"""Tiny char-level GPT with exposed attention weights (for SCI proxy measurement)."""
import math, torch, torch.nn as nn
import torch.nn.functional as F

class Head(nn.Module):
    def __init__(self, n_embd, head_size, block_size, dropout=0.1):
        super().__init__()
        self.key = nn.Linear(n_embd, head_size, bias=False)
        self.query = nn.Linear(n_embd, head_size, bias=False)
        self.value = nn.Linear(n_embd, head_size, bias=False)
        self.register_buffer('tril', torch.tril(torch.ones(block_size, block_size)))
        self.dropout = nn.Dropout(dropout)
        self.last_attn = None  # (B, T, T) saved after forward

    def forward(self, x):
        B, T, C = x.shape
        k, q, v = self.key(x), self.query(x), self.value(x)
        wei = q @ k.transpose(-2, -1) * (k.shape[-1] ** -0.5)
        wei = wei.masked_fill(self.tril[:T, :T] == 0, float('-inf'))
        wei = F.softmax(wei, dim=-1)
        self.last_attn = wei.detach()
        wei = self.dropout(wei)
        return wei @ v

class MHA(nn.Module):
    def __init__(self, n_embd, n_head, block_size):
        super().__init__()
        hs = n_embd // n_head
        self.heads = nn.ModuleList([Head(n_embd, hs, block_size) for _ in range(n_head)])
        self.proj = nn.Linear(n_embd, n_embd)
        self.dropout = nn.Dropout(0.1)

    def forward(self, x):
        out = torch.cat([h(x) for h in self.heads], dim=-1)
        return self.dropout(self.proj(out))

class Block(nn.Module):
    def __init__(self, n_embd, n_head, block_size):
        super().__init__()
        self.sa = MHA(n_embd, n_head, block_size)
        self.ffwd = nn.Sequential(
            nn.Linear(n_embd, 4 * n_embd), nn.GELU(),
            nn.Linear(4 * n_embd, n_embd), nn.Dropout(0.1))
        self.ln1 = nn.LayerNorm(n_embd)
        self.ln2 = nn.LayerNorm(n_embd)

    def forward(self, x):
        x = x + self.sa(self.ln1(x))
        x = x + self.ffwd(self.ln2(x))
        return x

class TinyGPT(nn.Module):
    def __init__(self, vocab_size, n_embd=96, n_head=4, n_layer=3, block_size=96):
        super().__init__()
        self.block_size = block_size
        self.tok_emb = nn.Embedding(vocab_size, n_embd)
        self.pos_emb = nn.Embedding(block_size, n_embd)
        self.blocks = nn.ModuleList([Block(n_embd, n_head, block_size) for _ in range(n_layer)])
        self.ln_f = nn.LayerNorm(n_embd)
        self.head = nn.Linear(n_embd, vocab_size)

    def forward(self, idx, targets=None):
        B, T = idx.shape
        x = self.tok_emb(idx) + self.pos_emb(torch.arange(T, device=idx.device))
        for b in self.blocks:
            x = b(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))
        return logits, loss

    def collect_attention(self):
        """Returns list over layers of (B, n_head, T, T) attention tensors from last forward."""
        out = []
        for blk in self.blocks:
            heads = torch.stack([h.last_attn for h in blk.sa.heads], dim=1)
            out.append(heads)
        return out