"""TinyGPT — a compact nanoGPT-style decoder-only transformer.

Deliberately small and dependency-light (pure PyTorch) so it runs on a free CPU,
while remaining a *real* autoregressive language model. The same module builds the
global model on the parameter server and the local replica on every worker.
"""

from __future__ import annotations

import math
from typing import Optional, Tuple

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

from swarm.config import ModelConfig


class CausalSelfAttention(nn.Module):
    def __init__(self, cfg: ModelConfig):
        super().__init__()
        assert cfg.n_embd % cfg.n_head == 0, "n_embd must be divisible by n_head"
        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 = nn.Dropout(cfg.dropout)
        self.resid_dropout = nn.Dropout(cfg.dropout)
        self.dropout = cfg.dropout
        # Causal mask: (1, 1, T, T) lower-triangular.
        mask = torch.tril(torch.ones(cfg.block_size, cfg.block_size))
        self.register_buffer("bias", mask.view(1, 1, cfg.block_size, cfg.block_size))

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        B, T, C = x.size()
        q, k, v = self.c_attn(x).split(self.n_embd, dim=2)
        # (B, nh, T, hs)
        k = k.view(B, T, self.n_head, C // self.n_head).transpose(1, 2)
        q = q.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)
        att = (q @ k.transpose(-2, -1)) * (1.0 / math.sqrt(k.size(-1)))
        att = att.masked_fill(self.bias[:, :, :T, :T] == 0, float("-inf"))
        att = F.softmax(att, dim=-1)
        att = self.attn_dropout(att)
        y = att @ v  # (B, nh, T, hs)
        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: ModelConfig):
        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: torch.Tensor) -> torch.Tensor:
        return self.dropout(self.c_proj(F.gelu(self.c_fc(x))))


class Block(nn.Module):
    def __init__(self, cfg: ModelConfig):
        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: torch.Tensor) -> torch.Tensor:
        x = x + self.attn(self.ln_1(x))
        x = x + self.mlp(self.ln_2(x))
        return x


class TinyGPT(nn.Module):
    def __init__(self, cfg: ModelConfig):
        super().__init__()
        assert cfg.vocab_size > 0, "ModelConfig.vocab_size must be set (use cfg.with_vocab)"
        self.cfg = cfg
        self.transformer = nn.ModuleDict(
            dict(
                wte=nn.Embedding(cfg.vocab_size, cfg.n_embd),
                wpe=nn.Embedding(cfg.block_size, cfg.n_embd),
                drop=nn.Dropout(cfg.dropout),
                h=nn.ModuleList([Block(cfg) for _ in range(cfg.n_layer)]),
                ln_f=nn.LayerNorm(cfg.n_embd),
            )
        )
        self.lm_head = nn.Linear(cfg.n_embd, cfg.vocab_size, bias=False)
        # Weight tying (standard for GPT).
        self.transformer.wte.weight = self.lm_head.weight
        self.apply(self._init_weights)

    @staticmethod
    def _init_weights(module: nn.Module) -> None:
        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 num_params(self) -> int:
        # Subtract position embedding once (wte is tied to lm_head, counted once).
        return sum(p.numel() for p in self.parameters())

    def forward(
        self, idx: torch.Tensor, targets: Optional[torch.Tensor] = None
    ) -> Tuple[torch.Tensor, Optional[torch.Tensor]]:
        B, T = idx.size()
        assert T <= self.cfg.block_size, f"sequence length {T} > block_size {self.cfg.block_size}"
        pos = torch.arange(0, T, dtype=torch.long, device=idx.device)
        tok_emb = self.transformer.wte(idx)
        pos_emb = self.transformer.wpe(pos)
        x = self.transformer.drop(tok_emb + pos_emb)
        for block in self.transformer.h:
            x = block(x)
        x = self.transformer.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 generate(
        self, idx: torch.Tensor, max_new_tokens: int, temperature: float = 1.0
    ) -> torch.Tensor:
        for _ in range(max_new_tokens):
            idx_cond = idx[:, -self.cfg.block_size :]
            logits, _ = self(idx_cond)
            logits = logits[:, -1, :] / max(temperature, 1e-8)
            probs = F.softmax(logits, dim=-1)
            idx_next = torch.multinomial(probs, num_samples=1)
            idx = torch.cat((idx, idx_next), dim=1)
        return idx


def build_model(cfg: ModelConfig) -> TinyGPT:
    """Construct a TinyGPT and put it in train mode."""
    model = TinyGPT(cfg)
    model.train()
    return model