import math
import torch
import torch.nn as nn
import torch.nn.functional as F
from transformers import PretrainedConfig, PreTrainedModel
from transformers.modeling_outputs import CausalLMOutputWithPast

def rotate_half(x):
    x1 = x[..., : x.shape[-1] // 2]
    x2 = x[..., x.shape[-1] // 2 :]
    return torch.cat((-x2, x1), dim=-1)

def apply_rope(q, k):
    dim = q.shape[-1]
    device = q.device
    seq_len = q.shape[-2]
    theta = 1.0 / (10000 ** (torch.arange(0, dim, 2, device=device).float() / dim))
    pos = torch.arange(seq_len, device=device).float()
    freqs = torch.einsum('i,j->ij', pos, theta)
    emb = torch.cat((freqs, freqs), dim=-1)
    cos = emb.cos()[None, None, :, :]
    sin = emb.sin()[None, None, :, :]
    q = (q * cos) + (rotate_half(q) * sin)
    k = (k * cos) + (rotate_half(k) * sin)
    return q, k

class ChessConfig(PretrainedConfig):
    model_type = "chess_transformer"
    def __init__(
        self,
        vocab_size=1682,
        n_embd=96,
        n_layer=8,
        n_head=8,
        n_ctx=336,
        n_inner=None,
        dropout=0.15,
        layer_norm_epsilon=1e-5,
        tie_weights=True,
        pad_token_id=0,
        bos_token_id=1,
        eos_token_id=2,
        **kwargs,
    ):
        super().__init__(
            pad_token_id=pad_token_id,
            bos_token_id=bos_token_id,
            eos_token_id=eos_token_id,
            **kwargs,
        )
        self.vocab_size = vocab_size
        self.n_embd = n_embd
        self.n_layer = n_layer
        self.n_head = n_head
        self.n_ctx = n_ctx
        self.n_inner = n_inner if n_inner is not None else int(3.5 * n_embd)
        self.dropout = dropout
        self.layer_norm_epsilon = layer_norm_epsilon
        self.tie_weights = tie_weights
        self.tie_word_embeddings = True

class MultiHeadAttention(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.n_head = config.n_head
        self.n_embd = config.n_embd
        self.head_dim = config.n_embd // config.n_head
        self.c_attn = nn.Linear(config.n_embd, 3 * config.n_embd)
        self.c_proj = nn.Linear(config.n_embd, config.n_embd)
        self.dropout = nn.Dropout(config.dropout)
        self.register_buffer("bias", torch.tril(torch.ones(config.n_ctx, config.n_ctx)).view(1, 1, config.n_ctx, config.n_ctx))

    def forward(self, x, attention_mask=None):
        batch_size, seq_len, _ = x.size()
        qkv = self.c_attn(x)
        q, k, v = qkv.split(self.n_embd, dim=2)
        q = q.view(batch_size, seq_len, self.n_head, self.head_dim).transpose(1, 2)
        k = k.view(batch_size, seq_len, self.n_head, self.head_dim).transpose(1, 2)
        v = v.view(batch_size, seq_len, self.n_head, self.head_dim).transpose(1, 2)
        q, k = apply_rope(q, k)
        attn_weights = (q @ k.transpose(-2, -1)) / math.sqrt(self.head_dim)
        attn_weights = attn_weights.masked_fill(self.bias[:, :, :seq_len, :seq_len] == 0, float("-inf"))
        if attention_mask is not None:
            attn_weights = attn_weights.masked_fill(attention_mask.view(batch_size, 1, 1, seq_len) == 0, float("-inf"))
        attn_weights = F.softmax(attn_weights, dim=-1)
        return self.c_proj((attn_weights @ v).transpose(1, 2).contiguous().view(batch_size, seq_len, self.n_embd))

class FeedForward(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.c_fc = nn.Linear(config.n_embd, config.n_inner)
        self.c_proj = nn.Linear(config.n_inner, config.n_embd)
        self.dropout = nn.Dropout(config.dropout)
    def forward(self, x):
        return self.dropout(self.c_proj(F.gelu(self.c_fc(x))))

class TransformerBlock(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.ln_1 = nn.LayerNorm(config.n_embd)
        self.attn = MultiHeadAttention(config)
        self.ln_2 = nn.LayerNorm(config.n_embd)
        self.mlp = FeedForward(config)
    def forward(self, x, attention_mask=None):
        x = x + self.attn(self.ln_1(x), attention_mask=attention_mask)
        x = x + self.mlp(self.ln_2(x))
        return x

class ChessForCausalLM(PreTrainedModel):
    config_class = ChessConfig
    def __init__(self, config):
        super().__init__(config)
        self.wte = nn.Embedding(config.vocab_size, config.n_embd)
        self.drop = nn.Dropout(config.dropout)
        self.h = nn.ModuleList([TransformerBlock(config) for _ in range(config.n_layer)])
        self.ln_f = nn.LayerNorm(config.n_embd)
        self.lm_head = nn.Linear(config.n_embd, config.vocab_size, bias=False)
        self.post_init()
        if config.tie_weights: self.lm_head.weight = self.wte.weight

    def forward(self, input_ids, attention_mask=None, labels=None, **kwargs):
        x = self.drop(self.wte(input_ids))
        for block in self.h: x = block(x, attention_mask=attention_mask)
        logits = self.lm_head(self.ln_f(x))
        loss = None
        if labels is not None:
            loss = F.cross_entropy(logits[..., :-1, :].contiguous().view(-1, logits.size(-1)), labels[..., 1:].contiguous().view(-1), ignore_index=-100)
        return CausalLMOutputWithPast(loss=loss, logits=logits)