File size: 16,689 Bytes
70cf84c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
"""
Chess Transformer Model for the Chess Challenge.

This module provides a simple GPT-style transformer architecture
designed to fit within the 1M parameter constraint.

Key components:
- ChessConfig: Configuration class for model hyperparameters
- ChessForCausalLM: The main model class for next-move prediction
"""

from __future__ import annotations

from pprint import pformat
from typing import Optional, Tuple, Union

import torch
import torch.nn as nn
import torch.nn.functional as F
from transformers.configuration_utils import PretrainedConfig
from transformers.modeling_outputs import CausalLMOutputWithPast
from transformers.modeling_utils import PreTrainedModel


class ChessConfig(PretrainedConfig):
    """
    Configuration class for the Chess Transformer model.

    This configuration is designed for a ~1M parameter model.
    Students can adjust these values to explore different architectures.

    Parameter budget breakdown (with default values):
    - Embeddings (vocab): 1200 x 128 = 153,600
    - Position Embeddings: 256 x 128 = 32,768
    - Transformer Layers: 6 x ~120,000 = ~720,000
    - LM Head (with weight tying): 0 (shared with embeddings)
    - Total: ~906,000 parameters

    Attributes:
        vocab_size: Size of the vocabulary (number of unique moves).
        n_embd: Embedding dimension (d_model).
        n_layer: Number of transformer layers.
        n_head: Number of attention heads.
        n_ctx: Maximum sequence length (context window).
        n_inner: Feed-forward inner dimension (default: 3 * n_embd).
        dropout: Dropout probability.
        layer_norm_epsilon: Epsilon for layer normalization.
        tie_weights: Whether to tie embedding and output weights.
    """

    model_type = "chess_transformer"

    def __init__(
        self,
        vocab_size: int = 1200,
        n_embd: int = 256,
        n_layer: int = 10,
        n_head_kv: int = 8,
        n_head_q_per_kv: int = 2,
        dim_head_qk: int = 32,
        dim_head_v: Optional[int] = None,
        n_ctx: int = 1024,
        n_inner: Optional[int] = None,
        dropout: float = 0.1,
        layer_norm_epsilon: float = 1e-5,
        tie_weights: bool = True,
        rope_theta: float = 1e4,
        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.dim_head_qk = dim_head_qk
        self.dim_head_v = dim_head_v or dim_head_qk

        self.n_head_kv = n_head_kv
        self.n_head_q_per_kv = n_head_q_per_kv

        self.vocab_size = vocab_size
        self.n_embd = n_embd
        self.n_layer = n_layer
        self.n_head_kv = n_head_kv
        self.n_ctx = n_ctx
        self.n_inner = n_inner if n_inner is not None else 3 * n_embd  # Reduced from 4x to 3x
        self.dropout = dropout
        self.layer_norm_epsilon = layer_norm_epsilon
        self.tie_weights = tie_weights
        self.rope_theta = rope_theta
        # Inform HF base class about tying behavior
        self.tie_word_embeddings = bool(tie_weights)

    @property
    def dim_q(self):
        return self.n_head_q * self.dim_head_qk

    @property
    def dim_k(self):
        return self.n_head_kv * self.dim_head_qk

    @property
    def dim_v(self):
        return self.n_head_kv * self.dim_head_v

    @property
    def n_head_q(self):
        return self.n_head_q_per_kv * self.n_head_kv

    def __repr__(self):
        cls = self.__class__.__name__
        fields = self.to_dict()
        return f"{cls}(\n{pformat(fields, indent=2)}\n)"

    __str__ = __repr__


def apply_rotary_emb(x: torch.Tensor, freqs_cis: torch.Tensor) -> torch.Tensor:
    """Applies rotary embeddings to input tensor x."""
    # Reshape x to complex numbers
    x_complex = torch.view_as_complex(x.float().reshape(*x.shape[:-1], -1, 2))
    freqs_cis = freqs_cis.view(1, x.size(1), 1, -1)
    # Perform rotation in complex space
    x_rotated = torch.view_as_real(x_complex * freqs_cis).flatten(3)
    return x_rotated.type_as(x)


class MultiHeadAttention(nn.Module):
    """
    Multi-head self-attention module.

    This is a standard scaled dot-product attention implementation
    with causal masking for autoregressive generation.
    """

    bias: torch.Tensor  # to restrict type to Tensor and not Module

    def __init__(self, config: ChessConfig):
        super().__init__()

        self._config = config

        self.proj_q = nn.Linear(config.n_embd, self.dim_q)
        self.proj_k = nn.Linear(config.n_embd, self.dim_k)
        self.proj_v = nn.Linear(config.n_embd, self.dim_v)

        self.proj_o = nn.Linear(self._n_head_q * self._dim_head_v, config.n_embd)

        # Causal mask (will be created on first forward pass)
        self.register_buffer(
            "bias",
            torch.ones(config.n_ctx, config.n_ctx, dtype=torch.bool)
            .tril(diagonal=0)
            .unsqueeze(0)
            .unsqueeze(0),
            persistent=False,
        )

    @property
    def dim_q(self):
        return self._config.dim_q

    @property
    def dim_k(self):
        return self._config.dim_k

    @property
    def dim_v(self):
        return self._config.dim_v

    @property
    def enable_gqa(self):
        return self._n_head_q_per_kv > 1

    @property
    def dropout_p(self):
        return self._config.dropout * self.training

    @property
    def _n_head_kv(self):
        return self._config.n_head_kv

    @property
    def _n_head_q(self):
        return self._config.n_head_q

    @property
    def _dim_head_qk(self):
        return self._config.dim_head_qk

    @property
    def _dim_head_v(self):
        return self._config.dim_head_v

    @property
    def _n_head_q_per_kv(self):
        return self._config.n_head_q_per_kv

    def forward(
        self,
        x: torch.Tensor,
        freqs_cis: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
    ) -> torch.Tensor:
        batch_size, seq_len, _ = x.size()

        # Compute Q, K, V
        q, k, v = (proj(x) for proj in (self.proj_q, self.proj_k, self.proj_v))

        # Reshape for multi-head attention
        q = q.unflatten(-1, (self._n_head_q, self._dim_head_qk))
        k = k.unflatten(-1, (self._n_head_kv, self._dim_head_qk))
        v = v.unflatten(-1, (self._n_head_kv, self._dim_head_v))

        q, k = (apply_rotary_emb(x, freqs_cis) for x in (q, k))

        q, k, v = (x.transpose(1, 2) for x in (q, k, v))

        attn_mask = self.bias[..., :seq_len, :seq_len]

        # merge causal mask with attention mask if provided
        if attention_mask is not None:
            attention_mask = (
                attention_mask.view(batch_size, 1, 1, seq_len)
                .expand(-1, -1, seq_len, -1)
                .to(torch.bool)
            )
            attn_mask = torch.logical_or(attention_mask, attn_mask)

        attn_output = (
            F.scaled_dot_product_attention(
                query=q,
                key=k,
                value=v,
                attn_mask=attn_mask,
                dropout_p=self.dropout_p,
                enable_gqa=self.enable_gqa,
            )
            .transpose(1, 2)
            .flatten(2)
        )

        return self.proj_o(attn_output)


class FeedForward(nn.Module):
    """
    Feed-forward network (MLP) module.

    Standard two-layer MLP with GELU activation.
    """

    def __init__(self, config: ChessConfig):
        super().__init__()

        self.proj_up = nn.Linear(config.n_embd, config.n_inner)
        self.proj_down = nn.Linear(config.n_inner, config.n_embd)
        self.dropout = nn.Dropout(config.dropout)

    def forward(self, x: torch.Tensor) -> torch.Tensor:
        x = self.proj_up(x)
        x = F.gelu(x)
        x = self.proj_down(x)
        x = self.dropout(x)
        return x


class TransformerBlock(nn.Module):
    """
    A single transformer block with attention and feed-forward layers.

    Uses pre-normalization (LayerNorm before attention/FFN) for better
    training stability.
    """

    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,
        freqs_cis: torch.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
    ) -> torch.Tensor:
        # Pre-norm attention
        x = x + self.attn(self.ln_1(x), freqs_cis=freqs_cis, attention_mask=attention_mask)
        # Pre-norm FFN
        x = x + self.mlp(self.ln_2(x))
        return x


class ChessForCausalLM(PreTrainedModel):
    """
    Chess Transformer for Causal Language Modeling (next-move prediction).

    This model is designed to predict the next chess move given a sequence
    of previous moves. It uses a GPT-style architecture with:
    - Token embeddings for chess moves
    - Learned positional embeddings
    - Stacked transformer blocks
    - Linear head for next-token prediction

    The model supports weight tying between the embedding layer and the
    output projection to save parameters.

    Example:
        >>> config = ChessConfig(vocab_size=1200, n_embd=128, n_layer=6)
        >>> model = ChessForCausalLM(config)
        >>> inputs = {"input_ids": torch.tensor([[1, 42, 87]])}
        >>> outputs = model(**inputs)
        >>> next_move_logits = outputs.logits[:, -1, :]
    """

    config_class = ChessConfig
    base_model_prefix = "transformer"
    supports_gradient_checkpointing = True
    # Suppress missing-key warning for tied lm_head when loading
    keys_to_ignore_on_load_missing = ["lm_head.weight"]
    freqs_cis: torch.Tensor

    def __init__(self, config: ChessConfig):
        super().__init__(config)

        # Token and position embeddings
        self.wte = nn.Embedding(config.vocab_size, config.n_embd)

        self.drop = nn.Dropout(config.dropout)

        # Transformer blocks
        self.h = nn.ModuleList([TransformerBlock(config) for _ in range(config.n_layer)])

        # Final layer norm
        self.ln_f = nn.LayerNorm(config.n_embd, eps=config.layer_norm_epsilon)

        # Output head
        self.lm_head = nn.Linear(config.n_embd, config.vocab_size, bias=False)

        freqs_cis = self._precompute_freqs_cis(config.dim_head_qk, config.n_ctx, config.rope_theta)
        self.register_buffer("freqs_cis", freqs_cis, persistent=False)

        # Declare tied weights for proper serialization
        if config.tie_weights:
            self._tied_weights_keys = ["lm_head.weight"]

        # Initialize weights
        self.post_init()

        # Tie weights if configured
        if config.tie_weights:
            self.tie_weights()

    def _precompute_freqs_cis(self, dim: int, end: int, theta: float):
        freqs = 1.0 / (theta ** (torch.arange(0, dim, 2)[: (dim // 2)].float() / dim))
        t = torch.arange(end)
        freqs = torch.outer(t, freqs).float()
        return torch.polar(torch.ones_like(freqs), freqs)

    def get_input_embeddings(self) -> nn.Module:
        return self.wte

    def set_input_embeddings(self, value: nn.Module):
        self.wte = value
        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):
        # Use HF helper to tie or clone depending on config
        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):
        """Initialize weights following GPT-2 style."""
        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.Tensor,
        attention_mask: Optional[torch.Tensor] = None,
        labels: Optional[torch.LongTensor] = None,
        return_dict: Optional[bool] = None,
        **kwargs,
    ) -> Union[Tuple, CausalLMOutputWithPast]:
        """
        Forward pass of the model.

        Args:
            input_ids: Token IDs of shape (batch_size, seq_len).
            attention_mask: Attention mask of shape (batch_size, seq_len).
            labels: Labels for language modeling loss.
            return_dict: Whether to return a ModelOutput object.

        Returns:
            CausalLMOutputWithPast containing loss (if labels provided) and logits.
        """
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        batch_size, seq_len = input_ids.size()

        # Get embeddings
        hidden_states = self.drop(self.wte(input_ids))

        freqs_cis = self.freqs_cis[:seq_len]

        # Pass through transformer blocks
        for block in self.h:
            hidden_states = block(hidden_states, freqs_cis=freqs_cis, attention_mask=attention_mask)

        # Final layer norm
        hidden_states = self.ln_f(hidden_states)

        # Get logits
        logits = self.lm_head(hidden_states)

        # Compute loss if labels are provided
        loss = None
        if labels is not None:
            # Shift logits and labels for next-token prediction
            shift_logits = logits[..., :-1, :].contiguous()
            shift_labels = labels[..., 1:].contiguous()

            # Flatten for cross-entropy
            ignore_index = self.config.pad_token_id or -100
            loss_fct = nn.CrossEntropyLoss(ignore_index=ignore_index)
            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,
        )

    @torch.no_grad()
    def generate_move(
        self,
        input_ids: torch.LongTensor,
        temperature: float = 1.0,
        top_k: Optional[int] = None,
        top_p: Optional[float] = None,
    ) -> int:
        """
        Generate the next move given a sequence of moves.

        Args:
            input_ids: Token IDs of shape (1, seq_len).
            temperature: Sampling temperature (1.0 = no change).
            top_k: If set, only sample from top k tokens.
            top_p: If set, use nucleus sampling with this threshold.

        Returns:
            The token ID of the predicted next move.
        """
        self.eval()

        # Get logits for the last position
        outputs = self(input_ids)
        logits = outputs.logits[:, -1, :] / temperature

        # Apply top-k filtering
        if top_k is not None:
            indices_to_remove = logits < torch.topk(logits, top_k)[0][..., -1, None]
            logits[indices_to_remove] = float("-inf")

        # Apply top-p (nucleus) filtering
        if top_p is not None:
            sorted_logits, sorted_indices = torch.sort(logits, descending=True)
            cumulative_probs = torch.cumsum(F.softmax(sorted_logits, dim=-1), dim=-1)

            # Remove tokens with cumulative probability above the threshold
            sorted_indices_to_remove = cumulative_probs > top_p
            sorted_indices_to_remove[..., 1:] = sorted_indices_to_remove[..., :-1].clone()
            sorted_indices_to_remove[..., 0] = 0

            indices_to_remove = sorted_indices_to_remove.scatter(
                dim=-1, index=sorted_indices, src=sorted_indices_to_remove
            )
            logits[indices_to_remove] = float("-inf")

        # Sample from the distribution
        probs = F.softmax(logits, dim=-1)
        next_token = torch.multinomial(probs, num_samples=1)
        return int(next_token.item())