notebooks/MCTN-MOSI.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "49e0ad50-a2a2-4b96-b360-6ebd4ba790f6",
   "metadata": {},
   "source": [
    "# Multimodal Cyclic Translation Network (MCTN) Model - MOSI"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1601d4a9-c176-4e94-9c84-250ae7ceb246",
   "metadata": {},
   "source": [
    "In this notebook, we implement the MCTN Model, and test it on CMU-MOSI dataset."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0e330111-2839-49fa-86f3-2204b6d24fc4",
   "metadata": {},
   "source": [
    "## Importing Libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "90273986-226c-4f13-a783-27d03eda1e21",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import sys\n",
    "import torch\n",
    "from torch import nn\n",
    "\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "sys.path.append(os.getcwd())\n",
    "sys.path.append(os.path.dirname(os.path.dirname(os.getcwd())))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3a8511b4-49ef-4df0-916e-07c8002ebcf9",
   "metadata": {},
   "source": [
    "## Importing Helper Functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "77dcaf0e-287e-4312-a80b-a67ca4221f73",
   "metadata": {},
   "outputs": [],
   "source": [
    "from helpers.data.get_data import get_dataloader\n",
    "from helpers.unimodals.common_models import GRU, MLP\n",
    "from helpers.fusions.MCTN import Encoder, Decoder\n",
    "from helpers.training.all_in_one import all_in_one_train\n",
    "from helpers.training.MCTN import train, test\n",
    "from helpers.utils.common_utils import get_default_device, save_results, format_time"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2bc745d1-4151-4894-af83-77780036ba7b",
   "metadata": {},
   "source": [
    "## Loading the Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "c7aa2aea-6160-4f43-a6f1-596f9c26bab7",
   "metadata": {},
   "outputs": [],
   "source": [
    "traindata, validdata, testdata = get_dataloader('/scratch/rrm9598/data/mosi_raw.pkl', robust_test=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "3e83b4bf-2838-4c0b-beb8-3d6150a369f3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "cuda\n"
     ]
    }
   ],
   "source": [
    "device = get_default_device()\n",
    "print(device)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "da5644c7-a89a-437f-8b75-3bf386b20dea",
   "metadata": {},
   "source": [
    "## Model Architecture"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "55070fb4-ea3e-4529-a717-7a0b3baf5275",
   "metadata": {},
   "outputs": [],
   "source": [
    "max_seq = 20\n",
    "feature_dim = 300\n",
    "hidden_dim = 32\n",
    "\n",
    "encoder0 = Encoder(feature_dim, hidden_dim, n_layers=1, dropout=0.0).to(device)\n",
    "decoder0 = Decoder(hidden_dim, feature_dim, n_layers=1, dropout=0.0).to(device)\n",
    "encoder1 = Encoder(hidden_dim, hidden_dim, n_layers=1, dropout=0.0).to(device)\n",
    "decoder1 = Decoder(hidden_dim, feature_dim, n_layers=1, dropout=0.0).to(device)\n",
    "\n",
    "reg_encoder = nn.GRU(hidden_dim, 32).to(device)\n",
    "head = MLP(32, 64, 1).to(device)\n",
    "\n",
    "allmodules = [encoder0, decoder0, encoder1, decoder1, reg_encoder, head]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1ccb00f7-9da1-43bb-b41e-f39eca505a76",
   "metadata": {},
   "source": [
    "## Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "334d11e7-12b3-476b-8a3e-ebc39c814460",
   "metadata": {},
   "outputs": [],
   "source": [
    "FILE_NAME = 'mosi_mctn_r0'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "49704ef0-fdb1-4f25-93a3-25083ce2d016",
   "metadata": {},
   "outputs": [],
   "source": [
    "def trainprocess():\n",
    "    global FILE_NAME\n",
    "    \n",
    "    train_losses, valid_losses = train(\n",
    "        traindata, validdata,\n",
    "        encoder0, decoder0, encoder1, decoder1,\n",
    "        reg_encoder, head,\n",
    "        criterion_t0=nn.MSELoss(), criterion_c=nn.MSELoss(),\n",
    "        criterion_t1=nn.MSELoss(), criterion_r=nn.L1Loss(),\n",
    "        max_seq_len=20,\n",
    "        mu_t0=0.01, mu_c=0.01, mu_t1=0.01,\n",
    "        dropout_p=0.15, early_stop=False, patience_num=15,\n",
    "        lr=1e-4, weight_decay=0.01, op_type=torch.optim.AdamW,\n",
    "        epoch=200, model_save=f'/scratch/rrm9598/Multimodal-Sentiment-Analysis/models/{FILE_NAME}.pt')\n",
    "    return train_losses, valid_losses"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "6a8c5934-2977-4f16-a472-f914e4677216",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch [1/200]\n",
      "Train Loss: 3.0502 | Regression Loss: 1.3307 | Embedding Loss: 1.7195\n",
      "Valid MAE: 1.4147 | Valid Accuracy: 57.4766% | Valid Recall: 61.1940%\n",
      "Saving Best\n",
      "\n",
      "Epoch [2/200]\n",
      "Train Loss: 3.0359 | Regression Loss: 1.3202 | Embedding Loss: 1.7156\n",
      "Valid MAE: 1.4157 | Valid Accuracy: 57.4766% | Valid Recall: 61.1940%\n",
      "\n",
      "Epoch [3/200]\n",
      "Train Loss: 3.0482 | Regression Loss: 1.3332 | Embedding Loss: 1.7149\n",
      "Valid MAE: 1.4135 | Valid Accuracy: 57.4766% | Valid Recall: 61.1940%\n",
      "Saving Best\n",
      "\n",
      "Epoch [4/200]\n",
      "Train Loss: 3.0388 | Regression Loss: 1.3224 | Embedding Loss: 1.7164\n",
      "Valid MAE: 1.4106 | Valid Accuracy: 57.4766% | Valid Recall: 61.1940%\n",
      "Saving Best\n",
      "\n",
      "Epoch [5/200]\n",
      "Train Loss: 3.0611 | Regression Loss: 1.3419 | Embedding Loss: 1.7191\n",
      "Valid MAE: 1.4133 | Valid Accuracy: 57.4766% | Valid Recall: 61.1940%\n",
      "\n",
      "Epoch [6/200]\n",
      "Train Loss: 3.0438 | Regression Loss: 1.3227 | Embedding Loss: 1.7212\n",
      "Valid MAE: 1.4137 | Valid Accuracy: 57.4766% | Valid Recall: 61.1940%\n",
      "\n",
      "Epoch [7/200]\n",
      "Train Loss: 3.0277 | Regression Loss: 1.3143 | Embedding Loss: 1.7134\n",
      "Valid MAE: 1.4116 | Valid Accuracy: 57.4766% | Valid Recall: 61.1940%\n",
      "\n",
      "Epoch [8/200]\n",
      "Train Loss: 3.0027 | Regression Loss: 1.2997 | Embedding Loss: 1.7030\n",
      "Valid MAE: 1.4105 | Valid Accuracy: 57.4766% | Valid Recall: 61.1940%\n",
      "Saving Best\n",
      "\n",
      "Epoch [9/200]\n",
      "Train Loss: 3.0155 | Regression Loss: 1.2979 | Embedding Loss: 1.7176\n",
      "Valid MAE: 1.4149 | Valid Accuracy: 57.4766% | Valid Recall: 61.1940%\n",
      "\n",
      "Epoch [10/200]\n",
      "Train Loss: 3.0451 | Regression Loss: 1.3154 | Embedding Loss: 1.7298\n",
      "Valid MAE: 1.4119 | Valid Accuracy: 57.4766% | Valid Recall: 61.1940%\n",
      "\n",
      "Epoch [11/200]\n",
      "Train Loss: 2.9917 | Regression Loss: 1.2825 | Embedding Loss: 1.7091\n",
      "Valid MAE: 1.4095 | Valid Accuracy: 57.4766% | Valid Recall: 61.1940%\n",
      "Saving Best\n",
      "\n",
      "Epoch [12/200]\n",
      "Train Loss: 2.9980 | Regression Loss: 1.2878 | Embedding Loss: 1.7102\n",
      "Valid MAE: 1.4016 | Valid Accuracy: 57.4766% | Valid Recall: 61.1940%\n",
      "Saving Best\n",
      "\n",
      "Epoch [13/200]\n",
      "Train Loss: 3.0150 | Regression Loss: 1.2726 | Embedding Loss: 1.7424\n",
      "Valid MAE: 1.3966 | Valid Accuracy: 58.8785% | Valid Recall: 62.1891%\n",
      "Saving Best\n",
      "\n",
      "Epoch [14/200]\n",
      "Train Loss: 2.9020 | Regression Loss: 1.1915 | Embedding Loss: 1.7105\n",
      "Valid MAE: 1.4159 | Valid Accuracy: 57.4766% | Valid Recall: 54.7264%\n",
      "\n",
      "Epoch [15/200]\n",
      "Train Loss: 2.8542 | Regression Loss: 1.1426 | Embedding Loss: 1.7116\n",
      "Valid MAE: 1.3750 | Valid Accuracy: 69.6262% | Valid Recall: 69.6517%\n",
      "Saving Best\n",
      "\n",
      "Epoch [16/200]\n",
      "Train Loss: 2.7430 | Regression Loss: 1.0362 | Embedding Loss: 1.7068\n",
      "Valid MAE: 1.3744 | Valid Accuracy: 72.8972% | Valid Recall: 71.1443%\n",
      "Saving Best\n",
      "\n",
      "Epoch [17/200]\n",
      "Train Loss: 2.6944 | Regression Loss: 0.9897 | Embedding Loss: 1.7047\n",
      "Valid MAE: 1.3381 | Valid Accuracy: 71.9626% | Valid Recall: 72.6368%\n",
      "Saving Best\n",
      "\n",
      "Epoch [18/200]\n",
      "Train Loss: 2.6816 | Regression Loss: 0.9722 | Embedding Loss: 1.7094\n",
      "Valid MAE: 1.3569 | Valid Accuracy: 75.2336% | Valid Recall: 74.1294%\n",
      "\n",
      "Epoch [19/200]\n",
      "Train Loss: 2.6834 | Regression Loss: 0.9709 | Embedding Loss: 1.7124\n",
      "Valid MAE: 1.3645 | Valid Accuracy: 75.2336% | Valid Recall: 73.6318%\n",
      "\n",
      "Epoch [20/200]\n",
      "Train Loss: 2.6544 | Regression Loss: 0.9523 | Embedding Loss: 1.7021\n",
      "Valid MAE: 1.3739 | Valid Accuracy: 74.2991% | Valid Recall: 72.6368%\n",
      "\n",
      "Epoch [21/200]\n",
      "Train Loss: 2.6177 | Regression Loss: 0.9152 | Embedding Loss: 1.7025\n",
      "Valid MAE: 1.3440 | Valid Accuracy: 73.3645% | Valid Recall: 74.6269%\n",
      "\n",
      "Epoch [22/200]\n",
      "Train Loss: 2.6601 | Regression Loss: 0.9460 | Embedding Loss: 1.7141\n",
      "Valid MAE: 1.3670 | Valid Accuracy: 76.1682% | Valid Recall: 74.6269%\n",
      "\n",
      "Epoch [23/200]\n",
      "Train Loss: 2.6597 | Regression Loss: 0.9444 | Embedding Loss: 1.7153\n",
      "Valid MAE: 1.3779 | Valid Accuracy: 69.1589% | Valid Recall: 67.1642%\n",
      "\n",
      "Epoch [24/200]\n",
      "Train Loss: 2.6522 | Regression Loss: 0.9507 | Embedding Loss: 1.7016\n",
      "Valid MAE: 1.3604 | Valid Accuracy: 76.6355% | Valid Recall: 77.6119%\n",
      "\n",
      "Epoch [25/200]\n",
      "Train Loss: 2.6114 | Regression Loss: 0.9116 | Embedding Loss: 1.6998\n",
      "Valid MAE: 1.3636 | Valid Accuracy: 75.2336% | Valid Recall: 73.6318%\n",
      "\n",
      "Epoch [26/200]\n",
      "Train Loss: 2.6208 | Regression Loss: 0.9187 | Embedding Loss: 1.7022\n",
      "Valid MAE: 1.3608 | Valid Accuracy: 76.6355% | Valid Recall: 75.6219%\n",
      "\n",
      "Epoch [27/200]\n",
      "Train Loss: 2.6380 | Regression Loss: 0.9107 | Embedding Loss: 1.7274\n",
      "Valid MAE: 1.3724 | Valid Accuracy: 71.9626% | Valid Recall: 70.1493%\n",
      "\n",
      "Epoch [28/200]\n",
      "Train Loss: 2.5829 | Regression Loss: 0.8826 | Embedding Loss: 1.7003\n",
      "Valid MAE: 1.3751 | Valid Accuracy: 71.4953% | Valid Recall: 69.6517%\n",
      "\n",
      "Epoch [29/200]\n",
      "Train Loss: 2.5707 | Regression Loss: 0.8678 | Embedding Loss: 1.7029\n",
      "Valid MAE: 1.3616 | Valid Accuracy: 74.7664% | Valid Recall: 73.6318%\n",
      "\n",
      "Epoch [30/200]\n",
      "Train Loss: 2.6004 | Regression Loss: 0.8887 | Embedding Loss: 1.7117\n",
      "Valid MAE: 1.3769 | Valid Accuracy: 77.1028% | Valid Recall: 75.6219%\n",
      "\n",
      "Epoch [31/200]\n",
      "Train Loss: 2.5715 | Regression Loss: 0.8681 | Embedding Loss: 1.7034\n",
      "Valid MAE: 1.3805 | Valid Accuracy: 68.6916% | Valid Recall: 66.6667%\n",
      "\n",
      "Epoch [32/200]\n",
      "Train Loss: 2.5731 | Regression Loss: 0.8729 | Embedding Loss: 1.7003\n",
      "Valid MAE: 1.3758 | Valid Accuracy: 70.0935% | Valid Recall: 68.1592%\n",
      "\n",
      "Epoch [33/200]\n",
      "Train Loss: 2.5718 | Regression Loss: 0.8664 | Embedding Loss: 1.7054\n",
      "Valid MAE: 1.3803 | Valid Accuracy: 68.6916% | Valid Recall: 66.6667%\n",
      "\n",
      "Epoch [34/200]\n",
      "Train Loss: 2.5400 | Regression Loss: 0.8379 | Embedding Loss: 1.7022\n",
      "Valid MAE: 1.3710 | Valid Accuracy: 70.5607% | Valid Recall: 68.6567%\n",
      "\n",
      "Epoch [35/200]\n",
      "Train Loss: 2.5449 | Regression Loss: 0.8390 | Embedding Loss: 1.7059\n",
      "Valid MAE: 1.3667 | Valid Accuracy: 77.1028% | Valid Recall: 76.1194%\n",
      "\n",
      "Epoch [36/200]\n",
      "Train Loss: 2.5746 | Regression Loss: 0.8554 | Embedding Loss: 1.7192\n",
      "Valid MAE: 1.3546 | Valid Accuracy: 78.5047% | Valid Recall: 78.6070%\n",
      "\n",
      "Epoch [37/200]\n",
      "Train Loss: 2.5486 | Regression Loss: 0.8447 | Embedding Loss: 1.7039\n",
      "Valid MAE: 1.3722 | Valid Accuracy: 73.8318% | Valid Recall: 72.1393%\n",
      "\n",
      "Epoch [38/200]\n",
      "Train Loss: 2.5268 | Regression Loss: 0.8256 | Embedding Loss: 1.7012\n",
      "Valid MAE: 1.3796 | Valid Accuracy: 69.1589% | Valid Recall: 67.1642%\n",
      "\n",
      "Epoch [39/200]\n",
      "Train Loss: 2.5204 | Regression Loss: 0.7985 | Embedding Loss: 1.7219\n",
      "Valid MAE: 1.3741 | Valid Accuracy: 70.0935% | Valid Recall: 68.1592%\n",
      "\n",
      "Epoch [40/200]\n",
      "Train Loss: 2.5538 | Regression Loss: 0.8287 | Embedding Loss: 1.7250\n",
      "Valid MAE: 1.3658 | Valid Accuracy: 73.8318% | Valid Recall: 72.1393%\n",
      "\n",
      "Epoch [41/200]\n",
      "Train Loss: 2.5580 | Regression Loss: 0.8339 | Embedding Loss: 1.7242\n",
      "Valid MAE: 1.3676 | Valid Accuracy: 72.8972% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [42/200]\n",
      "Train Loss: 2.5461 | Regression Loss: 0.8378 | Embedding Loss: 1.7082\n",
      "Valid MAE: 1.3563 | Valid Accuracy: 78.0374% | Valid Recall: 77.1144%\n",
      "\n",
      "Epoch [43/200]\n",
      "Train Loss: 2.5057 | Regression Loss: 0.8018 | Embedding Loss: 1.7039\n",
      "Valid MAE: 1.3640 | Valid Accuracy: 73.3645% | Valid Recall: 71.6418%\n",
      "\n",
      "Epoch [44/200]\n",
      "Train Loss: 2.4998 | Regression Loss: 0.7894 | Embedding Loss: 1.7104\n",
      "Valid MAE: 1.3629 | Valid Accuracy: 78.0374% | Valid Recall: 77.1144%\n",
      "\n",
      "Epoch [45/200]\n",
      "Train Loss: 2.5002 | Regression Loss: 0.8035 | Embedding Loss: 1.6967\n",
      "Valid MAE: 1.3640 | Valid Accuracy: 76.6355% | Valid Recall: 75.6219%\n",
      "\n",
      "Epoch [46/200]\n",
      "Train Loss: 2.4642 | Regression Loss: 0.7620 | Embedding Loss: 1.7021\n",
      "Valid MAE: 1.3622 | Valid Accuracy: 75.2336% | Valid Recall: 74.1294%\n",
      "\n",
      "Epoch [47/200]\n",
      "Train Loss: 2.4959 | Regression Loss: 0.7937 | Embedding Loss: 1.7022\n",
      "Valid MAE: 1.3588 | Valid Accuracy: 77.1028% | Valid Recall: 76.1194%\n",
      "\n",
      "Epoch [48/200]\n",
      "Train Loss: 2.5473 | Regression Loss: 0.8249 | Embedding Loss: 1.7224\n",
      "Valid MAE: 1.3634 | Valid Accuracy: 75.2336% | Valid Recall: 74.1294%\n",
      "\n",
      "Epoch [49/200]\n",
      "Train Loss: 2.4931 | Regression Loss: 0.7816 | Embedding Loss: 1.7115\n",
      "Valid MAE: 1.3691 | Valid Accuracy: 74.7664% | Valid Recall: 73.1343%\n",
      "\n",
      "Epoch [50/200]\n",
      "Train Loss: 2.4729 | Regression Loss: 0.7762 | Embedding Loss: 1.6967\n",
      "Valid MAE: 1.3733 | Valid Accuracy: 71.4953% | Valid Recall: 69.6517%\n",
      "\n",
      "Epoch [51/200]\n",
      "Train Loss: 2.4623 | Regression Loss: 0.7577 | Embedding Loss: 1.7046\n",
      "Valid MAE: 1.3691 | Valid Accuracy: 71.0280% | Valid Recall: 69.1542%\n",
      "\n",
      "Epoch [52/200]\n",
      "Train Loss: 2.4645 | Regression Loss: 0.7703 | Embedding Loss: 1.6942\n",
      "Valid MAE: 1.3849 | Valid Accuracy: 67.2897% | Valid Recall: 65.1741%\n",
      "\n",
      "Epoch [53/200]\n",
      "Train Loss: 2.4788 | Regression Loss: 0.7707 | Embedding Loss: 1.7081\n",
      "Valid MAE: 1.3803 | Valid Accuracy: 68.2243% | Valid Recall: 66.1692%\n",
      "\n",
      "Epoch [54/200]\n",
      "Train Loss: 2.4943 | Regression Loss: 0.7438 | Embedding Loss: 1.7505\n",
      "Valid MAE: 1.3729 | Valid Accuracy: 72.4299% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [55/200]\n",
      "Train Loss: 2.4950 | Regression Loss: 0.7844 | Embedding Loss: 1.7107\n",
      "Valid MAE: 1.3702 | Valid Accuracy: 73.8318% | Valid Recall: 72.6368%\n",
      "\n",
      "Epoch [56/200]\n",
      "Train Loss: 2.4724 | Regression Loss: 0.7669 | Embedding Loss: 1.7055\n",
      "Valid MAE: 1.3624 | Valid Accuracy: 74.2991% | Valid Recall: 73.1343%\n",
      "\n",
      "Epoch [57/200]\n",
      "Train Loss: 2.4725 | Regression Loss: 0.7723 | Embedding Loss: 1.7002\n",
      "Valid MAE: 1.3642 | Valid Accuracy: 73.3645% | Valid Recall: 71.6418%\n",
      "\n",
      "Epoch [58/200]\n",
      "Train Loss: 2.4616 | Regression Loss: 0.7485 | Embedding Loss: 1.7131\n",
      "Valid MAE: 1.3700 | Valid Accuracy: 71.4953% | Valid Recall: 69.6517%\n",
      "\n",
      "Epoch [59/200]\n",
      "Train Loss: 2.4633 | Regression Loss: 0.7531 | Embedding Loss: 1.7102\n",
      "Valid MAE: 1.3669 | Valid Accuracy: 72.4299% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [60/200]\n",
      "Train Loss: 2.4428 | Regression Loss: 0.7341 | Embedding Loss: 1.7087\n",
      "Valid MAE: 1.3680 | Valid Accuracy: 71.0280% | Valid Recall: 69.1542%\n",
      "\n",
      "Epoch [61/200]\n",
      "Train Loss: 2.4882 | Regression Loss: 0.7808 | Embedding Loss: 1.7074\n",
      "Valid MAE: 1.3698 | Valid Accuracy: 74.7664% | Valid Recall: 74.1294%\n",
      "\n",
      "Epoch [62/200]\n",
      "Train Loss: 2.4946 | Regression Loss: 0.7835 | Embedding Loss: 1.7112\n",
      "Valid MAE: 1.3689 | Valid Accuracy: 71.9626% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [63/200]\n",
      "Train Loss: 2.4308 | Regression Loss: 0.7352 | Embedding Loss: 1.6956\n",
      "Valid MAE: 1.3701 | Valid Accuracy: 71.9626% | Valid Recall: 70.1493%\n",
      "\n",
      "Epoch [64/200]\n",
      "Train Loss: 2.4863 | Regression Loss: 0.7832 | Embedding Loss: 1.7030\n",
      "Valid MAE: 1.3743 | Valid Accuracy: 71.9626% | Valid Recall: 70.1493%\n",
      "\n",
      "Epoch [65/200]\n",
      "Train Loss: 2.4402 | Regression Loss: 0.7255 | Embedding Loss: 1.7147\n",
      "Valid MAE: 1.3775 | Valid Accuracy: 66.8224% | Valid Recall: 64.6766%\n",
      "\n",
      "Epoch [66/200]\n",
      "Train Loss: 2.4764 | Regression Loss: 0.7638 | Embedding Loss: 1.7126\n",
      "Valid MAE: 1.3662 | Valid Accuracy: 72.8972% | Valid Recall: 72.1393%\n",
      "\n",
      "Epoch [67/200]\n",
      "Train Loss: 2.4870 | Regression Loss: 0.7846 | Embedding Loss: 1.7025\n",
      "Valid MAE: 1.3593 | Valid Accuracy: 75.2336% | Valid Recall: 74.6269%\n",
      "\n",
      "Epoch [68/200]\n",
      "Train Loss: 2.4753 | Regression Loss: 0.7720 | Embedding Loss: 1.7033\n",
      "Valid MAE: 1.3609 | Valid Accuracy: 73.3645% | Valid Recall: 72.6368%\n",
      "\n",
      "Epoch [69/200]\n",
      "Train Loss: 2.4552 | Regression Loss: 0.7439 | Embedding Loss: 1.7113\n",
      "Valid MAE: 1.3561 | Valid Accuracy: 73.8318% | Valid Recall: 73.1343%\n",
      "\n",
      "Epoch [70/200]\n",
      "Train Loss: 2.4178 | Regression Loss: 0.7107 | Embedding Loss: 1.7071\n",
      "Valid MAE: 1.3676 | Valid Accuracy: 72.4299% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [71/200]\n",
      "Train Loss: 2.4241 | Regression Loss: 0.7183 | Embedding Loss: 1.7058\n",
      "Valid MAE: 1.3775 | Valid Accuracy: 67.2897% | Valid Recall: 65.1741%\n",
      "\n",
      "Epoch [72/200]\n",
      "Train Loss: 2.4165 | Regression Loss: 0.7155 | Embedding Loss: 1.7010\n",
      "Valid MAE: 1.3727 | Valid Accuracy: 70.0935% | Valid Recall: 68.1592%\n",
      "\n",
      "Epoch [73/200]\n",
      "Train Loss: 2.4076 | Regression Loss: 0.6987 | Embedding Loss: 1.7089\n",
      "Valid MAE: 1.3707 | Valid Accuracy: 71.9626% | Valid Recall: 70.6468%\n",
      "\n",
      "Epoch [74/200]\n",
      "Train Loss: 2.4244 | Regression Loss: 0.7194 | Embedding Loss: 1.7050\n",
      "Valid MAE: 1.3535 | Valid Accuracy: 75.2336% | Valid Recall: 74.6269%\n",
      "\n",
      "Epoch [75/200]\n",
      "Train Loss: 2.4330 | Regression Loss: 0.7290 | Embedding Loss: 1.7040\n",
      "Valid MAE: 1.3619 | Valid Accuracy: 73.8318% | Valid Recall: 73.1343%\n",
      "\n",
      "Epoch [76/200]\n",
      "Train Loss: 2.4465 | Regression Loss: 0.7318 | Embedding Loss: 1.7147\n",
      "Valid MAE: 1.3613 | Valid Accuracy: 74.7664% | Valid Recall: 74.1294%\n",
      "\n",
      "Epoch [77/200]\n",
      "Train Loss: 2.3939 | Regression Loss: 0.6953 | Embedding Loss: 1.6985\n",
      "Valid MAE: 1.3623 | Valid Accuracy: 72.4299% | Valid Recall: 71.6418%\n",
      "\n",
      "Epoch [78/200]\n",
      "Train Loss: 2.4052 | Regression Loss: 0.6908 | Embedding Loss: 1.7144\n",
      "Valid MAE: 1.3734 | Valid Accuracy: 67.7570% | Valid Recall: 65.6716%\n",
      "\n",
      "Epoch [79/200]\n",
      "Train Loss: 2.4054 | Regression Loss: 0.6997 | Embedding Loss: 1.7056\n",
      "Valid MAE: 1.3670 | Valid Accuracy: 71.4953% | Valid Recall: 70.1493%\n",
      "\n",
      "Epoch [80/200]\n",
      "Train Loss: 2.3947 | Regression Loss: 0.6973 | Embedding Loss: 1.6974\n",
      "Valid MAE: 1.3615 | Valid Accuracy: 74.7664% | Valid Recall: 74.1294%\n",
      "\n",
      "Epoch [81/200]\n",
      "Train Loss: 2.3963 | Regression Loss: 0.7003 | Embedding Loss: 1.6961\n",
      "Valid MAE: 1.3703 | Valid Accuracy: 70.5607% | Valid Recall: 68.6567%\n",
      "\n",
      "Epoch [82/200]\n",
      "Train Loss: 2.3792 | Regression Loss: 0.6810 | Embedding Loss: 1.6982\n",
      "Valid MAE: 1.3704 | Valid Accuracy: 72.4299% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [83/200]\n",
      "Train Loss: 2.3997 | Regression Loss: 0.6935 | Embedding Loss: 1.7063\n",
      "Valid MAE: 1.3638 | Valid Accuracy: 74.2991% | Valid Recall: 73.6318%\n",
      "\n",
      "Epoch [84/200]\n",
      "Train Loss: 2.3634 | Regression Loss: 0.6633 | Embedding Loss: 1.7000\n",
      "Valid MAE: 1.3696 | Valid Accuracy: 69.6262% | Valid Recall: 68.1592%\n",
      "\n",
      "Epoch [85/200]\n",
      "Train Loss: 2.4131 | Regression Loss: 0.7068 | Embedding Loss: 1.7063\n",
      "Valid MAE: 1.3629 | Valid Accuracy: 73.8318% | Valid Recall: 73.1343%\n",
      "\n",
      "Epoch [86/200]\n",
      "Train Loss: 2.4053 | Regression Loss: 0.6882 | Embedding Loss: 1.7171\n",
      "Valid MAE: 1.3668 | Valid Accuracy: 72.8972% | Valid Recall: 72.1393%\n",
      "\n",
      "Epoch [87/200]\n",
      "Train Loss: 2.4123 | Regression Loss: 0.7018 | Embedding Loss: 1.7105\n",
      "Valid MAE: 1.3741 | Valid Accuracy: 67.2897% | Valid Recall: 65.6716%\n",
      "\n",
      "Epoch [88/200]\n",
      "Train Loss: 2.3765 | Regression Loss: 0.6765 | Embedding Loss: 1.7000\n",
      "Valid MAE: 1.3538 | Valid Accuracy: 75.7009% | Valid Recall: 75.6219%\n",
      "\n",
      "Epoch [89/200]\n",
      "Train Loss: 2.4203 | Regression Loss: 0.7096 | Embedding Loss: 1.7106\n",
      "Valid MAE: 1.3750 | Valid Accuracy: 67.2897% | Valid Recall: 65.6716%\n",
      "\n",
      "Epoch [90/200]\n",
      "Train Loss: 2.3747 | Regression Loss: 0.6707 | Embedding Loss: 1.7041\n",
      "Valid MAE: 1.3592 | Valid Accuracy: 74.2991% | Valid Recall: 73.6318%\n",
      "\n",
      "Epoch [91/200]\n",
      "Train Loss: 2.3559 | Regression Loss: 0.6510 | Embedding Loss: 1.7049\n",
      "Valid MAE: 1.3699 | Valid Accuracy: 68.2243% | Valid Recall: 66.6667%\n",
      "\n",
      "Epoch [92/200]\n",
      "Train Loss: 2.3725 | Regression Loss: 0.6604 | Embedding Loss: 1.7121\n",
      "Valid MAE: 1.3597 | Valid Accuracy: 74.7664% | Valid Recall: 74.6269%\n",
      "\n",
      "Epoch [93/200]\n",
      "Train Loss: 2.3572 | Regression Loss: 0.6603 | Embedding Loss: 1.6969\n",
      "Valid MAE: 1.3630 | Valid Accuracy: 75.2336% | Valid Recall: 74.6269%\n",
      "\n",
      "Epoch [94/200]\n",
      "Train Loss: 2.3583 | Regression Loss: 0.6518 | Embedding Loss: 1.7066\n",
      "Valid MAE: 1.3594 | Valid Accuracy: 74.7664% | Valid Recall: 74.6269%\n",
      "\n",
      "Epoch [95/200]\n",
      "Train Loss: 2.3972 | Regression Loss: 0.6884 | Embedding Loss: 1.7088\n",
      "Valid MAE: 1.3574 | Valid Accuracy: 75.2336% | Valid Recall: 75.1244%\n",
      "\n",
      "Epoch [96/200]\n",
      "Train Loss: 2.3643 | Regression Loss: 0.6630 | Embedding Loss: 1.7013\n",
      "Valid MAE: 1.3597 | Valid Accuracy: 74.2991% | Valid Recall: 73.6318%\n",
      "\n",
      "Epoch [97/200]\n",
      "Train Loss: 2.3594 | Regression Loss: 0.6610 | Embedding Loss: 1.6984\n",
      "Valid MAE: 1.3556 | Valid Accuracy: 74.2991% | Valid Recall: 74.1294%\n",
      "\n",
      "Epoch [98/200]\n",
      "Train Loss: 2.3411 | Regression Loss: 0.6304 | Embedding Loss: 1.7106\n",
      "Valid MAE: 1.3651 | Valid Accuracy: 71.4953% | Valid Recall: 70.6468%\n",
      "\n",
      "Epoch [99/200]\n",
      "Train Loss: 2.3604 | Regression Loss: 0.6533 | Embedding Loss: 1.7071\n",
      "Valid MAE: 1.3593 | Valid Accuracy: 74.7664% | Valid Recall: 74.1294%\n",
      "\n",
      "Epoch [100/200]\n",
      "Train Loss: 2.3777 | Regression Loss: 0.6458 | Embedding Loss: 1.7318\n",
      "Valid MAE: 1.3608 | Valid Accuracy: 74.7664% | Valid Recall: 74.6269%\n",
      "\n",
      "Epoch [101/200]\n",
      "Train Loss: 2.3385 | Regression Loss: 0.6415 | Embedding Loss: 1.6970\n",
      "Valid MAE: 1.3623 | Valid Accuracy: 69.6262% | Valid Recall: 68.6567%\n",
      "\n",
      "Epoch [102/200]\n",
      "Train Loss: 2.3544 | Regression Loss: 0.6579 | Embedding Loss: 1.6964\n",
      "Valid MAE: 1.3507 | Valid Accuracy: 75.2336% | Valid Recall: 75.1244%\n",
      "\n",
      "Epoch [103/200]\n",
      "Train Loss: 2.3495 | Regression Loss: 0.6496 | Embedding Loss: 1.6999\n",
      "Valid MAE: 1.3580 | Valid Accuracy: 72.8972% | Valid Recall: 72.6368%\n",
      "\n",
      "Epoch [104/200]\n",
      "Train Loss: 2.3438 | Regression Loss: 0.6475 | Embedding Loss: 1.6963\n",
      "Valid MAE: 1.3638 | Valid Accuracy: 72.8972% | Valid Recall: 72.6368%\n",
      "\n",
      "Epoch [105/200]\n",
      "Train Loss: 2.3652 | Regression Loss: 0.6474 | Embedding Loss: 1.7178\n",
      "Valid MAE: 1.3605 | Valid Accuracy: 72.4299% | Valid Recall: 72.1393%\n",
      "\n",
      "Epoch [106/200]\n",
      "Train Loss: 2.3143 | Regression Loss: 0.6144 | Embedding Loss: 1.6999\n",
      "Valid MAE: 1.3584 | Valid Accuracy: 72.4299% | Valid Recall: 72.1393%\n",
      "\n",
      "Epoch [107/200]\n",
      "Train Loss: 2.3349 | Regression Loss: 0.6298 | Embedding Loss: 1.7051\n",
      "Valid MAE: 1.3648 | Valid Accuracy: 69.1589% | Valid Recall: 67.6617%\n",
      "\n",
      "Epoch [108/200]\n",
      "Train Loss: 2.3777 | Regression Loss: 0.6704 | Embedding Loss: 1.7074\n",
      "Valid MAE: 1.3603 | Valid Accuracy: 71.9626% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [109/200]\n",
      "Train Loss: 2.3626 | Regression Loss: 0.6435 | Embedding Loss: 1.7191\n",
      "Valid MAE: 1.3587 | Valid Accuracy: 72.8972% | Valid Recall: 72.6368%\n",
      "\n",
      "Epoch [110/200]\n",
      "Train Loss: 2.3018 | Regression Loss: 0.6090 | Embedding Loss: 1.6928\n",
      "Valid MAE: 1.3533 | Valid Accuracy: 73.8318% | Valid Recall: 73.6318%\n",
      "\n",
      "Epoch [111/200]\n",
      "Train Loss: 2.3038 | Regression Loss: 0.6049 | Embedding Loss: 1.6988\n",
      "Valid MAE: 1.3527 | Valid Accuracy: 72.8972% | Valid Recall: 72.6368%\n",
      "\n",
      "Epoch [112/200]\n",
      "Train Loss: 2.3430 | Regression Loss: 0.6219 | Embedding Loss: 1.7211\n",
      "Valid MAE: 1.3579 | Valid Accuracy: 71.4953% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [113/200]\n",
      "Train Loss: 2.3287 | Regression Loss: 0.6170 | Embedding Loss: 1.7116\n",
      "Valid MAE: 1.3585 | Valid Accuracy: 73.3645% | Valid Recall: 73.1343%\n",
      "\n",
      "Epoch [114/200]\n",
      "Train Loss: 2.3634 | Regression Loss: 0.6407 | Embedding Loss: 1.7227\n",
      "Valid MAE: 1.3674 | Valid Accuracy: 71.0280% | Valid Recall: 70.1493%\n",
      "\n",
      "Epoch [115/200]\n",
      "Train Loss: 2.3163 | Regression Loss: 0.6129 | Embedding Loss: 1.7034\n",
      "Valid MAE: 1.3628 | Valid Accuracy: 71.0280% | Valid Recall: 70.6468%\n",
      "\n",
      "Epoch [116/200]\n",
      "Train Loss: 2.3169 | Regression Loss: 0.6041 | Embedding Loss: 1.7129\n",
      "Valid MAE: 1.3564 | Valid Accuracy: 71.4953% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [117/200]\n",
      "Train Loss: 2.2912 | Regression Loss: 0.5853 | Embedding Loss: 1.7059\n",
      "Valid MAE: 1.3657 | Valid Accuracy: 71.4953% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [118/200]\n",
      "Train Loss: 2.3004 | Regression Loss: 0.5901 | Embedding Loss: 1.7103\n",
      "Valid MAE: 1.3583 | Valid Accuracy: 71.9626% | Valid Recall: 71.6418%\n",
      "\n",
      "Epoch [119/200]\n",
      "Train Loss: 2.3009 | Regression Loss: 0.5968 | Embedding Loss: 1.7041\n",
      "Valid MAE: 1.3676 | Valid Accuracy: 69.1589% | Valid Recall: 68.1592%\n",
      "\n",
      "Epoch [120/200]\n",
      "Train Loss: 2.3118 | Regression Loss: 0.6079 | Embedding Loss: 1.7039\n",
      "Valid MAE: 1.3560 | Valid Accuracy: 72.8972% | Valid Recall: 72.6368%\n",
      "\n",
      "Epoch [121/200]\n",
      "Train Loss: 2.3195 | Regression Loss: 0.6184 | Embedding Loss: 1.7012\n",
      "Valid MAE: 1.3548 | Valid Accuracy: 71.9626% | Valid Recall: 71.6418%\n",
      "\n",
      "Epoch [122/200]\n",
      "Train Loss: 2.3252 | Regression Loss: 0.6264 | Embedding Loss: 1.6988\n",
      "Valid MAE: 1.3627 | Valid Accuracy: 71.4953% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [123/200]\n",
      "Train Loss: 2.3195 | Regression Loss: 0.6107 | Embedding Loss: 1.7088\n",
      "Valid MAE: 1.3561 | Valid Accuracy: 71.4953% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [124/200]\n",
      "Train Loss: 2.2663 | Regression Loss: 0.5662 | Embedding Loss: 1.7001\n",
      "Valid MAE: 1.3563 | Valid Accuracy: 71.0280% | Valid Recall: 70.6468%\n",
      "\n",
      "Epoch [125/200]\n",
      "Train Loss: 2.2630 | Regression Loss: 0.5676 | Embedding Loss: 1.6954\n",
      "Valid MAE: 1.3527 | Valid Accuracy: 71.0280% | Valid Recall: 70.6468%\n",
      "\n",
      "Epoch [126/200]\n",
      "Train Loss: 2.3074 | Regression Loss: 0.6051 | Embedding Loss: 1.7023\n",
      "Valid MAE: 1.3549 | Valid Accuracy: 71.0280% | Valid Recall: 70.6468%\n",
      "\n",
      "Epoch [127/200]\n",
      "Train Loss: 2.3058 | Regression Loss: 0.5893 | Embedding Loss: 1.7165\n",
      "Valid MAE: 1.3599 | Valid Accuracy: 71.0280% | Valid Recall: 70.6468%\n",
      "\n",
      "Epoch [128/200]\n",
      "Train Loss: 2.3318 | Regression Loss: 0.6165 | Embedding Loss: 1.7153\n",
      "Valid MAE: 1.3533 | Valid Accuracy: 71.4953% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [129/200]\n",
      "Train Loss: 2.2957 | Regression Loss: 0.5969 | Embedding Loss: 1.6989\n",
      "Valid MAE: 1.3572 | Valid Accuracy: 71.0280% | Valid Recall: 70.6468%\n",
      "\n",
      "Epoch [130/200]\n",
      "Train Loss: 2.2984 | Regression Loss: 0.5955 | Embedding Loss: 1.7029\n",
      "Valid MAE: 1.3578 | Valid Accuracy: 70.5607% | Valid Recall: 70.6468%\n",
      "\n",
      "Epoch [131/200]\n",
      "Train Loss: 2.2815 | Regression Loss: 0.5578 | Embedding Loss: 1.7237\n",
      "Valid MAE: 1.3552 | Valid Accuracy: 70.5607% | Valid Recall: 70.1493%\n",
      "\n",
      "Epoch [132/200]\n",
      "Train Loss: 2.2773 | Regression Loss: 0.5704 | Embedding Loss: 1.7069\n",
      "Valid MAE: 1.3536 | Valid Accuracy: 71.9626% | Valid Recall: 71.6418%\n",
      "\n",
      "Epoch [133/200]\n",
      "Train Loss: 2.2941 | Regression Loss: 0.5799 | Embedding Loss: 1.7142\n",
      "Valid MAE: 1.3492 | Valid Accuracy: 73.3645% | Valid Recall: 73.1343%\n",
      "\n",
      "Epoch [134/200]\n",
      "Train Loss: 2.2701 | Regression Loss: 0.5532 | Embedding Loss: 1.7169\n",
      "Valid MAE: 1.3599 | Valid Accuracy: 69.6262% | Valid Recall: 69.1542%\n",
      "\n",
      "Epoch [135/200]\n",
      "Train Loss: 2.3050 | Regression Loss: 0.5774 | Embedding Loss: 1.7276\n",
      "Valid MAE: 1.3503 | Valid Accuracy: 72.8972% | Valid Recall: 73.1343%\n",
      "\n",
      "Epoch [136/200]\n",
      "Train Loss: 2.3011 | Regression Loss: 0.5775 | Embedding Loss: 1.7236\n",
      "Valid MAE: 1.3579 | Valid Accuracy: 68.2243% | Valid Recall: 67.6617%\n",
      "\n",
      "Epoch [137/200]\n",
      "Train Loss: 2.3085 | Regression Loss: 0.5939 | Embedding Loss: 1.7146\n",
      "Valid MAE: 1.3450 | Valid Accuracy: 73.8318% | Valid Recall: 74.6269%\n",
      "\n",
      "Epoch [138/200]\n",
      "Train Loss: 2.2864 | Regression Loss: 0.5893 | Embedding Loss: 1.6971\n",
      "Valid MAE: 1.3565 | Valid Accuracy: 69.1589% | Valid Recall: 68.6567%\n",
      "\n",
      "Epoch [139/200]\n",
      "Train Loss: 2.2980 | Regression Loss: 0.5834 | Embedding Loss: 1.7146\n",
      "Valid MAE: 1.3587 | Valid Accuracy: 70.0935% | Valid Recall: 69.6517%\n",
      "\n",
      "Epoch [140/200]\n",
      "Train Loss: 2.2741 | Regression Loss: 0.5577 | Embedding Loss: 1.7164\n",
      "Valid MAE: 1.3585 | Valid Accuracy: 69.6262% | Valid Recall: 69.1542%\n",
      "\n",
      "Epoch [141/200]\n",
      "Train Loss: 2.2553 | Regression Loss: 0.5596 | Embedding Loss: 1.6957\n",
      "Valid MAE: 1.3519 | Valid Accuracy: 71.9626% | Valid Recall: 71.6418%\n",
      "\n",
      "Epoch [142/200]\n",
      "Train Loss: 2.2770 | Regression Loss: 0.5620 | Embedding Loss: 1.7150\n",
      "Valid MAE: 1.3547 | Valid Accuracy: 69.1589% | Valid Recall: 68.6567%\n",
      "\n",
      "Epoch [143/200]\n",
      "Train Loss: 2.2952 | Regression Loss: 0.5712 | Embedding Loss: 1.7241\n",
      "Valid MAE: 1.3531 | Valid Accuracy: 70.5607% | Valid Recall: 70.1493%\n",
      "\n",
      "Epoch [144/200]\n",
      "Train Loss: 2.2769 | Regression Loss: 0.5718 | Embedding Loss: 1.7050\n",
      "Valid MAE: 1.3526 | Valid Accuracy: 70.5607% | Valid Recall: 70.1493%\n",
      "\n",
      "Epoch [145/200]\n",
      "Train Loss: 2.2649 | Regression Loss: 0.5684 | Embedding Loss: 1.6965\n",
      "Valid MAE: 1.3483 | Valid Accuracy: 72.4299% | Valid Recall: 73.1343%\n",
      "\n",
      "Epoch [146/200]\n",
      "Train Loss: 2.2637 | Regression Loss: 0.5542 | Embedding Loss: 1.7095\n",
      "Valid MAE: 1.3473 | Valid Accuracy: 71.9626% | Valid Recall: 72.6368%\n",
      "\n",
      "Epoch [147/200]\n",
      "Train Loss: 2.2841 | Regression Loss: 0.5802 | Embedding Loss: 1.7039\n",
      "Valid MAE: 1.3573 | Valid Accuracy: 70.5607% | Valid Recall: 70.6468%\n",
      "\n",
      "Epoch [148/200]\n",
      "Train Loss: 2.2817 | Regression Loss: 0.5760 | Embedding Loss: 1.7057\n",
      "Valid MAE: 1.3556 | Valid Accuracy: 72.8972% | Valid Recall: 72.6368%\n",
      "\n",
      "Epoch [149/200]\n",
      "Train Loss: 2.2867 | Regression Loss: 0.5742 | Embedding Loss: 1.7126\n",
      "Valid MAE: 1.3522 | Valid Accuracy: 71.9626% | Valid Recall: 72.1393%\n",
      "\n",
      "Epoch [150/200]\n",
      "Train Loss: 2.2369 | Regression Loss: 0.5312 | Embedding Loss: 1.7056\n",
      "Valid MAE: 1.3627 | Valid Accuracy: 68.6916% | Valid Recall: 68.6567%\n",
      "\n",
      "Epoch [151/200]\n",
      "Train Loss: 2.2848 | Regression Loss: 0.5823 | Embedding Loss: 1.7026\n",
      "Valid MAE: 1.3492 | Valid Accuracy: 70.0935% | Valid Recall: 70.6468%\n",
      "\n",
      "Epoch [152/200]\n",
      "Train Loss: 2.2533 | Regression Loss: 0.5427 | Embedding Loss: 1.7107\n",
      "Valid MAE: 1.3485 | Valid Accuracy: 71.0280% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [153/200]\n",
      "Train Loss: 2.2334 | Regression Loss: 0.5190 | Embedding Loss: 1.7144\n",
      "Valid MAE: 1.3476 | Valid Accuracy: 71.0280% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [154/200]\n",
      "Train Loss: 2.2825 | Regression Loss: 0.5733 | Embedding Loss: 1.7092\n",
      "Valid MAE: 1.3442 | Valid Accuracy: 77.1028% | Valid Recall: 78.1095%\n",
      "\n",
      "Epoch [155/200]\n",
      "Train Loss: 2.2706 | Regression Loss: 0.5671 | Embedding Loss: 1.7035\n",
      "Valid MAE: 1.3493 | Valid Accuracy: 73.3645% | Valid Recall: 74.1294%\n",
      "\n",
      "Epoch [156/200]\n",
      "Train Loss: 2.2389 | Regression Loss: 0.5445 | Embedding Loss: 1.6945\n",
      "Valid MAE: 1.3540 | Valid Accuracy: 71.4953% | Valid Recall: 72.1393%\n",
      "\n",
      "Epoch [157/200]\n",
      "Train Loss: 2.2439 | Regression Loss: 0.5336 | Embedding Loss: 1.7103\n",
      "Valid MAE: 1.3558 | Valid Accuracy: 71.0280% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [158/200]\n",
      "Train Loss: 2.2558 | Regression Loss: 0.5474 | Embedding Loss: 1.7084\n",
      "Valid MAE: 1.3477 | Valid Accuracy: 71.0280% | Valid Recall: 71.6418%\n",
      "\n",
      "Epoch [159/200]\n",
      "Train Loss: 2.2623 | Regression Loss: 0.5469 | Embedding Loss: 1.7154\n",
      "Valid MAE: 1.3525 | Valid Accuracy: 70.0935% | Valid Recall: 70.6468%\n",
      "\n",
      "Epoch [160/200]\n",
      "Train Loss: 2.2451 | Regression Loss: 0.5410 | Embedding Loss: 1.7041\n",
      "Valid MAE: 1.3414 | Valid Accuracy: 75.2336% | Valid Recall: 76.1194%\n",
      "\n",
      "Epoch [161/200]\n",
      "Train Loss: 2.2358 | Regression Loss: 0.5346 | Embedding Loss: 1.7012\n",
      "Valid MAE: 1.3460 | Valid Accuracy: 71.4953% | Valid Recall: 71.6418%\n",
      "\n",
      "Epoch [162/200]\n",
      "Train Loss: 2.2646 | Regression Loss: 0.5701 | Embedding Loss: 1.6945\n",
      "Valid MAE: 1.3483 | Valid Accuracy: 72.4299% | Valid Recall: 73.1343%\n",
      "\n",
      "Epoch [163/200]\n",
      "Train Loss: 2.2639 | Regression Loss: 0.5468 | Embedding Loss: 1.7171\n",
      "Valid MAE: 1.3426 | Valid Accuracy: 74.7664% | Valid Recall: 75.6219%\n",
      "\n",
      "Epoch [164/200]\n",
      "Train Loss: 2.2889 | Regression Loss: 0.5923 | Embedding Loss: 1.6966\n",
      "Valid MAE: 1.3596 | Valid Accuracy: 71.4953% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [165/200]\n",
      "Train Loss: 2.2668 | Regression Loss: 0.5502 | Embedding Loss: 1.7166\n",
      "Valid MAE: 1.3466 | Valid Accuracy: 71.4953% | Valid Recall: 72.1393%\n",
      "\n",
      "Epoch [166/200]\n",
      "Train Loss: 2.2141 | Regression Loss: 0.5135 | Embedding Loss: 1.7006\n",
      "Valid MAE: 1.3509 | Valid Accuracy: 73.8318% | Valid Recall: 74.6269%\n",
      "\n",
      "Epoch [167/200]\n",
      "Train Loss: 2.2625 | Regression Loss: 0.5485 | Embedding Loss: 1.7140\n",
      "Valid MAE: 1.3523 | Valid Accuracy: 71.0280% | Valid Recall: 70.6468%\n",
      "\n",
      "Epoch [168/200]\n",
      "Train Loss: 2.2310 | Regression Loss: 0.5281 | Embedding Loss: 1.7029\n",
      "Valid MAE: 1.3571 | Valid Accuracy: 70.0935% | Valid Recall: 69.6517%\n",
      "\n",
      "Epoch [169/200]\n",
      "Train Loss: 2.2490 | Regression Loss: 0.5422 | Embedding Loss: 1.7067\n",
      "Valid MAE: 1.3449 | Valid Accuracy: 77.1028% | Valid Recall: 78.1095%\n",
      "\n",
      "Epoch [170/200]\n",
      "Train Loss: 2.2460 | Regression Loss: 0.5394 | Embedding Loss: 1.7065\n",
      "Valid MAE: 1.3495 | Valid Accuracy: 71.0280% | Valid Recall: 71.6418%\n",
      "\n",
      "Epoch [171/200]\n",
      "Train Loss: 2.2630 | Regression Loss: 0.5562 | Embedding Loss: 1.7067\n",
      "Valid MAE: 1.3470 | Valid Accuracy: 71.0280% | Valid Recall: 71.6418%\n",
      "\n",
      "Epoch [172/200]\n",
      "Train Loss: 2.2295 | Regression Loss: 0.5114 | Embedding Loss: 1.7181\n",
      "Valid MAE: 1.3471 | Valid Accuracy: 71.9626% | Valid Recall: 72.1393%\n",
      "\n",
      "Epoch [173/200]\n",
      "Train Loss: 2.2175 | Regression Loss: 0.5146 | Embedding Loss: 1.7030\n",
      "Valid MAE: 1.3424 | Valid Accuracy: 72.8972% | Valid Recall: 73.6318%\n",
      "\n",
      "Epoch [174/200]\n",
      "Train Loss: 2.1991 | Regression Loss: 0.5008 | Embedding Loss: 1.6983\n",
      "Valid MAE: 1.3441 | Valid Accuracy: 71.4953% | Valid Recall: 72.1393%\n",
      "\n",
      "Epoch [175/200]\n",
      "Train Loss: 2.2232 | Regression Loss: 0.5174 | Embedding Loss: 1.7058\n",
      "Valid MAE: 1.3518 | Valid Accuracy: 71.0280% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [176/200]\n",
      "Train Loss: 2.2449 | Regression Loss: 0.5285 | Embedding Loss: 1.7164\n",
      "Valid MAE: 1.3350 | Valid Accuracy: 73.8318% | Valid Recall: 74.6269%\n",
      "Saving Best\n",
      "\n",
      "Epoch [177/200]\n",
      "Train Loss: 2.2352 | Regression Loss: 0.5286 | Embedding Loss: 1.7066\n",
      "Valid MAE: 1.3409 | Valid Accuracy: 77.1028% | Valid Recall: 78.1095%\n",
      "\n",
      "Epoch [178/200]\n",
      "Train Loss: 2.2443 | Regression Loss: 0.5262 | Embedding Loss: 1.7180\n",
      "Valid MAE: 1.3388 | Valid Accuracy: 75.2336% | Valid Recall: 76.1194%\n",
      "\n",
      "Epoch [179/200]\n",
      "Train Loss: 2.2160 | Regression Loss: 0.5212 | Embedding Loss: 1.6948\n",
      "Valid MAE: 1.3389 | Valid Accuracy: 74.2991% | Valid Recall: 75.1244%\n",
      "\n",
      "Epoch [180/200]\n",
      "Train Loss: 2.2164 | Regression Loss: 0.4996 | Embedding Loss: 1.7169\n",
      "Valid MAE: 1.3481 | Valid Accuracy: 73.3645% | Valid Recall: 74.1294%\n",
      "\n",
      "Epoch [181/200]\n",
      "Train Loss: 2.2320 | Regression Loss: 0.5321 | Embedding Loss: 1.6999\n",
      "Valid MAE: 1.3490 | Valid Accuracy: 71.0280% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [182/200]\n",
      "Train Loss: 2.2132 | Regression Loss: 0.5189 | Embedding Loss: 1.6944\n",
      "Valid MAE: 1.3402 | Valid Accuracy: 73.8318% | Valid Recall: 74.6269%\n",
      "\n",
      "Epoch [183/200]\n",
      "Train Loss: 2.2676 | Regression Loss: 0.5529 | Embedding Loss: 1.7148\n",
      "Valid MAE: 1.3476 | Valid Accuracy: 69.6262% | Valid Recall: 70.1493%\n",
      "\n",
      "Epoch [184/200]\n",
      "Train Loss: 2.2330 | Regression Loss: 0.5248 | Embedding Loss: 1.7083\n",
      "Valid MAE: 1.3401 | Valid Accuracy: 73.3645% | Valid Recall: 74.1294%\n",
      "\n",
      "Epoch [185/200]\n",
      "Train Loss: 2.2366 | Regression Loss: 0.5117 | Embedding Loss: 1.7249\n",
      "Valid MAE: 1.3443 | Valid Accuracy: 73.3645% | Valid Recall: 74.1294%\n",
      "\n",
      "Epoch [186/200]\n",
      "Train Loss: 2.1983 | Regression Loss: 0.5006 | Embedding Loss: 1.6977\n",
      "Valid MAE: 1.3487 | Valid Accuracy: 72.4299% | Valid Recall: 72.6368%\n",
      "\n",
      "Epoch [187/200]\n",
      "Train Loss: 2.2136 | Regression Loss: 0.5041 | Embedding Loss: 1.7095\n",
      "Valid MAE: 1.3440 | Valid Accuracy: 73.8318% | Valid Recall: 74.6269%\n",
      "\n",
      "Epoch [188/200]\n",
      "Train Loss: 2.2184 | Regression Loss: 0.5065 | Embedding Loss: 1.7119\n",
      "Valid MAE: 1.3381 | Valid Accuracy: 72.4299% | Valid Recall: 73.1343%\n",
      "\n",
      "Epoch [189/200]\n",
      "Train Loss: 2.1790 | Regression Loss: 0.4768 | Embedding Loss: 1.7022\n",
      "Valid MAE: 1.3369 | Valid Accuracy: 74.2991% | Valid Recall: 75.1244%\n",
      "\n",
      "Epoch [190/200]\n",
      "Train Loss: 2.2387 | Regression Loss: 0.5191 | Embedding Loss: 1.7196\n",
      "Valid MAE: 1.3415 | Valid Accuracy: 73.3645% | Valid Recall: 74.1294%\n",
      "\n",
      "Epoch [191/200]\n",
      "Train Loss: 2.1924 | Regression Loss: 0.4858 | Embedding Loss: 1.7067\n",
      "Valid MAE: 1.3406 | Valid Accuracy: 73.8318% | Valid Recall: 74.6269%\n",
      "\n",
      "Epoch [192/200]\n",
      "Train Loss: 2.1770 | Regression Loss: 0.4764 | Embedding Loss: 1.7006\n",
      "Valid MAE: 1.3366 | Valid Accuracy: 74.2991% | Valid Recall: 75.1244%\n",
      "\n",
      "Epoch [193/200]\n",
      "Train Loss: 2.2287 | Regression Loss: 0.5211 | Embedding Loss: 1.7076\n",
      "Valid MAE: 1.3445 | Valid Accuracy: 72.4299% | Valid Recall: 73.1343%\n",
      "\n",
      "Epoch [194/200]\n",
      "Train Loss: 2.2186 | Regression Loss: 0.5152 | Embedding Loss: 1.7034\n",
      "Valid MAE: 1.3480 | Valid Accuracy: 70.5607% | Valid Recall: 71.1443%\n",
      "\n",
      "Epoch [195/200]\n",
      "Train Loss: 2.1919 | Regression Loss: 0.4938 | Embedding Loss: 1.6981\n",
      "Valid MAE: 1.3499 | Valid Accuracy: 70.0935% | Valid Recall: 70.6468%\n",
      "\n",
      "Epoch [196/200]\n",
      "Train Loss: 2.2068 | Regression Loss: 0.5094 | Embedding Loss: 1.6974\n",
      "Valid MAE: 1.3376 | Valid Accuracy: 73.3645% | Valid Recall: 74.1294%\n",
      "\n",
      "Epoch [197/200]\n",
      "Train Loss: 2.2016 | Regression Loss: 0.5021 | Embedding Loss: 1.6994\n",
      "Valid MAE: 1.3474 | Valid Accuracy: 72.8972% | Valid Recall: 73.6318%\n",
      "\n",
      "Epoch [198/200]\n",
      "Train Loss: 2.1842 | Regression Loss: 0.4726 | Embedding Loss: 1.7116\n",
      "Valid MAE: 1.3452 | Valid Accuracy: 71.9626% | Valid Recall: 72.1393%\n",
      "\n",
      "Epoch [199/200]\n",
      "Train Loss: 2.1774 | Regression Loss: 0.4731 | Embedding Loss: 1.7044\n",
      "Valid MAE: 1.3398 | Valid Accuracy: 73.3645% | Valid Recall: 74.1294%\n",
      "\n",
      "Epoch [200/200]\n",
      "Train Loss: 2.2154 | Regression Loss: 0.5170 | Embedding Loss: 1.6984\n",
      "Valid MAE: 1.3466 | Valid Accuracy: 71.4953% | Valid Recall: 71.6418%\n",
      "\n",
      "Training Time: 15m48s\n",
      "Training Params: 198809\n"
     ]
    }
   ],
   "source": [
    "train_losses, valid_losses = all_in_one_train(trainprocess, allmodules)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "6322ee5b-0442-47f3-a910-2533fee34036",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'Loss vs. No. of epochs')"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(train_losses, label='Train Loss')\n",
    "plt.plot(valid_losses, label='Validation Loss')\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('Loss')\n",
    "plt.legend()\n",
    "plt.grid(True)\n",
    "plt.title('Loss vs. No. of epochs')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "4ecbf0dc-8fad-490b-bd2a-5b0c265e91ab",
   "metadata": {},
   "outputs": [],
   "source": [
    "save_results(train_losses, valid_losses, f'/scratch/rrm9598/Multimodal-Sentiment-Analysis/results/{FILE_NAME}.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7c091121-ff36-40fc-ab64-a58a4f524e8b",
   "metadata": {},
   "source": [
    "## Testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "d01b591b-614f-433c-9ae7-10ac18a850fe",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test MAE: 1.3424 | Test Accuracy: 73.7609% | Test Recall: 74.2378%\n",
      "Inference Time: 833ms\n",
      "Inference Params: 198809\n"
     ]
    }
   ],
   "source": [
    "model = torch.load(f'/scratch/rrm9598/Multimodal-Sentiment-Analysis/models/{FILE_NAME}.pt').to(device)\n",
    "\n",
    "test(model, testdata, 'mosi', no_robust=True)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.11.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}