model.py

from __future__ import annotations
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

class ChessConfig(PretrainedConfig):
    model_type = "chess_transformer"
    def __init__(self, vocab_size=75, n_embd=96, n_layer=8, n_head=4, n_ctx=512, n_inner=None, dropout=0.1, tie_weights=True, **kwargs):
        super().__init__(**kwargs)
        self.vocab_size = vocab_size
        self.n_embd = n_embd       # 96 : Largeur confortable
        self.n_layer = n_layer     # 8 couches
        self.n_head = n_head
        self.n_ctx = n_ctx         # 512 : Contexte doublé pour les coordonnées
        self.n_inner = n_inner if n_inner is not None else 3 * n_embd
        self.dropout = dropout
        self.tie_weights = tie_weights
        self.tie_word_embeddings = tie_weights

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), persistent=False)
    def forward(self, x, mask=None):
        B, T, C = x.size()
        q, k, v = self.c_attn(x).split(self.n_embd, dim=2)
        k = k.view(B, T, self.n_head, self.head_dim).transpose(1, 2)
        q = q.view(B, T, self.n_head, self.head_dim).transpose(1, 2)
        v = v.view(B, T, self.n_head, self.head_dim).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'))
        if mask is not None: att = att.masked_fill(mask.unsqueeze(1).unsqueeze(2) == 0, float('-inf'))
        y = self.dropout(F.softmax(att, dim=-1)) @ v
        return self.c_proj(y.transpose(1, 2).contiguous().view(B, T, C))

class TransformerBlock(nn.Module):
    def __init__(self, config):
        super().__init__()
        self.ln1 = nn.LayerNorm(config.n_embd); self.attn = MultiHeadAttention(config)
        self.ln2 = nn.LayerNorm(config.n_embd); self.mlp = nn.Sequential(nn.Linear(config.n_embd, config.n_inner), nn.GELU(), nn.Linear(config.n_inner, config.n_embd), nn.Dropout(config.dropout))
    def forward(self, x, mask=None): return x + self.mlp(self.ln2(x + self.attn(self.ln1(x), mask)))

class ChessForCausalLM(PreTrainedModel):
    config_class = ChessConfig
    keys_to_ignore_on_load_missing = ["lm_head.weight"]
    def __init__(self, config):
        super().__init__(config)
        self.wte = nn.Embedding(config.vocab_size, config.n_embd)
        self.wpe = nn.Embedding(config.n_ctx, config.n_embd)
        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)
        if config.tie_weights: self.lm_head.weight = self.wte.weight
        self.post_init()
    def forward(self, input_ids, attention_mask=None, labels=None, **kwargs):
        x = self.wte(input_ids) + self.wpe(torch.arange(input_ids.size(1), device=input_ids.device))
        for b in self.h: x = b(x, attention_mask)
        logits = self.lm_head(self.ln_f(x))
        loss = None
        if labels is not None: loss = F.cross_entropy(logits.view(-1, logits.size(-1)), labels.view(-1), ignore_index=-100)
        return CausalLMOutputWithPast(loss=loss, logits=logits)