notebooks/Late_Fusion-MOSI.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "067cb8f5-a1c8-40d0-8524-0b28f8e4b586",
   "metadata": {},
   "source": [
    "# Late Fusion Model - CMU-MOSI Dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dc5714ca-83db-40ac-aef0-ee05dfddd1a3",
   "metadata": {},
   "source": [
    "In this notebook, we implement the Late Fusion Model, and test it on CMU-MOSI dataset."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fc015c8e-f315-40d9-823c-6fc3626a4510",
   "metadata": {},
   "source": [
    "## Importing Libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "d4a4adae-a451-493d-9ddb-0efb28537923",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import sys\n",
    "import torch\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": "52a35966-1fba-4107-a1fc-1137f26224bf",
   "metadata": {},
   "source": [
    "## Importing Helper Functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "b8d45f84-62c1-4315-9010-45d345dda90d",
   "metadata": {},
   "outputs": [],
   "source": [
    "from helpers.data.get_data import get_dataloader\n",
    "from helpers.unimodals.common_models import GRU, MLP\n",
    "from helpers.training.supervised_learning import train, test\n",
    "from helpers.fusions.common_fusions import Concat\n",
    "from helpers.utils.common_utils import get_default_device, save_results"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8dc9b148-2e2f-4cfa-a7d4-d53b639f5b18",
   "metadata": {},
   "source": [
    "## Loading the Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "7f193c63-0bfb-4248-a2a8-e81ca20a7c14",
   "metadata": {},
   "outputs": [],
   "source": [
    "traindata, validdata, testdata = get_dataloader(\n",
    "    '/scratch/rrm9598/data/mosi_raw.pkl', robust_test=False, data_type='mosi')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "75b021aa-09ec-4739-92d3-028a0ed54287",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "cuda\n"
     ]
    }
   ],
   "source": [
    "device = get_default_device()\n",
    "print(device)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a1b80fea-4d34-4db8-a1ed-9ca4e853bbd4",
   "metadata": {},
   "source": [
    "## Model Architecture"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "59286eb3-949a-4afb-b4f4-0b115b885f92",
   "metadata": {},
   "outputs": [],
   "source": [
    "encoders = [GRU(35, 70, dropout=True, has_padding=True, batch_first=True).to(device),\n",
    "            GRU(74, 200, dropout=True, has_padding=True, batch_first=True).to(device),\n",
    "            GRU(300, 600, dropout=True, has_padding=True, batch_first=True).to(device)]\n",
    "head = MLP(870, 870, 1).to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "63408731-2f89-4f01-b1d4-c169817fd8a8",
   "metadata": {},
   "outputs": [],
   "source": [
    "fusion = Concat().to(device)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4575e9d5-572c-4810-ba00-030f8ded6f05",
   "metadata": {},
   "source": [
    "## Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "f0b5dfe9-6677-4494-a013-7343ed43b731",
   "metadata": {},
   "outputs": [],
   "source": [
    "FILE_NAME = 'mosi_lf_r0'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "d9a645da-32a4-419a-9d64-2f499da509ac",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch [1/100]\n",
      "Train Loss: 1.1627\n",
      "Valid Loss: 1.1841\n",
      "Saving Best\n",
      "\n",
      "Epoch [2/100]\n",
      "Train Loss: 0.9498\n",
      "Valid Loss: 1.1025\n",
      "Saving Best\n",
      "\n",
      "Epoch [3/100]\n",
      "Train Loss: 0.8599\n",
      "Valid Loss: 1.0336\n",
      "Saving Best\n",
      "\n",
      "Epoch [4/100]\n",
      "Train Loss: 0.8178\n",
      "Valid Loss: 0.9631\n",
      "Saving Best\n",
      "\n",
      "Epoch [5/100]\n",
      "Train Loss: 0.7314\n",
      "Valid Loss: 0.9184\n",
      "Saving Best\n",
      "\n",
      "Epoch [6/100]\n",
      "Train Loss: 0.6660\n",
      "Valid Loss: 0.9861\n",
      "\n",
      "Epoch [7/100]\n",
      "Train Loss: 0.6252\n",
      "Valid Loss: 0.9293\n",
      "\n",
      "Epoch [8/100]\n",
      "Train Loss: 0.5926\n",
      "Valid Loss: 0.9898\n",
      "\n",
      "Epoch [9/100]\n",
      "Train Loss: 0.5541\n",
      "Valid Loss: 0.9332\n",
      "\n",
      "Epoch [10/100]\n",
      "Train Loss: 0.4970\n",
      "Valid Loss: 0.9177\n",
      "Saving Best\n",
      "\n",
      "Epoch [11/100]\n",
      "Train Loss: 0.4794\n",
      "Valid Loss: 0.9582\n",
      "\n",
      "Epoch [12/100]\n",
      "Train Loss: 0.4367\n",
      "Valid Loss: 0.9904\n",
      "\n",
      "Epoch [13/100]\n",
      "Train Loss: 0.4371\n",
      "Valid Loss: 0.9196\n",
      "\n",
      "Epoch [14/100]\n",
      "Train Loss: 0.3732\n",
      "Valid Loss: 0.9802\n",
      "\n",
      "Epoch [15/100]\n",
      "Train Loss: 0.3579\n",
      "Valid Loss: 0.9451\n",
      "\n",
      "Epoch [16/100]\n",
      "Train Loss: 0.3334\n",
      "Valid Loss: 0.9958\n",
      "\n",
      "Epoch [17/100]\n",
      "Train Loss: 0.3216\n",
      "Valid Loss: 0.9878\n",
      "\n",
      "Epoch [18/100]\n",
      "Train Loss: 0.2965\n",
      "Valid Loss: 0.9603\n",
      "\n",
      "Epoch [19/100]\n",
      "Train Loss: 0.3087\n",
      "Valid Loss: 0.9622\n",
      "\n",
      "Epoch [20/100]\n",
      "Train Loss: 0.2941\n",
      "Valid Loss: 0.9817\n",
      "\n",
      "Epoch [21/100]\n",
      "Train Loss: 0.2616\n",
      "Valid Loss: 0.9762\n",
      "\n",
      "Epoch [22/100]\n",
      "Train Loss: 0.2505\n",
      "Valid Loss: 0.9750\n",
      "\n",
      "Epoch [23/100]\n",
      "Train Loss: 0.2264\n",
      "Valid Loss: 0.9859\n",
      "\n",
      "Epoch [24/100]\n",
      "Train Loss: 0.2466\n",
      "Valid Loss: 0.9356\n",
      "\n",
      "Epoch [25/100]\n",
      "Train Loss: 0.2171\n",
      "Valid Loss: 0.9406\n",
      "\n",
      "Epoch [26/100]\n",
      "Train Loss: 0.2190\n",
      "Valid Loss: 0.9461\n",
      "\n",
      "Epoch [27/100]\n",
      "Train Loss: 0.2092\n",
      "Valid Loss: 0.9442\n",
      "\n",
      "Epoch [28/100]\n",
      "Train Loss: 0.2262\n",
      "Valid Loss: 0.9335\n",
      "\n",
      "Epoch [29/100]\n",
      "Train Loss: 0.1956\n",
      "Valid Loss: 0.9627\n",
      "\n",
      "Epoch [30/100]\n",
      "Train Loss: 0.1972\n",
      "Valid Loss: 0.9417\n",
      "\n",
      "Epoch [31/100]\n",
      "Train Loss: 0.2027\n",
      "Valid Loss: 0.9343\n",
      "\n",
      "Epoch [32/100]\n",
      "Train Loss: 0.1849\n",
      "Valid Loss: 0.9357\n",
      "\n",
      "Epoch [33/100]\n",
      "Train Loss: 0.1817\n",
      "Valid Loss: 0.9333\n",
      "\n",
      "Epoch [34/100]\n",
      "Train Loss: 0.1899\n",
      "Valid Loss: 0.9456\n",
      "\n",
      "Epoch [35/100]\n",
      "Train Loss: 0.1776\n",
      "Valid Loss: 0.9531\n",
      "\n",
      "Epoch [36/100]\n",
      "Train Loss: 0.1826\n",
      "Valid Loss: 0.9400\n",
      "\n",
      "Epoch [37/100]\n",
      "Train Loss: 0.1689\n",
      "Valid Loss: 0.9496\n",
      "\n",
      "Epoch [38/100]\n",
      "Train Loss: 0.1629\n",
      "Valid Loss: 0.9363\n",
      "\n",
      "Epoch [39/100]\n",
      "Train Loss: 0.1840\n",
      "Valid Loss: 0.9434\n",
      "\n",
      "Epoch [40/100]\n",
      "Train Loss: 0.1681\n",
      "Valid Loss: 0.9270\n",
      "\n",
      "Epoch [41/100]\n",
      "Train Loss: 0.1686\n",
      "Valid Loss: 0.9480\n",
      "\n",
      "Epoch [42/100]\n",
      "Train Loss: 0.1724\n",
      "Valid Loss: 0.9451\n",
      "\n",
      "Epoch [43/100]\n",
      "Train Loss: 0.1598\n",
      "Valid Loss: 0.9552\n",
      "\n",
      "Epoch [44/100]\n",
      "Train Loss: 0.1545\n",
      "Valid Loss: 0.9592\n",
      "\n",
      "Epoch [45/100]\n",
      "Train Loss: 0.1488\n",
      "Valid Loss: 0.9441\n",
      "\n",
      "Epoch [46/100]\n",
      "Train Loss: 0.1523\n",
      "Valid Loss: 0.9318\n",
      "\n",
      "Epoch [47/100]\n",
      "Train Loss: 0.1573\n",
      "Valid Loss: 0.9547\n",
      "\n",
      "Epoch [48/100]\n",
      "Train Loss: 0.1479\n",
      "Valid Loss: 0.9285\n",
      "\n",
      "Epoch [49/100]\n",
      "Train Loss: 0.1469\n",
      "Valid Loss: 0.9283\n",
      "\n",
      "Epoch [50/100]\n",
      "Train Loss: 0.1471\n",
      "Valid Loss: 0.9281\n",
      "\n",
      "Epoch [51/100]\n",
      "Train Loss: 0.1336\n",
      "Valid Loss: 0.9370\n",
      "\n",
      "Epoch [52/100]\n",
      "Train Loss: 0.1367\n",
      "Valid Loss: 0.9339\n",
      "\n",
      "Epoch [53/100]\n",
      "Train Loss: 0.1468\n",
      "Valid Loss: 0.9377\n",
      "\n",
      "Epoch [54/100]\n",
      "Train Loss: 0.1343\n",
      "Valid Loss: 0.9201\n",
      "\n",
      "Epoch [55/100]\n",
      "Train Loss: 0.1365\n",
      "Valid Loss: 0.9389\n",
      "\n",
      "Epoch [56/100]\n",
      "Train Loss: 0.1305\n",
      "Valid Loss: 0.9342\n",
      "\n",
      "Epoch [57/100]\n",
      "Train Loss: 0.1457\n",
      "Valid Loss: 0.9483\n",
      "\n",
      "Epoch [58/100]\n",
      "Train Loss: 0.1299\n",
      "Valid Loss: 0.9372\n",
      "\n",
      "Epoch [59/100]\n",
      "Train Loss: 0.1241\n",
      "Valid Loss: 0.9279\n",
      "\n",
      "Epoch [60/100]\n",
      "Train Loss: 0.1303\n",
      "Valid Loss: 0.9396\n",
      "\n",
      "Epoch [61/100]\n",
      "Train Loss: 0.1265\n",
      "Valid Loss: 0.9289\n",
      "\n",
      "Epoch [62/100]\n",
      "Train Loss: 0.1257\n",
      "Valid Loss: 0.9230\n",
      "\n",
      "Epoch [63/100]\n",
      "Train Loss: 0.1331\n",
      "Valid Loss: 0.9437\n",
      "\n",
      "Epoch [64/100]\n",
      "Train Loss: 0.1239\n",
      "Valid Loss: 0.9707\n",
      "\n",
      "Epoch [65/100]\n",
      "Train Loss: 0.1314\n",
      "Valid Loss: 0.9289\n",
      "\n",
      "Epoch [66/100]\n",
      "Train Loss: 0.1230\n",
      "Valid Loss: 0.9376\n",
      "\n",
      "Epoch [67/100]\n",
      "Train Loss: 0.1185\n",
      "Valid Loss: 0.9508\n",
      "\n",
      "Epoch [68/100]\n",
      "Train Loss: 0.1337\n",
      "Valid Loss: 0.9428\n",
      "\n",
      "Epoch [69/100]\n",
      "Train Loss: 0.1234\n",
      "Valid Loss: 0.9444\n",
      "\n",
      "Epoch [70/100]\n",
      "Train Loss: 0.1156\n",
      "Valid Loss: 0.9485\n",
      "\n",
      "Epoch [71/100]\n",
      "Train Loss: 0.1320\n",
      "Valid Loss: 0.9430\n",
      "\n",
      "Epoch [72/100]\n",
      "Train Loss: 0.1201\n",
      "Valid Loss: 0.9426\n",
      "\n",
      "Epoch [73/100]\n",
      "Train Loss: 0.1266\n",
      "Valid Loss: 0.9451\n",
      "\n",
      "Epoch [74/100]\n",
      "Train Loss: 0.1311\n",
      "Valid Loss: 0.9320\n",
      "\n",
      "Epoch [75/100]\n",
      "Train Loss: 0.1128\n",
      "Valid Loss: 0.9336\n",
      "\n",
      "Epoch [76/100]\n",
      "Train Loss: 0.1215\n",
      "Valid Loss: 0.9515\n",
      "\n",
      "Epoch [77/100]\n",
      "Train Loss: 0.1180\n",
      "Valid Loss: 0.9247\n",
      "\n",
      "Epoch [78/100]\n",
      "Train Loss: 0.1131\n",
      "Valid Loss: 0.9292\n",
      "\n",
      "Epoch [79/100]\n",
      "Train Loss: 0.1214\n",
      "Valid Loss: 0.9377\n",
      "\n",
      "Epoch [80/100]\n",
      "Train Loss: 0.1173\n",
      "Valid Loss: 0.9344\n",
      "\n",
      "Epoch [81/100]\n",
      "Train Loss: 0.1135\n",
      "Valid Loss: 0.9269\n",
      "\n",
      "Epoch [82/100]\n",
      "Train Loss: 0.1275\n",
      "Valid Loss: 0.9277\n",
      "\n",
      "Epoch [83/100]\n",
      "Train Loss: 0.1061\n",
      "Valid Loss: 0.9210\n",
      "\n",
      "Epoch [84/100]\n",
      "Train Loss: 0.1175\n",
      "Valid Loss: 0.9377\n",
      "\n",
      "Epoch [85/100]\n",
      "Train Loss: 0.1142\n",
      "Valid Loss: 0.9228\n",
      "\n",
      "Epoch [86/100]\n",
      "Train Loss: 0.1126\n",
      "Valid Loss: 0.9442\n",
      "\n",
      "Epoch [87/100]\n",
      "Train Loss: 0.1123\n",
      "Valid Loss: 0.9354\n",
      "\n",
      "Epoch [88/100]\n",
      "Train Loss: 0.1119\n",
      "Valid Loss: 0.9333\n",
      "\n",
      "Epoch [89/100]\n",
      "Train Loss: 0.1126\n",
      "Valid Loss: 0.9298\n",
      "\n",
      "Epoch [90/100]\n",
      "Train Loss: 0.1093\n",
      "Valid Loss: 0.9324\n",
      "\n",
      "Epoch [91/100]\n",
      "Train Loss: 0.1050\n",
      "Valid Loss: 0.9322\n",
      "\n",
      "Epoch [92/100]\n",
      "Train Loss: 0.1148\n",
      "Valid Loss: 0.9394\n",
      "\n",
      "Epoch [93/100]\n",
      "Train Loss: 0.1081\n",
      "Valid Loss: 0.9198\n",
      "\n",
      "Epoch [94/100]\n",
      "Train Loss: 0.0977\n",
      "Valid Loss: 0.9338\n",
      "\n",
      "Epoch [95/100]\n",
      "Train Loss: 0.1139\n",
      "Valid Loss: 0.9419\n",
      "\n",
      "Epoch [96/100]\n",
      "Train Loss: 0.1059\n",
      "Valid Loss: 0.9468\n",
      "\n",
      "Epoch [97/100]\n",
      "Train Loss: 0.1017\n",
      "Valid Loss: 0.9324\n",
      "\n",
      "Epoch [98/100]\n",
      "Train Loss: 0.1000\n",
      "Valid Loss: 0.9355\n",
      "\n",
      "Epoch [99/100]\n",
      "Train Loss: 0.1118\n",
      "Valid Loss: 0.9228\n",
      "\n",
      "Epoch [100/100]\n",
      "Train Loss: 0.0969\n",
      "Valid Loss: 0.9282\n",
      "\n",
      "Training Time: 2m53s\n",
      "Training Peak Mem: 1338.75390625\n",
      "Training Params: 2570311\n"
     ]
    }
   ],
   "source": [
    "train_losses, valid_losses = train(\n",
    "    encoders, fusion, head, traindata, validdata, 100, task=\"regression\", optimtype=torch.optim.AdamW,\n",
    "    early_stop=False, is_packed=True, lr=1e-3, save=f'/scratch/rrm9598/Multimodal-Sentiment-Analysis/models/{FILE_NAME}.pt', \n",
    "    weight_decay=0.01, objective=torch.nn.L1Loss()\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "b0e7c33a-a68d-4292-a2b2-14e8a2e53500",
   "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": "8bbd60df-8764-4fcd-a545-90860e5f8a2c",
   "metadata": {},
   "outputs": [],
   "source": [
    "save_results(train_losses, valid_losses, f'/scratch/rrm9598/Multimodal-Sentiment-Analysis/results/{FILE_NAME}.pkl')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a2a0ca4a-2c05-4ba2-a6aa-1db0e341598b",
   "metadata": {},
   "source": [
    "## Testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "d4352442-0d78-4195-bd04-2332af36a7f7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Recall: 70.5793% | Total Accuracy: 70.2624%\n",
      "Inference Time: 572ms\n",
      "Inference Params: 2570311\n"
     ]
    }
   ],
   "source": [
    "model = torch.load(f'/scratch/rrm9598/Multimodal-Sentiment-Analysis/models/{FILE_NAME}.pt').to(device)\n",
    "\n",
    "test(model, test_dataloaders_all=testdata, dataset='mosi', is_packed=True,\n",
    "     criterion=torch.nn.L1Loss(), task=\"posneg-classification\", no_robust=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "dd7d4419-cc6b-4022-9b93-49772c55a675",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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
}