"""A small decoder-only transformer (nanoGPT-style).

Deliberately tiny and standard: token + positional embeddings, a stack of
pre-LayerNorm blocks (causal self-attention + MLP), then a tied linear head.
One of these is trained per (corpus, scheme) cell — identical architecture
everywhere, so the only thing that differs across tokenizers is the vocabulary
(and therefore what the model can even represent).
"""

from __future__ import annotations

import math
from dataclasses import dataclass, asdict

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


@dataclass
class GPTConfig:
    vocab_size: int
    block_size: int = 128
    n_layer: int = 4
    n_head: int = 4
    n_embd: int = 128
    dropout: float = 0.1

    def to_dict(self) -> dict:
        return asdict(self)


class CausalSelfAttention(nn.Module):
    def __init__(self, cfg: GPTConfig):
        super().__init__()
        assert cfg.n_embd % cfg.n_head == 0
        self.n_head = cfg.n_head
        self.n_embd = cfg.n_embd
        self.c_attn = nn.Linear(cfg.n_embd, 3 * cfg.n_embd)
        self.c_proj = nn.Linear(cfg.n_embd, cfg.n_embd)
        self.attn_dropout = cfg.dropout
        self.resid_dropout = nn.Dropout(cfg.dropout)

    def forward(self, x):
        B, T, C = x.shape
        q, k, v = self.c_attn(x).split(self.n_embd, dim=2)
        # (B, nh, T, hd)
        q = q.view(B, T, self.n_head, C // self.n_head).transpose(1, 2)
        k = k.view(B, T, self.n_head, C // self.n_head).transpose(1, 2)
        v = v.view(B, T, self.n_head, C // self.n_head).transpose(1, 2)
        y = F.scaled_dot_product_attention(
            q, k, v, is_causal=True,
            dropout_p=self.attn_dropout if self.training else 0.0,
        )
        y = y.transpose(1, 2).contiguous().view(B, T, C)
        return self.resid_dropout(self.c_proj(y))


class MLP(nn.Module):
    def __init__(self, cfg: GPTConfig):
        super().__init__()
        self.c_fc = nn.Linear(cfg.n_embd, 4 * cfg.n_embd)
        self.c_proj = nn.Linear(4 * cfg.n_embd, cfg.n_embd)
        self.dropout = nn.Dropout(cfg.dropout)

    def forward(self, x):
        return self.dropout(self.c_proj(F.gelu(self.c_fc(x))))


class Block(nn.Module):
    def __init__(self, cfg: GPTConfig):
        super().__init__()
        self.ln_1 = nn.LayerNorm(cfg.n_embd)
        self.attn = CausalSelfAttention(cfg)
        self.ln_2 = nn.LayerNorm(cfg.n_embd)
        self.mlp = MLP(cfg)

    def forward(self, x):
        x = x + self.attn(self.ln_1(x))
        x = x + self.mlp(self.ln_2(x))
        return x


class GPT(nn.Module):
    def __init__(self, cfg: GPTConfig):
        super().__init__()
        self.cfg = cfg
        self.tok_emb = nn.Embedding(cfg.vocab_size, cfg.n_embd)
        self.pos_emb = nn.Embedding(cfg.block_size, cfg.n_embd)
        self.drop = nn.Dropout(cfg.dropout)
        self.blocks = nn.ModuleList([Block(cfg) for _ in range(cfg.n_layer)])
        self.ln_f = nn.LayerNorm(cfg.n_embd)
        self.lm_head = nn.Linear(cfg.n_embd, cfg.vocab_size, bias=False)
        self.tok_emb.weight = self.lm_head.weight  # weight tying
        self.apply(self._init_weights)

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

    def num_params(self) -> int:
        # subtract tied head (shares tok_emb storage) to avoid double counting
        return sum(p.numel() for p in self.parameters()) - self.lm_head.weight.numel()

    def forward(self, idx, targets=None):
        B, T = idx.shape
        pos = torch.arange(T, device=idx.device)
        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.lm_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 next_token_logits(self, idx):
        """Logits for the position after the (cropped) context — for the playground."""
        idx = idx[:, -self.cfg.block_size:]
        logits, _ = self(idx)
        return logits[:, -1, :]

    @torch.no_grad()
    def generate(self, idx, max_new_tokens, temperature=1.0, top_k=None):
        for _ in range(max_new_tokens):
            logits = self.next_token_logits(idx) / max(temperature, 1e-6)
            if top_k is not None:
                v, _ = torch.topk(logits, min(top_k, logits.size(-1)))
                logits[logits < v[:, [-1]]] = -float("inf")
            probs = F.softmax(logits, dim=-1)
            nxt = torch.multinomial(probs, num_samples=1)
            idx = torch.cat([idx, nxt], dim=1)
        return idx