""" Stage 2: MODEL ARCHITECTURE. A from-scratch GPT-style decoder-only transformer, built directly in PyTorch (no pretrained weights, no transformers.AutoModel -- this is the part that makes "built an LLM from scratch" an honest claim): - token + positional embeddings - N transformer blocks, each with: multi-head causal self-attention, a feed-forward MLP, and layer norm (pre-norm, like GPT-2) - a final layer norm + linear head tied to the token embedding weights Sizes (see PRESETS below): - "tiny" ~ a few M params -> for a fast CPU dry-run / sanity check - "small" ~ 30M params -> trains in well under an hour on an RTX 3060 - "medium" ~ 125M params -> GPT-1 scale, still fits an RTX 3060 12GB """ import math from dataclasses import dataclass import torch import torch.nn as nn from torch.nn import functional as F PRESETS = { "tiny": dict(n_layer=4, n_head=4, n_embd=128, block_size=128), # CPU dry-run "small": dict(n_layer=6, n_head=6, n_embd=384, block_size=256), # ~30M params "medium": dict(n_layer=12, n_head=12, n_embd=768, block_size=512), # ~125M params } @dataclass class GPTConfig: vocab_size: int = 8000 block_size: int = 256 n_layer: int = 6 n_head: int = 6 n_embd: int = 384 dropout: float = 0.1 class CausalSelfAttention(nn.Module): def __init__(self, config: GPTConfig): super().__init__() assert config.n_embd % config.n_head == 0 self.n_head = config.n_head self.head_dim = config.n_embd // config.n_head self.qkv = nn.Linear(config.n_embd, 3 * config.n_embd) self.proj = nn.Linear(config.n_embd, config.n_embd) self.attn_dropout = nn.Dropout(config.dropout) self.resid_dropout = nn.Dropout(config.dropout) mask = torch.tril(torch.ones(config.block_size, config.block_size)) self.register_buffer("mask", mask.view(1, 1, config.block_size, config.block_size)) def forward(self, x): B, T, C = x.shape qkv = self.qkv(x) q, k, v = qkv.split(C, dim=2) q = q.view(B, T, self.n_head, self.head_dim).transpose(1, 2) k = k.view(B, T, self.n_head, self.head_dim).transpose(1, 2) v = v.view(B, T, self.n_head, self.head_dim).transpose(1, 2) att = (q @ k.transpose(-2, -1)) / math.sqrt(self.head_dim) att = att.masked_fill(self.mask[:, :, :T, :T] == 0, float("-inf")) att = F.softmax(att, dim=-1) att = self.attn_dropout(att) y = att @ v y = y.transpose(1, 2).contiguous().view(B, T, C) return self.resid_dropout(self.proj(y)) class MLP(nn.Module): def __init__(self, config: GPTConfig): super().__init__() self.fc = nn.Linear(config.n_embd, 4 * config.n_embd) self.proj = nn.Linear(4 * config.n_embd, config.n_embd) self.dropout = nn.Dropout(config.dropout) def forward(self, x): return self.dropout(self.proj(F.gelu(self.fc(x)))) class Block(nn.Module): def __init__(self, config: GPTConfig): super().__init__() self.ln1 = nn.LayerNorm(config.n_embd) self.attn = CausalSelfAttention(config) self.ln2 = nn.LayerNorm(config.n_embd) self.mlp = MLP(config) def forward(self, x): x = x + self.attn(self.ln1(x)) x = x + self.mlp(self.ln2(x)) return x class GPT(nn.Module): def __init__(self, config: GPTConfig): super().__init__() self.config = config self.tok_emb = nn.Embedding(config.vocab_size, config.n_embd) self.pos_emb = nn.Embedding(config.block_size, config.n_embd) self.drop = nn.Dropout(config.dropout) self.blocks = nn.ModuleList([Block(config) for _ in range(config.n_layer)]) self.ln_f = nn.LayerNorm(config.n_embd) self.head = nn.Linear(config.n_embd, config.vocab_size, bias=False) self.head.weight = self.tok_emb.weight # weight tying self.apply(self._init_weights) n_params = sum(p.numel() for p in self.parameters()) print(f"GPT initialized: {n_params/1e6:.2f}M parameters " f"(n_layer={config.n_layer}, n_head={config.n_head}, " f"n_embd={config.n_embd}, vocab_size={config.vocab_size}, " f"block_size={config.block_size})") def _init_weights(self, module): if isinstance(module, nn.Linear): nn.init.normal_(module.weight, mean=0.0, std=0.02) if module.bias is not None: nn.init.zeros_(module.bias) elif isinstance(module, nn.Embedding): nn.init.normal_(module.weight, mean=0.0, std=0.02) def forward(self, idx, targets=None): B, T = idx.shape assert T <= self.config.block_size, "sequence longer than block_size" pos = torch.arange(0, T, device=idx.device).unsqueeze(0) x = self.drop(self.tok_emb(idx) + self.pos_emb(pos)) for block in self.blocks: x = block(x) x = self.ln_f(x) logits = self.head(x) loss = None if targets is not None: loss = F.cross_entropy(logits.view(-1, logits.size(-1)), targets.view(-1), ignore_index=-1) return logits, loss @torch.no_grad() def generate(self, idx, max_new_tokens, temperature=1.0, top_k=None, top_p=None): self.eval() for _ in range(max_new_tokens): idx_cond = idx[:, -self.config.block_size:] logits, _ = self(idx_cond) logits = logits[:, -1, :] / max(temperature, 1e-5) if top_k is not None: v, _ = torch.topk(logits, min(top_k, logits.size(-1))) logits[logits < v[:, [-1]]] = float("-inf") if top_p is not None: sorted_logits, sorted_idx = torch.sort(logits, descending=True) probs = F.softmax(sorted_logits, dim=-1) cum_probs = torch.cumsum(probs, dim=-1) remove = cum_probs > top_p remove[:, 1:] = remove[:, :-1].clone() remove[:, 0] = False sorted_logits[remove] = float("-inf") logits = torch.full_like(logits, float("-inf")).scatter(1, sorted_idx, sorted_logits) probs = F.softmax(logits, dim=-1) next_id = torch.multinomial(probs, num_samples=1) idx = torch.cat([idx, next_id], dim=1) return idx