 4:  50%|█████▉      | 97/195 [00:13<00:13,  7.01it/s, loss=0.227, v_num=0]\u001b[A\n"
     ]
    }
   ],
   "source": [
    "!python src/train_multiFX.py \\\n",
    "        resnet18 \\\n",
    "        dataset/generated/guitarset10/gen_multiFX_05252023 \\\n",
    "        --max_epoch 10 \\\n",
    "        --gpus 1 \\\n",
    "        --batch_size 16 \\\n",
    "        --val_check_interval 0.5 \\\n",
    "        --random_seed 42 \\\n",
    "        --log_class_loss true \\\n",
    "        --with_clean true"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from src.transforms import MelSpectrogramDBTransform\n",
    "from src.model.resnet import resnet18\n",
    "from src.model.pl_wrapper import LightningWrapper"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "LightningWrapper(\n",
       "  (model): resnet18(\n",
       "    (conv1): Conv2d(2, 16, kernel_size=(5, 5), stride=(2, 2), padding=(2, 2), bias=False)\n",
       "    (bn1): BatchNorm2d(16, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "    (relu): ReLU(inplace=True)\n",
       "    (maxpool): MaxPool2d(kernel_size=3, stride=(1, 2), padding=1, dilation=1, ceil_mode=False)\n",
       "    (layer_1): Sequential(\n",
       "      (0): BasicBlock(\n",
       "        (conv1): Conv2d(16, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace=True)\n",
       "        (conv2): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (downsample): Sequential(\n",
       "          (0): Conv2d(16, 32, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
       "          (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        )\n",
       "      )\n",
       "      (1): BasicBlock(\n",
       "        (conv1): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace=True)\n",
       "        (conv2): Conv2d(32, 32, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "    )\n",
       "    (layer_2): Sequential(\n",
       "      (0): BasicBlock(\n",
       "        (conv1): Conv2d(32, 32, kernel_size=(5, 1), stride=(2, 2), padding=(2, 0), bias=False)\n",
       "        (bn1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace=True)\n",
       "        (conv2): Conv2d(32, 32, kernel_size=(5, 1), stride=(1, 1), padding=(2, 0), bias=False)\n",
       "        (bn2): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (downsample): Sequential(\n",
       "          (0): Conv2d(32, 32, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
       "          (1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        )\n",
       "      )\n",
       "      (1): BasicBlock(\n",
       "        (conv1): Conv2d(32, 32, kernel_size=(5, 1), stride=(1, 1), padding=(2, 0), bias=False)\n",
       "        (bn1): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace=True)\n",
       "        (conv2): Conv2d(32, 32, kernel_size=(5, 1), stride=(1, 1), padding=(2, 0), bias=False)\n",
       "        (bn2): BatchNorm2d(32, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "    )\n",
       "    (layer_3): Sequential(\n",
       "      (0): BasicBlock(\n",
       "        (conv1): Conv2d(32, 64, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace=True)\n",
       "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (downsample): Sequential(\n",
       "          (0): Conv2d(32, 64, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
       "          (1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        )\n",
       "      )\n",
       "      (1): BasicBlock(\n",
       "        (conv1): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace=True)\n",
       "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "    )\n",
       "    (layer_4): Sequential(\n",
       "      (0): BasicBlock(\n",
       "        (conv1): Conv2d(64, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace=True)\n",
       "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (downsample): Sequential(\n",
       "          (0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
       "          (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        )\n",
       "      )\n",
       "      (1): BasicBlock(\n",
       "        (conv1): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "        (relu): ReLU(inplace=True)\n",
       "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
       "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
       "      )\n",
       "    )\n",
       "    (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))\n",
       "    (fc): Linear(in_features=128, out_features=13, bias=True)\n",
       "  )\n",
       "  (transform): MelSpectrogramDBTransform(\n",
       "    (specer): MelSpectrogram(\n",
       "      (spectrogram): Spectrogram()\n",
       "      (mel_scale): MelScale()\n",
       "    )\n",
       "    (dBer): AmplitudeToDB()\n",
       "  )\n",
       ")"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# load trained model, please change the path accordingly\n",
    "ckpt_path = \"lightning_logs/multiFX_resnet18_[1,5]_with_clean/version_0/checkpoints/epoch=4-step=373.ckpt\"\n",
    "\n",
    "model = LightningWrapper.load_from_checkpoint(\n",
    "    checkpoint_path=ckpt_path,\n",
    "    model=resnet18(in_channels=2, num_classes=13),\n",
    "    with_clean=True,\n",
    "    lr=1e-3,\n",
    "    transform=MelSpectrogramDBTransform(),\n",
    ")\n",
    "model.eval()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# load validation set\n",
    "valid_set = MultiFXDataset(\"dataset/generated/guitarset10/gen_multiFX_05252023\", \"valid\", None)\n",
    "valid_loader = torch.utils.data.DataLoader(valid_set, batch_size=64, num_workers=2)\n",
    "classes = list(valid_set.settings[\"fx_params\"].keys())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 14/14 [00:10<00:00,  1.32it/s]\n"
     ]
    }
   ],
   "source": [
    "# run inference\n",
    "\n",
    "model.cuda()\n",
    "y = torch.randn(1, 13, requires_grad=False).cuda()\n",
    "y_hat = torch.randn(1, 13, requires_grad=False).cuda()\n",
    "with torch.no_grad():\n",
    "    for batch in tqdm.tqdm(valid_loader):\n",
    "        x, y_batch = batch\n",
    "        y_hat_batch = model(x.cuda())\n",
    "        y = torch.vstack([y, y_batch.cuda()])\n",
    "        y_hat = torch.vstack([y_hat, y_hat_batch.cuda()])\n",
    "        del x, y_batch, y_hat_batch\n",
    "        torch.cuda.empty_cache()\n",
    "y_hat = torch.sigmoid(y_hat) > 0.5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1600x1600 with 16 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# generate per-class confusion matrix\n",
    "\n",
    "fig, ax = plt.subplots(4, 4)\n",
    "fig.set_size_inches(16, 16)\n",
    "cms = sklearn.metrics.multilabel_confusion_matrix(\n",
    "    y[1:].cpu().to(torch.int), y_hat[1:].cpu().to(torch.int)\n",
    ")\n",
    "for i in range(len(classes)):\n",
    "    disp = sklearn.metrics.ConfusionMatrixDisplay(\n",
    "        cms[i], display_labels=None,\n",
    "    )\n",
    "    axis = ax[i // 4][i % 4]\n",
    "    disp.plot(ax=axis, colorbar=False)\n",
    "    TP, FP, FN, TN = cms[i][0][0], cms[i][0][1], cms[i][1][0], cms[i][1][1]\n",
    "    P = TP / (TP + FP)\n",
    "    R = TP / (TP + FN)\n",
    "    axis.set_title(f\"{classes[i]} $F_1$: ${2 * (P * R) / (P + R):.3f}$\")\n",
    "    axis.set_xlabel(\"\")\n",
    "    axis.set_ylabel(\"\")\n",
    "ax[3][1].axis(\"off\")\n",
    "ax[3][2].axis(\"off\")\n",
    "ax[3][3].axis(\"off\")\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "sfx_learner",
   "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": "3.10.11"
  },
  "vscode": {
   "interpreter": {
    "hash": "6624702b29e579fef95ef694d7db523ba62a38cd4b7b4921e207e4d20600c502"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}