File size: 7,929 Bytes
95a9cb5 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 | from __future__ import annotations
import math
from typing import Optional, Tuple, Union
import torch
import torch.nn as nn
import torch.nn.functional as F
from transformers import AutoConfig, AutoModelForCausalLM, PretrainedConfig, PreTrainedModel
from transformers.modeling_outputs import CausalLMOutputWithPast
class ChessConfig(PretrainedConfig):
model_type = "chess_square_transformer"
def __init__(
self,
vocab_size: int = 72,
n_embd: int = 128,
n_layer: int = 5,
n_head: int = 4,
n_ctx: int = 256,
n_inner: Optional[int] = None,
dropout: float = 0.1,
layer_norm_epsilon: float = 1e-5,
tie_weights: bool = True,
pad_token_id: int = 0,
bos_token_id: int = 1,
eos_token_id: int = 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 3 * n_embd
self.dropout = dropout
self.layer_norm_epsilon = layer_norm_epsilon
self.tie_weights = tie_weights
self.tie_word_embeddings = bool(tie_weights)
class MultiHeadAttention(nn.Module):
def __init__(self, config: ChessConfig):
super().__init__()
if config.n_embd % config.n_head != 0:
raise ValueError("n_embd must be divisible by n_head")
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: torch.Tensor, attention_mask: Optional[torch.Tensor] = None) -> torch.Tensor:
bsz, seq_len, _ = x.size()
qkv = self.c_attn(x)
q, k, v = qkv.split(self.n_embd, dim=2)
q = q.view(bsz, seq_len, self.n_head, self.head_dim).transpose(1, 2)
k = k.view(bsz, seq_len, self.n_head, self.head_dim).transpose(1, 2)
v = v.view(bsz, seq_len, self.n_head, self.head_dim).transpose(1, 2)
attn = torch.matmul(q, k.transpose(-2, -1)) / math.sqrt(self.head_dim)
causal_mask = self.bias[:, :, :seq_len, :seq_len]
attn = attn.masked_fill(causal_mask == 0, float("-inf"))
if attention_mask is not None:
attention_mask = attention_mask.unsqueeze(1).unsqueeze(2)
attn = attn.masked_fill(attention_mask == 0, float("-inf"))
attn = F.softmax(attn, dim=-1)
attn = self.dropout(attn)
y = torch.matmul(attn, v)
y = y.transpose(1, 2).contiguous().view(bsz, seq_len, self.n_embd)
y = self.c_proj(y)
return y
class FeedForward(nn.Module):
def __init__(self, config: ChessConfig):
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: torch.Tensor) -> torch.Tensor:
x = self.c_fc(x)
x = F.gelu(x)
x = self.c_proj(x)
x = self.dropout(x)
return x
class TransformerBlock(nn.Module):
def __init__(self, config: ChessConfig):
super().__init__()
self.ln_1 = nn.LayerNorm(config.n_embd, eps=config.layer_norm_epsilon)
self.attn = MultiHeadAttention(config)
self.ln_2 = nn.LayerNorm(config.n_embd, eps=config.layer_norm_epsilon)
self.mlp = FeedForward(config)
def forward(self, x: torch.Tensor, attention_mask: Optional[torch.Tensor] = None) -> torch.Tensor:
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
base_model_prefix = "transformer"
supports_gradient_checkpointing = True
keys_to_ignore_on_load_missing = ["lm_head.weight"]
def __init__(self, config: ChessConfig):
super().__init__(config)
self.wte = nn.Embedding(config.vocab_size, config.n_embd)
self.wpe = nn.Embedding(config.n_ctx, 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, eps=config.layer_norm_epsilon)
self.lm_head = nn.Linear(config.n_embd, config.vocab_size, bias=False)
if config.tie_weights:
self._tied_weights_keys = ["lm_head.weight"]
self.post_init()
if config.tie_weights:
self.tie_weights()
def get_input_embeddings(self) -> nn.Module:
return self.wte
def set_input_embeddings(self, new_embeddings: nn.Module):
self.wte = new_embeddings
if getattr(self.config, "tie_weights", False):
self.tie_weights()
def get_output_embeddings(self) -> nn.Module:
return self.lm_head
def set_output_embeddings(self, new_embeddings: nn.Module):
self.lm_head = new_embeddings
def tie_weights(self):
if getattr(self.config, "tie_weights", False) or getattr(self.config, "tie_word_embeddings", False):
self._tie_or_clone_weights(self.lm_head, self.wte)
def _init_weights(self, module: nn.Module):
if isinstance(module, nn.Linear):
torch.nn.init.normal_(module.weight, mean=0.0, std=0.02)
if module.bias is not None:
torch.nn.init.zeros_(module.bias)
elif isinstance(module, nn.Embedding):
torch.nn.init.normal_(module.weight, mean=0.0, std=0.02)
elif isinstance(module, nn.LayerNorm):
torch.nn.init.ones_(module.weight)
torch.nn.init.zeros_(module.bias)
def forward(
self,
input_ids: torch.LongTensor,
attention_mask: Optional[torch.Tensor] = None,
position_ids: Optional[torch.LongTensor] = None,
labels: Optional[torch.LongTensor] = None,
return_dict: Optional[bool] = None,
**kwargs,
) -> Union[Tuple, CausalLMOutputWithPast]:
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
bsz, seq_len = input_ids.size()
device = input_ids.device
if position_ids is None:
position_ids = torch.arange(seq_len, device=device).unsqueeze(0).expand(bsz, -1)
tok = self.wte(input_ids)
pos = self.wpe(position_ids)
x = self.drop(tok + pos)
for block in self.h:
x = block(x, attention_mask=attention_mask)
x = self.ln_f(x)
logits = self.lm_head(x)
loss = None
if labels is not None:
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))
if not return_dict:
output = (logits,)
return ((loss,) + output) if loss is not None else output
return CausalLMOutputWithPast(
loss=loss,
logits=logits,
past_key_values=None,
hidden_states=None,
attentions=None,
)
AutoConfig.register("chess_square_transformer", ChessConfig)
AutoModelForCausalLM.register(ChessConfig, ChessForCausalLM)
|