main.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "metadata": {}
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'\\nAuthor: Philip Paterson\\nReferenced PyTorch documentation\\n'"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from vit_pytorch.vit import ViT\n",
    "from vit_pytorch.extractor import Extractor\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.image as mpimg\n",
    "from MyTransformer import MyTransformer\n",
    "\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "from torch import Tensor\n",
    "from torchvision import transforms\n",
    "from torch.nn import Transformer\n",
    "import math\n",
    "from torchtext.data.utils import get_tokenizer\n",
    "from collections import Counter\n",
    "from torchtext.vocab import Vocab\n",
    "import io\n",
    "from torch.nn.utils.rnn import pad_sequence\n",
    "from MyTransformer import MyTransformer\n",
    "from torch.utils.data import DataLoader, Dataset\n",
    "import numpy as np\n",
    "from util.bleu import get_bleu\n",
    "import os\n",
    "import json\n",
    "from PIL import Image\n",
    "from torchvision.transforms import v2\n",
    "\"\"\"\n",
    "Author: Philip Paterson\n",
    "Referenced PyTorch documentation\n",
    "\"\"\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "metadata": {}
   },
   "outputs": [],
   "source": [
    "class ImageCaptioner(nn.Module):\n",
    "    def __init__(self, vocab_size):\n",
    "\n",
    "        NUM_ENCODER_LAYERS = 3\n",
    "        NUM_DECODER_LAYERS = 3\n",
    "        EMB_SIZE = 512 # Embedding dimension: 512\n",
    "        NHEAD = 16 # No. of Attention head: 8\n",
    "        FFN_HID_DIM = 512 # Feedforward dimension: 512\n",
    "        DROPOUT = 0.1\n",
    "\n",
    "        super(ImageCaptioner, self).__init__()\n",
    "\n",
    "        self.v = ViT(\n",
    "            image_size=256, #Size of the image\n",
    "            patch_size=16, # number of patches\n",
    "            num_classes=1000, #number of classes,\n",
    "            dim=EMB_SIZE, # Embedding dimension\n",
    "            depth=NUM_ENCODER_LAYERS, # No of encoder layers\n",
    "            heads=NHEAD, #No of MLP heads\n",
    "            mlp_dim=FFN_HID_DIM, # Dimension of Feedforwards\n",
    "            dropout=0.1,\n",
    "            emb_dropout=0.1\n",
    "        )\n",
    "\n",
    "        self.vision_encoder = Extractor(self.v)\n",
    "        \n",
    "        # self.transformer = Transformer(\n",
    "        #     d_model=EMB_SIZE,\n",
    "        #     nhead=NHEAD,\n",
    "        #     num_encoder_layers=0,\n",
    "        #     num_decoder_layers=NUM_DECODER_LAYERS,\n",
    "        #     dim_feedforward=FFN_HID_DIM,\n",
    "        #     dropout=DROPOUT\n",
    "        # )\n",
    "\n",
    "        self.decoder = nn.TransformerDecoder(\n",
    "            nn.TransformerDecoderLayer(\n",
    "                d_model=EMB_SIZE, \n",
    "                nhead=NHEAD,\n",
    "                dim_feedforward=FFN_HID_DIM,\n",
    "                dropout=DROPOUT),\n",
    "                num_layers=NUM_DECODER_LAYERS\n",
    "        )\n",
    "\n",
    "        self.generator = nn.Linear(EMB_SIZE, vocab_size)\n",
    "        self.tgt_tok_emb = TokenEmbedding(vocab_size, EMB_SIZE)\n",
    "        self.positional_encoding = PositionalEncoding(\n",
    "            EMB_SIZE, dropout=DROPOUT)\n",
    "        \n",
    "    @property\n",
    "    def device(self):\n",
    "        return next(self.parameters()).device\n",
    "\n",
    "    def generate_square_subsequent_mask(self, sz):\n",
    "        mask = (torch.triu(torch.ones((sz, sz), device=self.device)) == 1).transpose(0, 1)\n",
    "        mask = mask.float().masked_fill(mask == 0, float('-inf')).masked_fill(mask == 1, float(0.0))\n",
    "        return mask\n",
    "\n",
    "    def create_mask(self, src, tgt, PAD_IDX):\n",
    "        # src_seq_len = src.shape[0]\n",
    "        tgt_seq_len = tgt.shape[0]\n",
    "\n",
    "        tgt_mask = self.generate_square_subsequent_mask(tgt_seq_len)\n",
    "        # src_mask = torch.zeros((src_seq_len, src_seq_len), device=self.device).type(torch.bool)\n",
    "\n",
    "        # src_padding_mask = (src == PAD_IDX).transpose(0, 1)\n",
    "        tgt_padding_mask = (tgt == PAD_IDX).transpose(0, 1)\n",
    "        # return src_mask, tgt_mask, src_padding_mask, tgt_padding_mask\n",
    "        return tgt_mask, tgt_padding_mask\n",
    "    \n",
    "    # def decode(self, tgt: Tensor, memory: Tensor, tgt_mask: Tensor):\n",
    "    #     return self.transformer.decoder(self.positional_encoding(\n",
    "    #                       self.tgt_tok_emb(tgt)), memory,\n",
    "    #                       tgt_mask)\n",
    "\n",
    "    def decode(self, tgt: Tensor, memory: Tensor, tgt_mask: Tensor, tgt_padding_mask):\n",
    "        return self.decoder(\n",
    "            self.positional_encoding(self.tgt_tok_emb(tgt)), \n",
    "            memory, \n",
    "            tgt_mask=tgt_mask, \n",
    "            tgt_key_padding_mask=tgt_padding_mask\n",
    "        )\n",
    "\n",
    "    def forward(self, src, trg, pad_idx : torch.Tensor):\n",
    "        \"\"\" Forward method for transformer\n",
    "        \n",
    "        :param src: batch of images of shape \n",
    "            (Batch,Colorchannel,Height,Width)\n",
    "        :param trg: Target input sequence, right-shifted by one\n",
    "            Target batch matrix is (N_T - 1) x Batch Size\n",
    "        :param pad_idx: padding index\n",
    "        \"\"\"\n",
    "        # Here, src is the batch of images of shape \n",
    "        # (Batch,Colorchannel,Height,Width)\n",
    "        logits, memory = self.vision_encoder(src)\n",
    "\n",
    "        tgt_mask, tgt_padding_mask = self.create_mask(src, trg, pad_idx)\n",
    "\n",
    "        outs = self.decode(trg, memory.transpose(0, 1), tgt_mask, tgt_padding_mask)\n",
    "\n",
    "        return self.generator(outs)\n",
    "\n",
    "\n",
    "# helper Module that adds positional encoding to the token embedding to introduce a notion of word order.\n",
    "class PositionalEncoding(nn.Module):\n",
    "    def __init__(self,\n",
    "                 emb_size: int,\n",
    "                 dropout: float,\n",
    "                 maxlen: int = 5000):\n",
    "        super(PositionalEncoding, self).__init__()\n",
    "        den = torch.exp(- torch.arange(0, emb_size, 2)* math.log(10000) / emb_size)\n",
    "        pos = torch.arange(0, maxlen).reshape(maxlen, 1)\n",
    "        pos_embedding = torch.zeros((maxlen, emb_size))\n",
    "        pos_embedding[:, 0::2] = torch.sin(pos * den)\n",
    "        pos_embedding[:, 1::2] = torch.cos(pos * den)\n",
    "        pos_embedding = pos_embedding.unsqueeze(-2)\n",
    "\n",
    "        self.dropout = nn.Dropout(dropout)\n",
    "        self.register_buffer('pos_embedding', pos_embedding)\n",
    "\n",
    "    def forward(self, token_embedding: Tensor):\n",
    "        return self.dropout(token_embedding + self.pos_embedding[:token_embedding.size(0), :])\n",
    "\n",
    "# helper Module to convert tensor of input indices into corresponding tensor of token embeddings\n",
    "class TokenEmbedding(nn.Module):\n",
    "    def __init__(self, vocab_size: int, emb_size):\n",
    "        super(TokenEmbedding, self).__init__()\n",
    "        self.embedding = nn.Embedding(vocab_size, emb_size)\n",
    "        self.emb_size = emb_size\n",
    "\n",
    "    def forward(self, tokens: Tensor):\n",
    "        return self.embedding(tokens.long()) * math.sqrt(self.emb_size)\n",
    "\n",
    "class DatasetFlickr8K(Dataset):\n",
    "    def __init__(self, data : dict, vocab : dict, data_dir : str, all_captions_used : bool = True):\n",
    "        CAPTIONS_PER_IMG = 5\n",
    "        \n",
    "        self.data = data # The json dataset dict\n",
    "        self.vocab = vocab\n",
    "        self.data_dir = data_dir\n",
    "\n",
    "        self.img_transform = transforms.Compose([\n",
    "            transforms.Resize((256, 256)),\n",
    "            transforms.ToTensor(),\n",
    "            transforms.Normalize(0.5, 0.5),\n",
    "            v2.RandomHorizontalFlip(p=0.5)\n",
    "        ])\n",
    "\n",
    "        self.captions_per_img = CAPTIONS_PER_IMG\n",
    "\n",
    "        self.all_captions_used = all_captions_used\n",
    "    \n",
    "    def __len__(self) -> int:\n",
    "        if not self.all_captions_used:\n",
    "            return len(self.data['images'])\n",
    "\n",
    "        length = self.captions_per_img*len(self.data['images'])\n",
    "        return length\n",
    "\n",
    "    def __getitem__(self, idx) -> tuple:\n",
    "        if idx > len(self):\n",
    "            # Referenced https://docs.python.org/3/library/exceptions.html\n",
    "            # for raising exceptions\n",
    "            raise IndexError(\"ERROR: Invalid index; out of bounds\")\n",
    "\n",
    "        # index = idx\n",
    "\n",
    "        # Get the indices (ASSUMING ALL IMAGES HAVE EXACTLY 5 CAPTIONS!)\n",
    "        if not self.all_captions_used: # Could use for training data\n",
    "            img_idx = idx\n",
    "            img_data = self.data['images'][img_idx]\n",
    "            caption = max(img_data['sentences'], key=len) # Get the longest caption\n",
    "        else: # If not training data \n",
    "            img_idx = idx // self.captions_per_img\n",
    "            caption_idx = idx % self.captions_per_img\n",
    "\n",
    "            img_data = self.data['images'][img_idx]\n",
    "            caption = img_data['sentences'][caption_idx]\n",
    "\n",
    "        # Get the image\n",
    "        img_name = self.data_dir + 'Images/' + img_data['filename']\n",
    "        with Image.open(img_name) as raw_img:\n",
    "            transformed_img = self.img_transform(raw_img)\n",
    "\n",
    "        # Get the caption tokens\n",
    "        tokens = caption['tokens']\n",
    "\n",
    "        # Encode caption\n",
    "        encoded_caption = encode_tokens(tokens=tokens, vocab=self.vocab)\n",
    "\n",
    "        return transformed_img, torch.tensor(encoded_caption)\n",
    "    \n",
    "    def test_captions(self):\n",
    "        for img_data in self.data['images']:\n",
    "            if len(img_data['sentences']) != 5:\n",
    "                print(\"ERROR: NON-FIVE CAPTION COUNT\")\n",
    "                return\n",
    "        print(\"All 5!\")    \n",
    "\n",
    "# Functions on Getting Data\n",
    "def build_vocab(filepath, tokenizer):\n",
    "    my_counter = Counter()\n",
    "    with io.open(filepath, encoding=\"utf8\") as filehandle:\n",
    "        for str in filehandle:\n",
    "            my_counter.update(tokenizer(str))\n",
    "    return Vocab(my_counter, specials=['<unk>', '<pad>', '<bos>', '<eos>'])\n",
    "\n",
    "def encode_tokens(tokens : list, vocab : dict) -> list:\n",
    "    '''\n",
    "    Encodes the given token sequence\n",
    "    :param words: list of words\n",
    "    :param vocab: vocab dictionary \n",
    "    '''\n",
    "    tok_arr = [vocab['<bos>']]\n",
    "    for token in tokens:\n",
    "        if token in vocab:\n",
    "            tok_arr.append(vocab[token])\n",
    "        else:\n",
    "            tok_arr.append(vocab['<unk>'])\n",
    "    tok_arr.append(vocab['<eos>'])\n",
    "\n",
    "    return tok_arr\n",
    "\n",
    "def decode_tokens(tokenized_sentence, vocab):\n",
    "    '''\n",
    "    Takes a tokenized sentence and converts it to a sentence of words.\n",
    "    '''\n",
    "    \n",
    "    # Referenced https://pieriantraining.com/reversing-keys-and-values-in-a-python-dictionary/ \n",
    "    # for reversing a dictionary\n",
    "    reversed_vocab = {value: key for key, value in vocab.items()}\n",
    "\n",
    "    # Build the translated sentence\n",
    "    sentence = []\n",
    "    for token in tokenized_sentence:\n",
    "        word = reversed_vocab[token]\n",
    "        if '<' not in word:\n",
    "            sentence.append(word)\n",
    "\n",
    "    return sentence\n",
    "\n",
    "def tokenize_text(iterator, vocab, tokenizer) -> list:\n",
    "    '''\n",
    "    Tokenizes all sentences in the given iterator\n",
    "    '''\n",
    "    tokenized_text = []\n",
    "    for sentence in iterator:\n",
    "        tokenized_sentence = torch.Tensor(encode_tokens(sentence, vocab))\n",
    "        tokenized_text.append(tokenized_sentence)\n",
    "    return tokenized_text\n",
    "\n",
    "def create_batch(each_data_batch, PAD_IDX):\n",
    "    '''\n",
    "    Creates a batch of the\n",
    "    '''\n",
    "    img_batch, caption_batch = [], []\n",
    "    i = 0\n",
    "    for (img, caption) in each_data_batch:\n",
    "        img_batch.append(img)\n",
    "        caption_batch.append(caption)\n",
    "        i += 1\n",
    "\n",
    "    # Pad the sequences\n",
    "    caption_batch = pad_sequence(caption_batch, padding_value=PAD_IDX)\n",
    "\n",
    "    return torch.stack(img_batch, dim=0), caption_batch\n",
    "\n",
    "\n",
    "def get_dataloader(data : dict, vocab : dict, data_dir : str, batch_size : int, all_captions_used : bool) -> DataLoader:\n",
    "    '''\n",
    "    Helper function to create dataloader\n",
    "    '''\n",
    "    data = DatasetFlickr8K(\n",
    "        data=data,\n",
    "        vocab=vocab,\n",
    "        data_dir=data_dir,\n",
    "        all_captions_used=all_captions_used\n",
    "    )\n",
    "\n",
    "    # Return the dataloader\n",
    "    dataloader = DataLoader(\n",
    "        data,\n",
    "        batch_size=batch_size,\n",
    "        shuffle=True,\n",
    "        collate_fn=lambda batch: create_batch(batch, vocab['<pad>'])\n",
    "    )    \n",
    "\n",
    "    return dataloader\n",
    "\n",
    "def run_one_epoch(epoch_index : int, model, dataloader : DataLoader, loss_func, optimizer, pad_idx, split : str, device : str = 'cpu'):\n",
    "    '''\n",
    "    Performs one epoch of either training with backpropagation or validation or testing inference\n",
    "    '''\n",
    "    running_loss = 0.\n",
    "\n",
    "    for src, tgt in dataloader:\n",
    "        # Change the device and format\n",
    "        tgt = tgt.type(torch.LongTensor)\n",
    "        src = src.to(device=device)\n",
    "        tgt = tgt.to(device=device)\n",
    "        \n",
    "        # Zero the gradients for every batch\n",
    "        optimizer.zero_grad()\n",
    "\n",
    "        tgt_input = tgt[:-1]\n",
    "        tgt_out = tgt[1:]\n",
    "\n",
    "        # Make the predictions from the forward pass\n",
    "        logits = model(src=src, trg=tgt_input, pad_idx=torch.tensor(pad_idx, device=device))\n",
    "\n",
    "        # Compute the loss\n",
    "        loss = loss_func(logits.reshape(-1, logits.shape[-1]), tgt_out.reshape(-1))\n",
    "        \n",
    "        # If we're training the model\n",
    "        if split == 'TRAIN':\n",
    "            # Compute loss gradients using backward\n",
    "            loss.backward()\n",
    "\n",
    "            # Make adjustments to the learning weights\n",
    "            optimizer.step()\n",
    "\n",
    "        # Sum the losses and accuracies\n",
    "        running_loss += loss.item()\n",
    "\n",
    "    avg_loss = running_loss / len(dataloader)\n",
    "\n",
    "    return avg_loss\n",
    "\n",
    "def train(\n",
    "        model: nn.Module,\n",
    "        train_dataloader : DataLoader,\n",
    "        valid_dataloader : DataLoader,\n",
    "        test_dataloader : DataLoader,\n",
    "        loss_func, optimizer,\n",
    "        pad_idx,\n",
    "        tgt_vocab,\n",
    "        device = 'cpu'\n",
    "    ):\n",
    "    '''\n",
    "    The main train function\n",
    "    '''\n",
    "    EPOCHS = 20\n",
    "\n",
    "    \n",
    "    \n",
    "    # Initialize parameters\n",
    "    for p in model.parameters():\n",
    "        if p.dim() > 1:\n",
    "            nn.init.xavier_uniform_(p)\n",
    "\n",
    "    # Loop variables\n",
    "    avg_train_losses = []\n",
    "    avg_valid_losses = []\n",
    "    bleu_scores = []\n",
    "    best_bleu_score = 0\n",
    "\n",
    "    for epoch in range(EPOCHS):\n",
    "        print(f\"---Epoch {epoch + 1}---\")\n",
    "        \n",
    "        # Run the training\n",
    "        model.train(True)\n",
    "        \n",
    "        avg_train_loss = run_one_epoch(\n",
    "            epoch,\n",
    "            model,\n",
    "            train_dataloader,\n",
    "            loss_func,\n",
    "            optimizer,\n",
    "            pad_idx,\n",
    "            'TRAIN',\n",
    "            device\n",
    "        )\n",
    "                \n",
    "        # Evaluate the model off of the validation dataset\n",
    "        model.eval()\n",
    "        with torch.no_grad():\n",
    "            avg_valid_loss = run_one_epoch(\n",
    "                epoch,\n",
    "                model,\n",
    "                valid_dataloader,\n",
    "                loss_func,\n",
    "                optimizer,\n",
    "                pad_idx,\n",
    "                'VALID',\n",
    "                device\n",
    "            )\n",
    "\n",
    "        # Calculate the bleu score\n",
    "        with torch.no_grad():\n",
    "            bleu_score = calc_bleu_score(\n",
    "                model=model,\n",
    "                dataloader=test_dataloader,\n",
    "                device=device,\n",
    "                vocab=tgt_vocab,\n",
    "            )\n",
    "\n",
    "        # Save the best model parameters\n",
    "        if bleu_score > best_bleu_score:\n",
    "            best_bleu_score = bleu_score\n",
    "            torch.save(model.state_dict(), 'saved_model.pt')\n",
    "        \n",
    "        # Append the losses and bleu score\n",
    "        avg_train_losses.append(avg_train_loss)\n",
    "        avg_valid_losses.append(avg_valid_loss)\n",
    "        bleu_scores.append(bleu_score)\n",
    "\n",
    "        # Print the per epoch losses and accuracies\n",
    "        print(f\"Loss: | Train: {avg_train_loss:.5f} | Validation: {avg_valid_loss:.5f}\")\n",
    "        print(f\"BLEU Score: {bleu_score}\")\n",
    "    \n",
    "    stats = {\n",
    "        'training' : {\n",
    "            'losses' : avg_train_losses,\n",
    "        },\n",
    "        'validation' : {\n",
    "            'losses' : avg_valid_losses,\n",
    "        },\n",
    "        'testing' : {\n",
    "            'bleu' : bleu_scores\n",
    "        },\n",
    "        'epochs' : np.arange(1, EPOCHS+1).tolist()\n",
    "    }\n",
    "\n",
    "    return stats\n",
    "\n",
    "def test_captioner(data_dir : str, data_fname : str, vocab_path : str, model_param_path : str):\n",
    "    '''\n",
    "    TEST FUNCTION For testing the translator model.\n",
    "\n",
    "    Only works for testing translation from German to English.\n",
    "    '''\n",
    "\n",
    "    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
    "\n",
    "    \n",
    "    TOKENIZER_LANGUAGE = 'de_core_news_sm'\n",
    "\n",
    "    # Get the tokenizer\n",
    "    tokenizer = get_tokenizer('spacy', language=TOKENIZER_LANGUAGE)\n",
    "\n",
    "    # Get the vocabulary\n",
    "    vocab = torch.load(vocab_path, tokenizer).stoi\n",
    "    \n",
    "    # Get the data\n",
    "    data_json = json.load(open(data_dir + data_fname))\n",
    "\n",
    "    data = DatasetFlickr8K(\n",
    "        data=data_json,\n",
    "        vocab=vocab,\n",
    "        data_dir=data_dir,\n",
    "        all_captions_used=True\n",
    "    )\n",
    "    \n",
    "    dataloader = DataLoader(\n",
    "        data,\n",
    "        batch_size=1,\n",
    "        shuffle=False,\n",
    "        collate_fn=lambda batch: create_batch(batch, vocab['<pad>'])\n",
    "    )\n",
    "\n",
    "    model = ImageCaptioner(\n",
    "        vocab_size=len(vocab)\n",
    "    ).to(device)\n",
    "\n",
    "    model.load_state_dict(torch.load(model_param_path, map_location=device))\n",
    "\n",
    "    model.eval()\n",
    "\n",
    "    # Perform inference\n",
    "    with torch.no_grad():\n",
    "        bleu_score = calc_bleu_score(\n",
    "            model=model,\n",
    "            dataloader=dataloader,\n",
    "            device=device,\n",
    "            vocab=vocab\n",
    "        )\n",
    "    \n",
    "    print(f\"Testing BLEU Score: {bleu_score}\")\n",
    "    return bleu_score\n",
    "\n",
    "def calc_bleu_score(model : nn.Module, dataloader : DataLoader, device, vocab):\n",
    "    '''\n",
    "    Helper function to calculate the BLEU score\n",
    "    '''\n",
    "    \n",
    "    SENTENCE_GENERATION_MAX = 500\n",
    "    \n",
    "    pad_idx = vocab['<pad>']\n",
    "\n",
    "    running_bleu = 0.\n",
    "    sent_idx = 0 # Keeps track of how many sentences generated\n",
    "    for src, tgt in dataloader:\n",
    "        # Stops the generation of sentences\n",
    "        if sent_idx >= SENTENCE_GENERATION_MAX:\n",
    "            break\n",
    "\n",
    "        tgt = tgt.type(torch.LongTensor)\n",
    "        src = src.to(device=device)\n",
    "        tgt = tgt.to(device=device)\n",
    "\n",
    "        # Initialize the next token and predicted sentence\n",
    "        next_tok = vocab['<bos>']\n",
    "        pred_sent_beginning = [[vocab['<bos>']]]\n",
    "        pred_sent = torch.tensor(pred_sent_beginning).to(device=device) # Initialize the predicted sentence\n",
    "\n",
    "        # Iteratively test to build the prediction sentence; cap the max sentence length to 10 extra tokens\n",
    "        while len(pred_sent) <= len(tgt) + 10 and next_tok != vocab['<eos>']:\n",
    "            logits = model(src=src, trg=pred_sent, pad_idx=torch.tensor(pad_idx, device=device))\n",
    "            next_tok = torch.argmax(logits[-1][0], dim=-1)\n",
    "            pred_sent = torch.cat((pred_sent, next_tok.unsqueeze(0).unsqueeze(0)), dim=0)\n",
    "\n",
    "        # Convert the tokenized predicted and target sentences to list of words\n",
    "        trg_sent_list = decode_tokens(tgt.squeeze(1).tolist(), vocab)\n",
    "        pred_sent_list = decode_tokens(pred_sent.squeeze(1).tolist(), vocab)\n",
    "\n",
    "       \n",
    "\n",
    "        # Get the bleu scores\n",
    "        bleu = get_bleu(hypotheses=pred_sent_list, reference=trg_sent_list)\n",
    "\n",
    "        if sent_idx % 100 == 5:\n",
    "            # Debug Print Visualization of Sentences\n",
    "            print(f\"Intermediary bleu: {bleu}\")\n",
    "            print(\"Target sentence list   :\", trg_sent_list) # Visualizes sentences\n",
    "            print(\"Predicted sentence list:\", pred_sent_list) # Visualizes sentences\n",
    "\n",
    "        running_bleu += bleu\n",
    "        \n",
    "        sent_idx += 1\n",
    "    \n",
    "    average_total_bleu = running_bleu / SENTENCE_GENERATION_MAX # Calculate the average bleu score\n",
    "\n",
    "    return average_total_bleu\n",
    "\n",
    "\n",
    "def line_plot(xvals, xlab : str, ylab: str, yvals0, ylab0 : str, dir : str, yvals1 = None, ylab1 : str = None):\n",
    "    \"\"\"\n",
    "    line_plot Creates a line plot of the xvals vs yvals\n",
    "    \"\"\"\n",
    "    fig = plt.figure()\n",
    "    \n",
    "    if yvals1:\n",
    "        plt.plot(xvals, yvals0, label = ylab0)\n",
    "        plt.plot(xvals, yvals1, label = ylab1)\n",
    "        plt.legend()\n",
    "    else:\n",
    "        plt.plot(xvals, yvals0)\n",
    "    \n",
    "    plt.title(f\"{ylab} vs {xlab}\")\n",
    "    plt.xlabel(xlab)\n",
    "    plt.ylabel(ylab)\n",
    "    file_path = os.path.join(dir, f\"{ylab.replace(' ', '_')}_vs_{xlab.replace(' ', '_')}_plot.png\")\n",
    "    plt.savefig(file_path)\n",
    "\n",
    "def plot_performance(stats : dict):\n",
    "    '''\n",
    "    Plots the Programming Assignment 4 plots.\n",
    "    '''\n",
    "    \n",
    "    plot_dir = './plots'\n",
    "\n",
    "    # Training and validation loss over epochs plot \n",
    "    line_plot(\n",
    "        xvals = stats['epochs'],\n",
    "        xlab='Epochs',\n",
    "        ylab='Average Loss',\n",
    "        yvals0=stats['training']['losses'],\n",
    "        ylab0='Training Losses',\n",
    "        yvals1=stats['validation']['losses'],\n",
    "        ylab1='Validation Losses',\n",
    "        dir = plot_dir\n",
    "    )\n",
    "\n",
    "    # Testing BLEU score over epochs plot\n",
    "    line_plot(\n",
    "        xvals = stats['epochs'],\n",
    "        xlab='Epochs',\n",
    "        ylab='BLEU Score',\n",
    "        yvals0=stats['testing']['bleu'],\n",
    "        ylab0='BLEU Score for Test Dataset',\n",
    "        yvals1=None,\n",
    "        ylab1=None,\n",
    "        dir = plot_dir\n",
    "    )\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Main Function"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "metadata": {}
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/philip/anaconda3/lib/python3.11/site-packages/torch/cuda/__init__.py:141: UserWarning: CUDA initialization: CUDA unknown error - this may be due to an incorrectly set up environment, e.g. changing env variable CUDA_VISIBLE_DEVICES after program start. Setting the available devices to be zero. (Triggered internally at ../c10/cuda/CUDAFunctions.cpp:108.)\n",
      "  return torch._C._cuda_getDeviceCount() > 0\n"
     ]
    }
   ],
   "source": [
    "# Training JSON file (‘training data.json’) contains 6000 entities\n",
    "# validation JSON (‘val data.json’) contains 1000 entities\n",
    "# test JSON (‘test data.json’) contains 500 entities\n",
    "datapath = 'Final_Project_Spring_24/'\n",
    "training_data_fname = 'training_data.json'\n",
    "test_data_fname = 'test_data.json'\n",
    "val_data_fname = 'val_data.json'\n",
    "english_vocab_path = 'English_vocab.pth'\n",
    "\n",
    "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
    "BATCH_SIZE = 32"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "metadata": {}
   },
   "outputs": [],
   "source": [
    "# Load the data\n",
    "training_json = json.load(open(datapath + training_data_fname))\n",
    "test_json = json.load(open(datapath + test_data_fname))\n",
    "val_json = json.load(open(datapath + val_data_fname))\n",
    "\n",
    "# Get the tokenizer\n",
    "en_tokenizer = get_tokenizer('spacy', language='en_core_web_sm')\n",
    "\n",
    "# Get the vocabulary\n",
    "en_vocab = torch.load(english_vocab_path, en_tokenizer).stoi\n",
    "\n",
    "# Get the padding index\n",
    "pad_idx = en_vocab['<pad>']\n",
    "\n",
    "# Get the dataloaders for train, test, and validation\n",
    "train_dataloader = get_dataloader(\n",
    "    data=training_json,\n",
    "    vocab=en_vocab,\n",
    "    data_dir=datapath,\n",
    "    batch_size=BATCH_SIZE,\n",
    "    all_captions_used = False\n",
    ")\n",
    "\n",
    "test_dataloader = get_dataloader(\n",
    "    data=test_json,\n",
    "    vocab=en_vocab,\n",
    "    data_dir=datapath,\n",
    "    batch_size=1,\n",
    "    all_captions_used = True\n",
    ")\n",
    "\n",
    "valid_dataloader = get_dataloader(\n",
    "    data=val_json,\n",
    "    vocab=en_vocab,\n",
    "    data_dir=datapath,\n",
    "    batch_size=BATCH_SIZE,\n",
    "    all_captions_used = True\n",
    ")\n",
    "\n",
    "model = ImageCaptioner(vocab_size=len(en_vocab)).to(device)\n",
    "\n",
    "# Define the loss function and optimizer\n",
    "loss_fn = torch.nn.CrossEntropyLoss(ignore_index=pad_idx)\n",
    "optimizer = torch.optim.Adam(model.parameters(), lr=0.0001, betas=(0.9, 0.98), eps=1e-9)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "metadata": {}
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "---Epoch 1---\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/philip/anaconda3/lib/python3.11/site-packages/torch/nn/functional.py:5109: UserWarning: Support for mismatched key_padding_mask and attn_mask is deprecated. Use same type for both instead.\n",
      "  warnings.warn(\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['two', 'girls', 'are', 'posing', 'behind', 'a', 'tree', 'with', 'big', 'smiles']\n",
      "Predicted sentence list: ['a', 'man', 'in', 'a', 'black', 'dog', 'is', 'in', 'a', 'man', 'in', 'a', 'man']\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['a', 'basketball', 'player', 'in', 'the', 'air', 'about', 'to', 'make', 'a', 'shot']\n",
      "Predicted sentence list: ['a', 'man', 'in', 'a', 'man', 'in', 'a', 'man', 'in', 'a', 'man', 'in', 'a', 'man']\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['a', 'black', 'and', 'white', 'dog', 'is', 'playing', 'with', 'a', 'large', 'brown', 'cow']\n",
      "Predicted sentence list: ['a', 'man', 'in', 'a', 'dog', 'in', 'a', 'man', 'in', 'a', 'man', 'in', 'a', 'man']\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['two', 'kids', 'smiling', 'and', 'holding', 'a', 'card']\n",
      "Predicted sentence list: ['a', 'man', 'in', 'a', 'man', 'in', 'a', 'man', 'in', 'a', 'man', 'in', 'a', 'man', 'in', 'a', 'man']\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['a', 'person', 'flipping', 'upside', 'down']\n",
      "Predicted sentence list: ['a', 'man', 'in', 'a', 'man', 'in', 'a', 'man', 'in', 'a', 'man', 'in', 'a', 'man']\n",
      "Loss: | Train: 5.51438 | Validation: 4.66735\n",
      "BLEU Score: 0.0715070797678644\n",
      "---Epoch 2---\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['a', 'four', 'wheeler', 'jumping', 'in', 'an', 'empty', 'field']\n",
      "Predicted sentence list: ['a', 'man', 'in', 'a', 'blue', 'shirt', 'is', 'jumping', 'on', 'a', 'blue', 'shirt']\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['a', 'woman', 'outside', 'covers', 'her', 'face', 'with', 'fabric']\n",
      "Predicted sentence list: ['a', 'man', 'in', 'a', 'red', 'shirt', 'is', 'standing', 'in', 'a', 'red', 'shirt', 'and', 'a', 'red', 'shirt', 'and', 'white', 'shirt']\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['a', 'large', 'brown', 'dog', 'and', 'a', 'larger', 'black', 'dog', 'playing', 'together', 'with', 'a', 'ball']\n",
      "Predicted sentence list: ['a', 'man', 'in', 'a', 'red', 'shirt', 'is', 'standing', 'in', 'a', 'red', 'shirt', 'and', 'white', 'shirt', 'and', 'a', 'red', 'shirt']\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['two', 'men', 'resting', 'a', 'snowy', 'mountain']\n",
      "Predicted sentence list: ['a', 'man', 'in', 'a', 'blue', 'shirt', 'is', 'jumping', 'on', 'a', 'blue', 'shirt']\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['a', 'sponsored', 'race', 'car', 'is', 'driving', 'with', 'lights', 'on', 'a', 'wet']\n",
      "Predicted sentence list: ['a', 'man', 'in', 'a', 'blue', 'shirt', 'is', 'jumping', 'on', 'a', 'blue', 'shirt']\n",
      "Loss: | Train: 4.10697 | Validation: 4.17113\n",
      "BLEU Score: 0.7170620859913712\n",
      "---Epoch 3---\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['a', 'group', 'of', 'children', 'dressed', 'in', 'costume', 'pose', 'for', 'a', 'picture']\n",
      "Predicted sentence list: ['a', 'man', 'in', 'a', 'red', 'shirt', 'is', 'standing', 'on', 'a', 'red', 'shirt']\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['several', 'people', 'in', 'a', 'band', 'jumping', 'up', 'and', 'down', 'to', 'the', 'music']\n",
      "Predicted sentence list: ['a', 'man', 'in', 'a', 'red', 'shirt', 'is', 'standing', 'on', 'a', 'red', 'shirt']\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['a', 'woman', 'paints', 'a', 'picture', 'on', 'a', 'girl', 's', 'face']\n",
      "Predicted sentence list: ['a', 'man', 'in', 'a', 'red', 'shirt', 'is', 'standing', 'on', 'a', 'beach']\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['a', 'mother', 'helps', 'her', 'little', 'boy', 'eat', 'his', 'food', 'by', 'feeding', 'him', 'by', 'hand']\n",
      "Predicted sentence list: ['a', 'man', 'in', 'a', 'red', 'shirt', 'is', 'standing', 'on', 'a', 'red', 'shirt']\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['two', 'dogs', 'chase', 'another', 'dog', 'that', 'has', 'a', 'ball']\n",
      "Predicted sentence list: ['a', 'man', 'in', 'a', 'red', 'shirt', 'is', 'running', 'on', 'a', 'field']\n",
      "Loss: | Train: 3.71779 | Validation: 3.98390\n",
      "BLEU Score: 1.164531083905814\n",
      "---Epoch 4---\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[5], line 2\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[38;5;66;03m# Train the model\u001b[39;00m\n\u001b[0;32m----> 2\u001b[0m stats \u001b[38;5;241m=\u001b[39m \u001b[43mtrain\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m      3\u001b[0m \u001b[43m    \u001b[49m\u001b[43mmodel\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mmodel\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m      4\u001b[0m \u001b[43m    \u001b[49m\u001b[43mtrain_dataloader\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mtrain_dataloader\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m      5\u001b[0m \u001b[43m    \u001b[49m\u001b[43mvalid_dataloader\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mvalid_dataloader\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m      6\u001b[0m \u001b[43m    \u001b[49m\u001b[43mtest_dataloader\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mtest_dataloader\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m      7\u001b[0m \u001b[43m    \u001b[49m\u001b[43mloss_func\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mloss_fn\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m      8\u001b[0m \u001b[43m    \u001b[49m\u001b[43moptimizer\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43moptimizer\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m      9\u001b[0m \u001b[43m    \u001b[49m\u001b[43mpad_idx\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mpad_idx\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m     10\u001b[0m \u001b[43m    \u001b[49m\u001b[43mtgt_vocab\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43men_vocab\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m     11\u001b[0m \u001b[43m    \u001b[49m\u001b[43mdevice\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mdevice\u001b[49m\n\u001b[1;32m     12\u001b[0m \u001b[43m)\u001b[49m\n",
      "Cell \u001b[0;32mIn[2], line 381\u001b[0m, in \u001b[0;36mtrain\u001b[0;34m(model, train_dataloader, valid_dataloader, test_dataloader, loss_func, optimizer, pad_idx, tgt_vocab, device)\u001b[0m\n\u001b[1;32m    379\u001b[0m model\u001b[38;5;241m.\u001b[39meval()\n\u001b[1;32m    380\u001b[0m \u001b[38;5;28;01mwith\u001b[39;00m torch\u001b[38;5;241m.\u001b[39mno_grad():\n\u001b[0;32m--> 381\u001b[0m     avg_valid_loss \u001b[38;5;241m=\u001b[39m \u001b[43mrun_one_epoch\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m    382\u001b[0m \u001b[43m        \u001b[49m\u001b[43mepoch\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    383\u001b[0m \u001b[43m        \u001b[49m\u001b[43mmodel\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    384\u001b[0m \u001b[43m        \u001b[49m\u001b[43mvalid_dataloader\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    385\u001b[0m \u001b[43m        \u001b[49m\u001b[43mloss_func\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    386\u001b[0m \u001b[43m        \u001b[49m\u001b[43moptimizer\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    387\u001b[0m \u001b[43m        \u001b[49m\u001b[43mpad_idx\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    388\u001b[0m \u001b[43m        \u001b[49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43mVALID\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m,\u001b[49m\n\u001b[1;32m    389\u001b[0m \u001b[43m        \u001b[49m\u001b[43mdevice\u001b[49m\n\u001b[1;32m    390\u001b[0m \u001b[43m    \u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    392\u001b[0m \u001b[38;5;66;03m# Calculate the bleu score\u001b[39;00m\n\u001b[1;32m    393\u001b[0m \u001b[38;5;28;01mwith\u001b[39;00m torch\u001b[38;5;241m.\u001b[39mno_grad():\n",
      "Cell \u001b[0;32mIn[2], line 298\u001b[0m, in \u001b[0;36mrun_one_epoch\u001b[0;34m(epoch_index, model, dataloader, loss_func, optimizer, pad_idx, split, device)\u001b[0m\n\u001b[1;32m    293\u001b[0m \u001b[38;5;250m\u001b[39m\u001b[38;5;124;03m'''\u001b[39;00m\n\u001b[1;32m    294\u001b[0m \u001b[38;5;124;03mPerforms one epoch of either training with backpropagation or validation or testing inference\u001b[39;00m\n\u001b[1;32m    295\u001b[0m \u001b[38;5;124;03m'''\u001b[39;00m\n\u001b[1;32m    296\u001b[0m running_loss \u001b[38;5;241m=\u001b[39m \u001b[38;5;241m0.\u001b[39m\n\u001b[0;32m--> 298\u001b[0m \u001b[38;5;28;01mfor\u001b[39;00m src, tgt \u001b[38;5;129;01min\u001b[39;00m dataloader:\n\u001b[1;32m    299\u001b[0m     \u001b[38;5;66;03m# Change the device and format\u001b[39;00m\n\u001b[1;32m    300\u001b[0m     tgt \u001b[38;5;241m=\u001b[39m tgt\u001b[38;5;241m.\u001b[39mtype(torch\u001b[38;5;241m.\u001b[39mLongTensor)\n\u001b[1;32m    301\u001b[0m     src \u001b[38;5;241m=\u001b[39m src\u001b[38;5;241m.\u001b[39mto(device\u001b[38;5;241m=\u001b[39mdevice)\n",
      "File \u001b[0;32m~/anaconda3/lib/python3.11/site-packages/torch/utils/data/dataloader.py:631\u001b[0m, in \u001b[0;36m_BaseDataLoaderIter.__next__\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    628\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_sampler_iter \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n\u001b[1;32m    629\u001b[0m     \u001b[38;5;66;03m# TODO(https://github.com/pytorch/pytorch/issues/76750)\u001b[39;00m\n\u001b[1;32m    630\u001b[0m     \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_reset()  \u001b[38;5;66;03m# type: ignore[call-arg]\u001b[39;00m\n\u001b[0;32m--> 631\u001b[0m data \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_next_data\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    632\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_num_yielded \u001b[38;5;241m+\u001b[39m\u001b[38;5;241m=\u001b[39m \u001b[38;5;241m1\u001b[39m\n\u001b[1;32m    633\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_dataset_kind \u001b[38;5;241m==\u001b[39m _DatasetKind\u001b[38;5;241m.\u001b[39mIterable \u001b[38;5;129;01mand\u001b[39;00m \\\n\u001b[1;32m    634\u001b[0m         \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_IterableDataset_len_called \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m \u001b[38;5;129;01mand\u001b[39;00m \\\n\u001b[1;32m    635\u001b[0m         \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_num_yielded \u001b[38;5;241m>\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_IterableDataset_len_called:\n",
      "File \u001b[0;32m~/anaconda3/lib/python3.11/site-packages/torch/utils/data/dataloader.py:675\u001b[0m, in \u001b[0;36m_SingleProcessDataLoaderIter._next_data\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    673\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21m_next_data\u001b[39m(\u001b[38;5;28mself\u001b[39m):\n\u001b[1;32m    674\u001b[0m     index \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_next_index()  \u001b[38;5;66;03m# may raise StopIteration\u001b[39;00m\n\u001b[0;32m--> 675\u001b[0m     data \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_dataset_fetcher\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mfetch\u001b[49m\u001b[43m(\u001b[49m\u001b[43mindex\u001b[49m\u001b[43m)\u001b[49m  \u001b[38;5;66;03m# may raise StopIteration\u001b[39;00m\n\u001b[1;32m    676\u001b[0m     \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_pin_memory:\n\u001b[1;32m    677\u001b[0m         data \u001b[38;5;241m=\u001b[39m _utils\u001b[38;5;241m.\u001b[39mpin_memory\u001b[38;5;241m.\u001b[39mpin_memory(data, \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_pin_memory_device)\n",
      "File \u001b[0;32m~/anaconda3/lib/python3.11/site-packages/torch/utils/data/_utils/fetch.py:51\u001b[0m, in \u001b[0;36m_MapDatasetFetcher.fetch\u001b[0;34m(self, possibly_batched_index)\u001b[0m\n\u001b[1;32m     49\u001b[0m         data \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mdataset\u001b[38;5;241m.\u001b[39m__getitems__(possibly_batched_index)\n\u001b[1;32m     50\u001b[0m     \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[0;32m---> 51\u001b[0m         data \u001b[38;5;241m=\u001b[39m \u001b[43m[\u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mdataset\u001b[49m\u001b[43m[\u001b[49m\u001b[43midx\u001b[49m\u001b[43m]\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;28;43;01mfor\u001b[39;49;00m\u001b[43m \u001b[49m\u001b[43midx\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;129;43;01min\u001b[39;49;00m\u001b[43m \u001b[49m\u001b[43mpossibly_batched_index\u001b[49m\u001b[43m]\u001b[49m\n\u001b[1;32m     52\u001b[0m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[1;32m     53\u001b[0m     data \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mdataset[possibly_batched_index]\n",
      "File \u001b[0;32m~/anaconda3/lib/python3.11/site-packages/torch/utils/data/_utils/fetch.py:51\u001b[0m, in \u001b[0;36m<listcomp>\u001b[0;34m(.0)\u001b[0m\n\u001b[1;32m     49\u001b[0m         data \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mdataset\u001b[38;5;241m.\u001b[39m__getitems__(possibly_batched_index)\n\u001b[1;32m     50\u001b[0m     \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[0;32m---> 51\u001b[0m         data \u001b[38;5;241m=\u001b[39m [\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mdataset\u001b[49m\u001b[43m[\u001b[49m\u001b[43midx\u001b[49m\u001b[43m]\u001b[49m \u001b[38;5;28;01mfor\u001b[39;00m idx \u001b[38;5;129;01min\u001b[39;00m possibly_batched_index]\n\u001b[1;32m     52\u001b[0m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[1;32m     53\u001b[0m     data \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mdataset[possibly_batched_index]\n",
      "Cell \u001b[0;32mIn[2], line 185\u001b[0m, in \u001b[0;36mDatasetFlickr8K.__getitem__\u001b[0;34m(self, idx)\u001b[0m\n\u001b[1;32m    183\u001b[0m img_name \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mdata_dir \u001b[38;5;241m+\u001b[39m \u001b[38;5;124m'\u001b[39m\u001b[38;5;124mImages/\u001b[39m\u001b[38;5;124m'\u001b[39m \u001b[38;5;241m+\u001b[39m img_data[\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mfilename\u001b[39m\u001b[38;5;124m'\u001b[39m]\n\u001b[1;32m    184\u001b[0m \u001b[38;5;28;01mwith\u001b[39;00m Image\u001b[38;5;241m.\u001b[39mopen(img_name) \u001b[38;5;28;01mas\u001b[39;00m raw_img:\n\u001b[0;32m--> 185\u001b[0m     transformed_img \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mimg_transform\u001b[49m\u001b[43m(\u001b[49m\u001b[43mraw_img\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    187\u001b[0m \u001b[38;5;66;03m# Get the caption tokens\u001b[39;00m\n\u001b[1;32m    188\u001b[0m tokens \u001b[38;5;241m=\u001b[39m caption[\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mtokens\u001b[39m\u001b[38;5;124m'\u001b[39m]\n",
      "File \u001b[0;32m~/anaconda3/lib/python3.11/site-packages/torchvision/transforms/transforms.py:95\u001b[0m, in \u001b[0;36mCompose.__call__\u001b[0;34m(self, img)\u001b[0m\n\u001b[1;32m     93\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21m__call__\u001b[39m(\u001b[38;5;28mself\u001b[39m, img):\n\u001b[1;32m     94\u001b[0m     \u001b[38;5;28;01mfor\u001b[39;00m t \u001b[38;5;129;01min\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mtransforms:\n\u001b[0;32m---> 95\u001b[0m         img \u001b[38;5;241m=\u001b[39m \u001b[43mt\u001b[49m\u001b[43m(\u001b[49m\u001b[43mimg\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m     96\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m img\n",
      "File \u001b[0;32m~/anaconda3/lib/python3.11/site-packages/torch/nn/modules/module.py:1511\u001b[0m, in \u001b[0;36mModule._wrapped_call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1509\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_compiled_call_impl(\u001b[38;5;241m*\u001b[39margs, \u001b[38;5;241m*\u001b[39m\u001b[38;5;241m*\u001b[39mkwargs)  \u001b[38;5;66;03m# type: ignore[misc]\u001b[39;00m\n\u001b[1;32m   1510\u001b[0m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[0;32m-> 1511\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_call_impl\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m~/anaconda3/lib/python3.11/site-packages/torch/nn/modules/module.py:1520\u001b[0m, in \u001b[0;36mModule._call_impl\u001b[0;34m(self, *args, **kwargs)\u001b[0m\n\u001b[1;32m   1515\u001b[0m \u001b[38;5;66;03m# If we don't have any hooks, we want to skip the rest of the logic in\u001b[39;00m\n\u001b[1;32m   1516\u001b[0m \u001b[38;5;66;03m# this function, and just call forward.\u001b[39;00m\n\u001b[1;32m   1517\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m (\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_forward_pre_hooks\n\u001b[1;32m   1518\u001b[0m         \u001b[38;5;129;01mor\u001b[39;00m _global_backward_pre_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_backward_hooks\n\u001b[1;32m   1519\u001b[0m         \u001b[38;5;129;01mor\u001b[39;00m _global_forward_hooks \u001b[38;5;129;01mor\u001b[39;00m _global_forward_pre_hooks):\n\u001b[0;32m-> 1520\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mforward_call\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43margs\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[38;5;241;43m*\u001b[39;49m\u001b[43mkwargs\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   1522\u001b[0m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[1;32m   1523\u001b[0m     result \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;01mNone\u001b[39;00m\n",
      "File \u001b[0;32m~/anaconda3/lib/python3.11/site-packages/torchvision/transforms/transforms.py:354\u001b[0m, in \u001b[0;36mResize.forward\u001b[0;34m(self, img)\u001b[0m\n\u001b[1;32m    346\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21mforward\u001b[39m(\u001b[38;5;28mself\u001b[39m, img):\n\u001b[1;32m    347\u001b[0m \u001b[38;5;250m    \u001b[39m\u001b[38;5;124;03m\"\"\"\u001b[39;00m\n\u001b[1;32m    348\u001b[0m \u001b[38;5;124;03m    Args:\u001b[39;00m\n\u001b[1;32m    349\u001b[0m \u001b[38;5;124;03m        img (PIL Image or Tensor): Image to be scaled.\u001b[39;00m\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m    352\u001b[0m \u001b[38;5;124;03m        PIL Image or Tensor: Rescaled image.\u001b[39;00m\n\u001b[1;32m    353\u001b[0m \u001b[38;5;124;03m    \"\"\"\u001b[39;00m\n\u001b[0;32m--> 354\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mF\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mresize\u001b[49m\u001b[43m(\u001b[49m\u001b[43mimg\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43msize\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43minterpolation\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mmax_size\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mantialias\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m~/anaconda3/lib/python3.11/site-packages/torchvision/transforms/functional.py:467\u001b[0m, in \u001b[0;36mresize\u001b[0;34m(img, size, interpolation, max_size, antialias)\u001b[0m\n\u001b[1;32m    465\u001b[0m         warnings\u001b[38;5;241m.\u001b[39mwarn(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mAnti-alias option is always applied for PIL Image input. Argument antialias is ignored.\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n\u001b[1;32m    466\u001b[0m     pil_interpolation \u001b[38;5;241m=\u001b[39m pil_modes_mapping[interpolation]\n\u001b[0;32m--> 467\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mF_pil\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mresize\u001b[49m\u001b[43m(\u001b[49m\u001b[43mimg\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43msize\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43moutput_size\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43minterpolation\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mpil_interpolation\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    469\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m F_t\u001b[38;5;241m.\u001b[39mresize(img, size\u001b[38;5;241m=\u001b[39moutput_size, interpolation\u001b[38;5;241m=\u001b[39minterpolation\u001b[38;5;241m.\u001b[39mvalue, antialias\u001b[38;5;241m=\u001b[39mantialias)\n",
      "File \u001b[0;32m~/anaconda3/lib/python3.11/site-packages/torchvision/transforms/_functional_pil.py:250\u001b[0m, in \u001b[0;36mresize\u001b[0;34m(img, size, interpolation)\u001b[0m\n\u001b[1;32m    247\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m (\u001b[38;5;28misinstance\u001b[39m(size, \u001b[38;5;28mlist\u001b[39m) \u001b[38;5;129;01mand\u001b[39;00m \u001b[38;5;28mlen\u001b[39m(size) \u001b[38;5;241m==\u001b[39m \u001b[38;5;241m2\u001b[39m):\n\u001b[1;32m    248\u001b[0m     \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mTypeError\u001b[39;00m(\u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mGot inappropriate size arg: \u001b[39m\u001b[38;5;132;01m{\u001b[39;00msize\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m\"\u001b[39m)\n\u001b[0;32m--> 250\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43mimg\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mresize\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;28;43mtuple\u001b[39;49m\u001b[43m(\u001b[49m\u001b[43msize\u001b[49m\u001b[43m[\u001b[49m\u001b[43m:\u001b[49m\u001b[43m:\u001b[49m\u001b[38;5;241;43m-\u001b[39;49m\u001b[38;5;241;43m1\u001b[39;49m\u001b[43m]\u001b[49m\u001b[43m)\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43minterpolation\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m~/anaconda3/lib/python3.11/site-packages/PIL/Image.py:2164\u001b[0m, in \u001b[0;36mImage.resize\u001b[0;34m(self, size, resample, box, reducing_gap)\u001b[0m\n\u001b[1;32m   2160\u001b[0m     \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mValueError\u001b[39;00m(msg)\n\u001b[1;32m   2162\u001b[0m size \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mtuple\u001b[39m(size)\n\u001b[0;32m-> 2164\u001b[0m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mload\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   2165\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m box \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n\u001b[1;32m   2166\u001b[0m     box \u001b[38;5;241m=\u001b[39m (\u001b[38;5;241m0\u001b[39m, \u001b[38;5;241m0\u001b[39m) \u001b[38;5;241m+\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39msize\n",
      "File \u001b[0;32m~/anaconda3/lib/python3.11/site-packages/PIL/ImageFile.py:291\u001b[0m, in \u001b[0;36mImageFile.load\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    288\u001b[0m         \u001b[38;5;28;01mraise\u001b[39;00m \u001b[38;5;167;01mOSError\u001b[39;00m(msg)\n\u001b[1;32m    290\u001b[0m b \u001b[38;5;241m=\u001b[39m b \u001b[38;5;241m+\u001b[39m s\n\u001b[0;32m--> 291\u001b[0m n, err_code \u001b[38;5;241m=\u001b[39m \u001b[43mdecoder\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mdecode\u001b[49m\u001b[43m(\u001b[49m\u001b[43mb\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    292\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m n \u001b[38;5;241m<\u001b[39m \u001b[38;5;241m0\u001b[39m:\n\u001b[1;32m    293\u001b[0m     \u001b[38;5;28;01mbreak\u001b[39;00m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "# Train the model\n",
    "stats = train(\n",
    "    model=model,\n",
    "    train_dataloader=train_dataloader,\n",
    "    valid_dataloader=valid_dataloader,\n",
    "    test_dataloader=test_dataloader,\n",
    "    loss_func=loss_fn,\n",
    "    optimizer=optimizer,\n",
    "    pad_idx=pad_idx,\n",
    "    tgt_vocab=en_vocab,\n",
    "    device=device\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "metadata": {}
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['a', 'group', 'of', 'children', 'in', 'uniforms', 'stand', 'outside']\n",
      "Predicted sentence list: ['a', 'boy', 'in', 'a', 'blue', 'shirt', 'is', 'riding', 'a', 'skateboard', 'down', 'a', 'dirt', 'road']\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['a', 'dog', 'prepares', 'to', 'catch', 'a', 'thrown', 'object', 'in', 'a', 'field', 'with', 'nearby', 'cars']\n",
      "Predicted sentence list: ['a', 'girl', 'in', 'a', 'green', 'shirt', 'and', 'jeans', 'is', 'sitting', 'on', 'a', 'swing']\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['a', 'bald', 'basketball', 'player', 'is', 'standing', 'on', 'a', 'court', 'with', 'the', 'ball', 'in', 'his', 'right', 'hand']\n",
      "Predicted sentence list: ['a', 'man', 'in', 'a', 'white', 'shirt', 'is', 'sitting', 'on', 'a', 'bench', 'with', 'a', 'woman', 'in', 'a', 'white', 'shirt', 'and', 'black', 'shirt']\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['a', 'brown', 'and', 'white', 'dog', 'is', 'jumping', 'with', 'a', 'red', 'frisbee', 'in', 'its', 'mouth']\n",
      "Predicted sentence list: ['a', 'dog', 'is', 'running', 'in', 'a', 'field']\n",
      "Intermediary bleu: 0\n",
      "Target sentence list   : ['a', 'man', 'wakeboards', 'in', 'a', 'red', 'vest', 'and', 'smiles', 'for', 'the', 'camera']\n",
      "Predicted sentence list: ['a', 'girl', 'in', 'a', 'pink', 'shirt', 'is', 'jumping', 'up', 'at', 'a', 'playground']\n",
      "Testing BLEU Score: 2.1560281220942032\n",
      "Final BLEU: 2.1560281220942032\n"
     ]
    }
   ],
   "source": [
    "final_bleu = test_captioner(\n",
    "    datapath,\n",
    "    test_data_fname,\n",
    "    vocab_path=english_vocab_path,\n",
    "    model_param_path='saved_model_lol.pt'\n",
    ")\n",
    "print(\"Final BLEU:\", final_bleu)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Create the performance plots\n",
    "plot_performance(stats)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "base",
   "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": "3.11.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}