import torch
import torch.nn as nn
import torch.nn.functional as F
from transformers import PreTrainedModel
from transformers.modeling_outputs import CausalLMOutputWithPast
from .configuration_dumbmath import DumbMathConfig

# ---------------------------------------------------------
# あなたのコンポーネント定義 (そのまま使用)
# ---------------------------------------------------------
class RMSNorm(nn.Module):
    def __init__(self, dim, eps=1e-6):
        super().__init__()
        self.eps = eps
        self.weight = nn.Parameter(torch.ones(dim))

    def forward(self, x):
        variance = x.pow(2).mean(-1, keepdim=True)
        return x * torch.rsqrt(variance + self.eps) * self.weight

class SwiGLUMLP(nn.Module):
    def __init__(self, d_model, d_ff):
        super().__init__()
        self.w_gate = nn.Linear(d_model, d_ff, bias=False)
        self.w_up = nn.Linear(d_model, d_ff, bias=False)
        self.w_down = nn.Linear(d_ff, d_model, bias=False)

    def forward(self, x):
        return self.w_down(F.silu(self.w_gate(x)) * self.w_up(x))

class MultiHeadAttention(nn.Module):
    def __init__(self, d_model, n_heads):
        super().__init__()
        self.n_heads = n_heads
        self.d_model = d_model
        self.head_dim = d_model // n_heads

        self.q_proj = nn.Linear(d_model, d_model, bias=False)
        self.k_proj = nn.Linear(d_model, d_model, bias=False)
        self.v_proj = nn.Linear(d_model, d_model, bias=False)
        self.out_proj = nn.Linear(d_model, d_model, bias=False)

    def forward(self, x):
        b, t, c = x.size()
        q = self.q_proj(x).view(b, t, self.n_heads, self.head_dim).transpose(1, 2)
        k = self.k_proj(x).view(b, t, self.n_heads, self.head_dim).transpose(1, 2)
        v = self.v_proj(x).view(b, t, self.n_heads, self.head_dim).transpose(1, 2)

        attn_out = F.scaled_dot_product_attention(q, k, v, attn_mask=None, is_causal=True)
        attn_out = attn_out.transpose(1, 2).contiguous().view(b, t, c)
        return self.out_proj(attn_out)

class DecoderBlock(nn.Module):
    def __init__(self, d_model, n_heads, d_ff):
        super().__init__()
        self.attn_norm = RMSNorm(d_model)
        self.attn = MultiHeadAttention(d_model, n_heads)
        self.ffn_norm = RMSNorm(d_model)
        self.ffn = SwiGLUMLP(d_model, d_ff)

    def forward(self, x):
        x = x + self.attn(self.attn_norm(x))
        x = x + self.ffn(self.ffn_norm(x))
        return x

# ---------------------------------------------------------
# Hugging Face規格のラッパークラス
# ---------------------------------------------------------
class DumbMathForCausalLM(PreTrainedModel):
    config_class = DumbMathConfig
    base_model_prefix = "model"

    def __init__(self, config):
        super().__init__(config)
        self.token_embedding = nn.Embedding(config.vocab_size, config.d_model)
        self.pos_embedding = nn.Parameter(torch.zeros(1, config.max_len, config.d_model))

        self.layers = nn.ModuleList([
            DecoderBlock(config.d_model, config.n_heads, config.d_ff) for _ in range(config.n_layers)
        ])

        self.ln_f = RMSNorm(config.d_model)
        self.head = nn.Linear(config.d_model, config.vocab_size, bias=False)

        # 重み初期化の適用
        self.post_init()

    def get_input_embeddings(self):
        return self.token_embedding

    def set_input_embeddings(self, value):
        self.token_embedding = value

    def forward(self, input_ids, labels=None, attention_mask=None, **kwargs):
        b, t = input_ids.size()
        x = self.token_embedding(input_ids) + self.pos_embedding[:, :t, :]

        for layer in self.layers:
            x = layer(x)

        x = self.ln_f(x)
        logits = self.head(x)

        # ロス計算（Trainerに対応）
        loss = None
        if labels is not None:
            # 標準的なCausal LMのロス計算（右シフト処理はTrainerが自動実行またはここで対応）
            shift_logits = logits[..., :-1, :].contiguous()
            shift_labels = labels[..., 1:].contiguous()
            loss_fct = nn.CrossEntropyLoss(ignore_index=-100)
            loss = loss_fct(shift_logits.view(-1, shift_logits.size(-1)), shift_labels.view(-1))

        return CausalLMOutputWithPast(
            loss=loss,
            logits=logits
        )

    # 推論時(model.generate)に必要なメソッド定義
    def prepare_inputs_for_generation(self, input_ids, **kwargs):
        return {"input_ids": input_ids}