GPT.py

# GPT.py
import tiktoken
import torch
import torch.nn as nn
from torch.utils.data import Dataset, DataLoader


class GPTDatasetV1(Dataset):
    """Custom Dataset for GPT Model training.

    This dataset tokenizes the input text and creates input-target pairs
    using a sliding window approach with the specified stride and max length.

    Args:
        txt (str): The input text to be tokenized.
        tokenizer: The tokenizer used to encode the text.
        max_length (int): The maximum length of the input sequences.
        stride (int): The stride size for creating overlapping sequences.
    """
    def __init__(self, txt, tokenizer, max_length, stride):
        self.input_ids = []
        self.target_ids = []

        # Tokenize the entire text
        token_ids = tokenizer.encode(txt)

        if len(token_ids) <= max_length:
            # Check data is short.
            input_chunk = token_ids[:-1]
            target_chunk = token_ids[1:]
            self.input_ids.append(torch.tensor(input_chunk))
            self.target_ids.append(torch.tensor(target_chunk))
        else:
            # Use a sliding window to chunk the book into overlapping sequences of max_length
            for i in range(0, len(token_ids) - max_length, stride):
                input_chunk = token_ids[i:i + max_length]
                target_chunk = token_ids[i + 1: i + max_length + 1]
                self.input_ids.append(torch.tensor(input_chunk))
                self.target_ids.append(torch.tensor(target_chunk))

    def __len__(self):
        return len(self.input_ids)

    def __getitem__(self, idx):
        return self.input_ids[idx], self.target_ids[idx]


def create_dataloader_v1(prompts, batch_size=4, max_length=256, stride=128, shuffle=True, drop_last=True):
    tokenizer = tiktoken.get_encoding("gpt2")
    combined_text = " ".join(prompts)  # Combine list of texts into a single string
    dataset = GPTDatasetV1(combined_text, tokenizer, max_length, stride)
    return DataLoader(dataset, batch_size=batch_size, shuffle=shuffle, drop_last=drop_last)


class MultiHeadAttention(nn.Module):
    """Multi-Head Attention mechanism for the Transformer model.

    Args:
        d_in (int): The dimensionality of the input.
        d_out (int): The dimensionality of the output.
        context_length (int): The length of the context for the attention mechanism.
        dropout (float): The dropout rate.
        num_heads (int): The number of attention heads.
        qkv_bias (bool, optional): Whether to include bias in the QKV projections. Defaults to False.
    """
    def __init__(self, d_in, d_out, context_length, dropout, num_heads, qkv_bias=False):
        super().__init__()
        assert d_out % num_heads == 0, "d_out must be divisible by num_heads"
        self.d_out = d_out
        self.num_heads = num_heads
        self.head_dim = d_out // num_heads

        self.W_query = nn.Linear(d_in, d_out, bias=qkv_bias)
        self.W_key = nn.Linear(d_in, d_out, bias=qkv_bias)
        self.W_value = nn.Linear(d_in, d_out, bias=qkv_bias)
        self.out_proj = nn.Linear(d_out, d_out)
        self.dropout = nn.Dropout(dropout)
        self.register_buffer("mask", torch.triu(torch.ones(context_length, context_length), diagonal=1))

    def forward(self, x):
        """Forward pass for the Multi-Head Attention mechanism.

        Args:
            x (torch.Tensor): The input tensor.

        Returns:
            torch.Tensor: The output tensor after applying multi-head attention.
        """
        b, num_tokens, d_in = x.shape
        keys = self.W_key(x)
        queries = self.W_query(x)
        values = self.W_value(x)

        keys = keys.view(b, num_tokens, self.num_heads, self.head_dim)
        values = values.view(b, num_tokens, self.num_heads, self.head_dim)
        queries = queries.view(b, num_tokens, self.num_heads, self.head_dim)

        keys = keys.transpose(1, 2)
        queries = queries.transpose(1, 2)
        values = values.transpose(1, 2)

        attn_scores = queries @ keys.transpose(2, 3)
        mask_bool = self.mask.bool()[:num_tokens, :num_tokens]
        attn_scores.masked_fill_(mask_bool, -torch.inf)
        attn_weights = torch.softmax(attn_scores / keys.shape[-1]**0.5, dim=-1)
        attn_weights = self.dropout(attn_weights)

        context_vec = (attn_weights @ values).transpose(1, 2)
        context_vec = context_vec.contiguous().view(b, num_tokens, self.d_out)
        context_vec = self.out_proj(context_vec)

        return context_vec


class LayerNorm(nn.Module):
    """Layer Normalization for the Transformer model.

    Args:
        emb_dim (int): The dimensionality of the embeddings.
    """
    def __init__(self, emb_dim):
        super().__init__()
        self.eps = 1e-5
        self.scale = nn.Parameter(torch.ones(emb_dim))
        self.shift = nn.Parameter(torch.zeros(emb_dim))

    def forward(self, x):
        """Forward pass for Layer Normalization.

        Args:
            x (torch.Tensor): The input tensor.

        Returns:
            torch.Tensor: The normalized tensor.
        """
        mean = x.mean(dim=-1, keepdim=True)
        var = x.var(dim=-1, keepdim=True, unbiased=False)
        norm_x = (x - mean) / torch.sqrt(var + self.eps)
        return self.scale * norm_x + self.shift


class GELU(nn.Module):
    """Gaussian Error Linear Unit activation function."""

    def forward(self, x):
        """Forward pass for GELU activation.

        Args:
            x (torch.Tensor): The input tensor.

        Returns:
            torch.Tensor: The tensor after applying GELU activation.
        """
        return 0.5 * x * (1 + torch.tanh(
            torch.sqrt(torch.tensor(2.0 / torch.pi)) *
            (x + 0.044715 * torch.pow(x, 3))
        ))


class FeedForward(nn.Module):
    """Feed Forward Neural Network for the Transformer model.

    Args:
        cfg (dict): Configuration dictionary containing model parameters.
    """
    def __init__(self, cfg):
        super().__init__()
        self.layers = nn.Sequential(
            nn.Linear(cfg["emb_dim"], 4 * cfg["emb_dim"]),
            GELU(),
            nn.Linear(4 * cfg["emb_dim"], cfg["emb_dim"]),
        )

    def forward(self, x):
        """Forward pass for the Feed Forward network.

        Args:
            x (torch.Tensor): The input tensor.

        Returns:
            torch.Tensor: The output tensor after applying the Feed Forward network.
        """
        return self.layers(x)


class TransformerBlock(nn.Module):
    """Single Transformer Block consisting of Multi-Head Attention and Feed Forward network.

    Args:
        cfg (dict): Configuration dictionary containing model parameters.
    """
    def __init__(self, cfg):
        super().__init__()
        self.att = MultiHeadAttention(
            d_in=cfg["emb_dim"],
            d_out=cfg["emb_dim"],
            context_length=cfg["context_length"],
            num_heads=cfg["n_heads"],
            dropout=cfg["drop_rate"],
            qkv_bias=cfg["qkv_bias"])
        self.ff = FeedForward(cfg)
        self.norm1 = LayerNorm(cfg["emb_dim"])
        self.norm2 = LayerNorm(cfg["emb_dim"])
        self.drop_shortcut = nn.Dropout(cfg["drop_rate"])

    def forward(self, x):
        """Forward pass for the Transformer Block.

        Args:
            x (torch.Tensor): The input tensor.

        Returns:
            torch.Tensor: The output tensor after applying the Transformer Block.
        """
        shortcut = x
        x = self.norm1(x)
        x = self.att(x)
        x = self.drop_shortcut(x)
        x = x + shortcut

        shortcut = x
        x = self.norm2(x)
        x = self.ff(x)
        x = self.drop_shortcut(x)
        x = x + shortcut

        return x


class GPTModel(nn.Module):
    """GPT Model consisting of multiple Transformer Blocks.

    Args:
        cfg (dict): Configuration dictionary containing model parameters.
    """
    def __init__(self, cfg):
        super().__init__()
        self.tok_emb = nn.Embedding(cfg["vocab_size"], cfg["emb_dim"])
        self.pos_emb = nn.Embedding(cfg["context_length"], cfg["emb_dim"])
        self.drop_emb = nn.Dropout(cfg["drop_rate"])
        self.trf_blocks = nn.Sequential(
            *[TransformerBlock(cfg) for _ in range(cfg["n_layers"])]
        )
        self.final_norm = LayerNorm(cfg["emb_dim"])
        self.out_head = nn.Linear(cfg["emb_dim"], cfg["vocab_size"], bias=False)

    def forward(self, in_idx):
        batch_size, seq_len = in_idx.shape
        tok_embeds = self.tok_emb(in_idx)
        pos_embeds = self.pos_emb(torch.arange(seq_len, device=in_idx.device))
        x = tok_embeds + pos_embeds
        x = self.drop_emb(x)
        x = self.trf_blocks(x)
        x = self.final_norm(x)
        logits = self.out_head(x)

        return logits