model.py

"""A small GPT-style decoder-only transformer, from scratch in PyTorch."""

import math

import torch
import torch.nn as nn
import torch.nn.functional as F


class MultiHeadSelfAttention(nn.Module):
    def __init__(self, d_model: int, n_heads: int, dropout: float = 0.1):
        super().__init__()
        assert d_model % n_heads == 0
        self.n_heads = n_heads
        self.head_dim = d_model // n_heads

        self.qkv = nn.Linear(d_model, 3 * d_model)
        self.out_proj = nn.Linear(d_model, d_model)
        self.dropout = nn.Dropout(dropout)

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        B, T, C = x.shape

        qkv = self.qkv(x)  # (B, T, 3*C)
        q, k, v = qkv.chunk(3, dim=-1)

        # Reshape to (B, n_heads, T, head_dim)
        q = q.view(B, T, self.n_heads, self.head_dim).transpose(1, 2)
        k = k.view(B, T, self.n_heads, self.head_dim).transpose(1, 2)
        v = v.view(B, T, self.n_heads, self.head_dim).transpose(1, 2)

        # Scaled dot-product attention with causal mask
        scale = math.sqrt(self.head_dim)
        attn = (q @ k.transpose(-2, -1)) / scale  # (B, nh, T, T)

        # Causal mask: prevent attending to future positions
        mask = torch.triu(torch.ones(T, T, device=x.device), diagonal=1).bool()
        attn = attn.masked_fill(mask, float("-inf"))

        attn = F.softmax(attn, dim=-1)
        attn = self.dropout(attn)

        out = attn @ v  # (B, nh, T, head_dim)
        out = out.transpose(1, 2).contiguous().view(B, T, C)
        return self.out_proj(out)


class FeedForward(nn.Module):
    def __init__(self, d_model: int, d_ff: int, dropout: float = 0.1):
        super().__init__()
        self.net = nn.Sequential(
            nn.Linear(d_model, d_ff),
            nn.GELU(),
            nn.Linear(d_ff, d_model),
            nn.Dropout(dropout),
        )

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        return self.net(x)


class TransformerBlock(nn.Module):
    def __init__(self, d_model: int, n_heads: int, d_ff: int, dropout: float = 0.1):
        super().__init__()
        self.ln1 = nn.LayerNorm(d_model)
        self.attn = MultiHeadSelfAttention(d_model, n_heads, dropout)
        self.ln2 = nn.LayerNorm(d_model)
        self.ff = FeedForward(d_model, d_ff, dropout)

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        # Pre-norm architecture (like GPT-2)
        x = x + self.attn(self.ln1(x))
        x = x + self.ff(self.ln2(x))
        return x


class TinyLLM(nn.Module):
    """A small decoder-only transformer language model.

    Architecture mirrors GPT-2 but much smaller:
    - Token embeddings + learned positional embeddings
    - N transformer blocks with pre-norm
    - Final layer norm + linear head to vocab
    """

    def __init__(
        self,
        vocab_size: int,
        d_model: int = 256,
        n_heads: int = 8,
        n_layers: int = 6,
        d_ff: int = 1024,
        max_seq_len: int = 256,
        dropout: float = 0.1,
    ):
        super().__init__()
        self.max_seq_len = max_seq_len

        self.token_emb = nn.Embedding(vocab_size, d_model)
        self.pos_emb = nn.Embedding(max_seq_len, d_model)
        self.drop = nn.Dropout(dropout)

        self.blocks = nn.ModuleList(
            [TransformerBlock(d_model, n_heads, d_ff, dropout) for _ in range(n_layers)]
        )
        self.ln_f = nn.LayerNorm(d_model)
        self.head = nn.Linear(d_model, vocab_size, bias=False)

        # Weight tying: share token embedding weights with output head
        self.head.weight = self.token_emb.weight

        self._init_weights()

    def _init_weights(self):
        for module in self.modules():
            if isinstance(module, nn.Linear):
                torch.nn.init.normal_(module.weight, mean=0.0, std=0.02)
                if module.bias is not None:
                    torch.nn.init.zeros_(module.bias)
            elif isinstance(module, nn.Embedding):
                torch.nn.init.normal_(module.weight, mean=0.0, std=0.02)

    def forward(
        self, idx: torch.Tensor, targets: torch.Tensor | None = None
    ) -> tuple[torch.Tensor, torch.Tensor | None]:
        B, T = idx.shape
        assert T <= self.max_seq_len, f"Sequence length {T} exceeds max {self.max_seq_len}"

        # Embeddings
        tok_emb = self.token_emb(idx)  # (B, T, d_model)
        pos = torch.arange(T, device=idx.device)
        pos_emb = self.pos_emb(pos)  # (T, d_model)
        x = self.drop(tok_emb + pos_emb)

        # Transformer blocks
        for block in self.blocks:
            x = block(x)

        x = self.ln_f(x)
        logits = self.head(x)  # (B, T, vocab_size)

        loss = None
        if targets is not None:
            loss = F.cross_entropy(
                logits.view(-1, logits.size(-1)),
                targets.view(-1),
            )

        return logits, loss

    def param_count(self) -> int:
        return sum(p.numel() for p in self.parameters() if p.requires_grad)

    @torch.no_grad()
    def generate(
        self,
        idx: torch.Tensor,
        max_new_tokens: int = 200,
        temperature: float = 0.8,
        top_k: int = 40,
    ) -> torch.Tensor:
        """Autoregressive generation with temperature and top-k sampling."""
        for _ in range(max_new_tokens):
            # Crop to max_seq_len
            idx_cond = idx[:, -self.max_seq_len :]
            logits, _ = self(idx_cond)
            logits = logits[:, -1, :] / temperature

            # Top-k filtering
            if top_k > 0:
                v, _ = torch.topk(logits, min(top_k, logits.size(-1)))
                logits[logits < v[:, [-1]]] = float("-inf")

            probs = F.softmax(logits, dim=-1)
            next_token = torch.multinomial(probs, num_samples=1)
            idx = torch.cat([idx, next_token], dim=1)

        return idx