File size: 18,220 Bytes
a73d8fb |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 |
"""
GeoMotionGPT Model
This module contains the model implementation for GeoMotionGPT, integrating:
1. Motion Tokenizer (DVQ-GSST VQ-VAE)
2. Language Model (fine-tuned GPT-2 for motion-to-text)
Usage:
```python
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("zy22b/GeoMotionGPT", trust_remote_code=True)
motion_tokenizer = model.motion_tokenizer
# Tokenize motion
motion_tokens = motion_tokenizer.encode(motion_features)
# Generate text
text = model.generate_from_motion(motion_tokens)
```
"""
import torch
import torch.nn as nn
import torch.nn.functional as F
from typing import Optional, Tuple, List, Union
from transformers import PreTrainedModel, GPT2LMHeadModel, GPT2Config
from transformers.modeling_outputs import CausalLMOutputWithCrossAttentions
# Handle both package and standalone imports
try:
from .configuration_geomotiongpt import GeoMotionGPTConfig
except ImportError:
from configuration_geomotiongpt import GeoMotionGPTConfig
# =====================================================
# Motion Tokenizer Components (DVQ-GSST)
# =====================================================
class Swish(nn.Module):
"""Swish activation function."""
def forward(self, x):
return x * torch.sigmoid(x)
class ResConv1DBlock(nn.Module):
"""Single residual convolution block."""
def __init__(self, n_in, n_state, dilation=1, activation='relu', norm=None):
super().__init__()
padding = dilation
self.norm = norm
if norm == "LN":
self.norm1 = nn.LayerNorm(n_in)
self.norm2 = nn.LayerNorm(n_in)
elif norm == "GN":
self.norm1 = nn.GroupNorm(num_groups=32, num_channels=n_in, eps=1e-6, affine=True)
self.norm2 = nn.GroupNorm(num_groups=32, num_channels=n_in, eps=1e-6, affine=True)
elif norm == "BN":
self.norm1 = nn.BatchNorm1d(num_features=n_in, eps=1e-6, affine=True)
self.norm2 = nn.BatchNorm1d(num_features=n_in, eps=1e-6, affine=True)
else:
self.norm1 = nn.Identity()
self.norm2 = nn.Identity()
if activation == "relu":
self.activation1 = nn.ReLU()
self.activation2 = nn.ReLU()
elif activation == "silu":
self.activation1 = Swish()
self.activation2 = Swish()
elif activation == "gelu":
self.activation1 = nn.GELU()
self.activation2 = nn.GELU()
self.conv1 = nn.Conv1d(n_in, n_state, 3, 1, padding, dilation)
self.conv2 = nn.Conv1d(n_state, n_in, 1, 1, 0)
def forward(self, x):
x_orig = x
if self.norm == "LN":
x = self.norm1(x.transpose(-2, -1))
x = self.activation1(x.transpose(-2, -1))
else:
x = self.norm1(x)
x = self.activation1(x)
x = self.conv1(x)
if self.norm == "LN":
x = self.norm2(x.transpose(-2, -1))
x = self.activation2(x.transpose(-2, -1))
else:
x = self.norm2(x)
x = self.activation2(x)
x = self.conv2(x)
return x + x_orig
class Resnet1D(nn.Module):
"""1D ResNet block composed of multiple ResConv1DBlocks."""
def __init__(self, n_in, n_depth, dilation_growth_rate=1,
reverse_dilation=True, activation='relu', norm=None):
super().__init__()
blocks = [
ResConv1DBlock(n_in, n_in, dilation=dilation_growth_rate ** depth,
activation=activation, norm=norm)
for depth in range(n_depth)
]
if reverse_dilation:
blocks = blocks[::-1]
self.model = nn.Sequential(*blocks)
def forward(self, x):
return self.model(x)
class MotionEncoder(nn.Module):
"""Encoder for motion features with temporal downsampling."""
def __init__(self, input_dim=263, hidden_dim=512, nb_code=512,
down_t=3, stride_t=2, depth=3, dilation_growth_rate=3,
activation='relu', norm=None):
super().__init__()
blocks = []
filter_t, pad_t = stride_t * 2, stride_t // 2
blocks.append(nn.Conv1d(input_dim, hidden_dim, 3, 1, 1))
blocks.append(nn.ReLU())
for _ in range(down_t):
block = nn.Sequential(
nn.Conv1d(hidden_dim, hidden_dim, filter_t, stride_t, pad_t),
Resnet1D(hidden_dim, depth, dilation_growth_rate,
reverse_dilation=False, activation=activation, norm=norm),
)
blocks.append(block)
blocks.append(nn.Conv1d(hidden_dim, nb_code, 3, 1, 1))
self.model = nn.Sequential(*blocks)
def forward(self, x):
return self.model(x)
class MotionDecoder(nn.Module):
"""Decoder for reconstructing motion from quantized features."""
def __init__(self, output_dim=263, hidden_dim=512, code_dim=512,
down_t=3, stride_t=2, depth=3, dilation_growth_rate=3,
activation='relu', norm=None):
super().__init__()
blocks = []
blocks.append(nn.Conv1d(code_dim, hidden_dim, 3, 1, 1))
blocks.append(nn.ReLU())
for _ in range(down_t):
block = nn.Sequential(
Resnet1D(hidden_dim, depth, dilation_growth_rate,
reverse_dilation=True, activation=activation, norm=norm),
nn.Upsample(scale_factor=2, mode='nearest'),
nn.Conv1d(hidden_dim, hidden_dim, 3, 1, 1)
)
blocks.append(block)
blocks.append(nn.Conv1d(hidden_dim, hidden_dim, 3, 1, 1))
blocks.append(nn.ReLU())
blocks.append(nn.Conv1d(hidden_dim, output_dim, 3, 1, 1))
self.model = nn.Sequential(*blocks)
def forward(self, x):
return self.model(x)
class GumbelSoftmaxQuantizer(nn.Module):
"""Gumbel-Softmax Straight-Through quantizer for VQ-VAE."""
def __init__(self, nb_code=512, code_dim=512):
super().__init__()
self.nb_code = nb_code
self.code_dim = code_dim
self.codebook = nn.Embedding(nb_code, code_dim)
nn.init.uniform_(self.codebook.weight, -1.0 / nb_code, 1.0 / nb_code)
self.tau = 0.4
def quantize(self, x):
"""Quantize encoder output to discrete indices."""
return x.argmax(dim=-1)
def dequantize(self, indices):
"""Convert indices back to embeddings."""
return self.codebook(indices)
def forward(self, x_encoder):
"""Forward pass with Gumbel-Softmax sampling."""
N, C, T = x_encoder.shape
x = x_encoder.permute(0, 2, 1).contiguous().view(-1, C)
# Gumbel-Softmax with straight-through
y_hard_st = F.gumbel_softmax(x, tau=self.tau, hard=True, dim=-1)
x_quantized = torch.matmul(y_hard_st, self.codebook.weight)
return x_quantized.view(N, T, -1).permute(0, 2, 1).contiguous()
class MotionTokenizer(nn.Module):
"""
DVQ-GSST Motion Tokenizer.
Converts continuous motion features (263-dim HumanML3D format) to discrete tokens.
Args:
config: GeoMotionGPTConfig containing motion tokenizer parameters
Example:
```python
motion = torch.randn(1, 100, 263) # (batch, time, features)
tokens = motion_tokenizer.encode(motion) # (batch, time//8)
```
"""
def __init__(self, config: GeoMotionGPTConfig):
super().__init__()
self.config = config
self.encoder = MotionEncoder(
input_dim=config.motion_input_dim,
hidden_dim=config.motion_hidden_dim,
nb_code=config.motion_vocab_size,
down_t=config.motion_down_t,
depth=config.motion_depth,
dilation_growth_rate=config.motion_dilation_growth_rate,
)
self.decoder = MotionDecoder(
output_dim=config.motion_input_dim,
hidden_dim=config.motion_hidden_dim,
code_dim=config.motion_vocab_size,
down_t=config.motion_down_t,
depth=config.motion_depth,
dilation_growth_rate=config.motion_dilation_growth_rate,
)
self.quantizer = GumbelSoftmaxQuantizer(
nb_code=config.motion_vocab_size,
code_dim=config.motion_vocab_size,
)
def encode(self, motion: torch.Tensor) -> torch.Tensor:
"""
Encode motion features to discrete tokens.
Args:
motion: Motion features of shape (batch, time, 263)
Returns:
Token indices of shape (batch, time // downsample_ratio)
"""
# (batch, time, 263) -> (batch, 263, time)
x = motion.permute(0, 2, 1).float()
# Encode
x_enc = self.encoder(x) # (batch, nb_code, time')
# (batch, nb_code, time') -> (batch, time', nb_code)
x_enc = x_enc.permute(0, 2, 1).contiguous()
N, T, C = x_enc.shape
# Get token indices
indices = self.quantizer.quantize(x_enc.view(-1, C))
return indices.view(N, T)
def decode(self, tokens: torch.Tensor) -> torch.Tensor:
"""
Decode tokens back to motion features.
Args:
tokens: Token indices of shape (batch, time')
Returns:
Motion features of shape (batch, time, 263)
"""
# Get embeddings from tokens
x = self.quantizer.dequantize(tokens) # (batch, time', code_dim)
# (batch, time', code_dim) -> (batch, code_dim, time')
x = x.permute(0, 2, 1).contiguous()
# Decode
x_out = self.decoder(x) # (batch, 263, time)
# (batch, 263, time) -> (batch, time, 263)
return x_out.permute(0, 2, 1)
def forward(self, motion: torch.Tensor):
"""Forward pass for training (encode -> quantize -> decode)."""
x = motion.permute(0, 2, 1).float()
x_enc = self.encoder(x)
x_quant = self.quantizer(x_enc)
x_dec = self.decoder(x_quant)
return x_dec.permute(0, 2, 1)
# =====================================================
# Main GeoMotionGPT Model
# =====================================================
class GeoMotionGPTPreTrainedModel(PreTrainedModel):
"""Base class for GeoMotionGPT models."""
config_class = GeoMotionGPTConfig
base_model_prefix = "geomotiongpt"
supports_gradient_checkpointing = True
def _init_weights(self, module):
"""Initialize weights."""
if isinstance(module, (nn.Linear, nn.Conv1d)):
module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
if module.bias is not None:
module.bias.data.zero_()
elif isinstance(module, nn.Embedding):
module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
if module.padding_idx is not None:
module.weight.data[module.padding_idx].zero_()
elif isinstance(module, nn.LayerNorm):
module.bias.data.zero_()
module.weight.data.fill_(1.0)
class GeoMotionGPTForCausalLM(GeoMotionGPTPreTrainedModel):
"""
GeoMotionGPT Model for motion-to-text generation.
This model combines:
1. A VQ-VAE motion tokenizer (DVQ-GSST) for converting motion to discrete tokens
2. A fine-tuned GPT-2 model for generating text from motion tokens
Example:
```python
from transformers import AutoModelForCausalLM
import torch
# Load model
model = AutoModelForCausalLM.from_pretrained(
"zy22b/GeoMotionGPT",
trust_remote_code=True
)
# Access motion tokenizer
motion_tokenizer = model.motion_tokenizer
# Tokenize motion (batch, time, 263) -> (batch, tokens)
motion = torch.randn(1, 100, 263)
motion_tokens = motion_tokenizer.encode(motion)
# Generate text from motion tokens
text = model.generate_text(motion_tokens)
```
"""
_tied_weights_keys = ["lm_head.weight"]
def __init__(self, config: GeoMotionGPTConfig):
super().__init__(config)
# Motion tokenizer
self.motion_tokenizer = MotionTokenizer(config)
# Build GPT-2 config
gpt2_config = GPT2Config(
vocab_size=config.vocab_size,
n_positions=config.n_positions,
n_embd=config.n_embd,
n_layer=config.n_layer,
n_head=config.n_head,
n_inner=config.n_inner,
activation_function=config.activation_function,
resid_pdrop=config.resid_pdrop,
embd_pdrop=config.embd_pdrop,
attn_pdrop=config.attn_pdrop,
layer_norm_epsilon=config.layer_norm_epsilon,
initializer_range=config.initializer_range,
bos_token_id=config.bos_token_id,
eos_token_id=config.eos_token_id,
)
# Language model (GPT-2)
self.language_model = GPT2LMHeadModel(gpt2_config)
# Motion token embeddings (separate from text embeddings)
mot_embed_dim = int(config.n_embd // config.n_head * config.mot_factor) * config.n_head
self.motion_embed = nn.Embedding(
config.motion_vocab_size + 3, # +3 for special tokens (BOT, EOT, PAD)
mot_embed_dim
)
self.motion_head = nn.Linear(mot_embed_dim, config.motion_vocab_size + 3, bias=False)
# Projection layers for multi-modal fusion
self.motion_to_text_proj = nn.Linear(mot_embed_dim, config.n_embd)
self.text_to_motion_proj = nn.Linear(config.n_embd, mot_embed_dim)
# Initialize weights
self.post_init()
def get_input_embeddings(self):
return self.language_model.transformer.wte
def set_input_embeddings(self, value):
self.language_model.transformer.wte = value
def get_output_embeddings(self):
return self.language_model.lm_head
def set_output_embeddings(self, new_embeddings):
self.language_model.lm_head = new_embeddings
def encode_motion(self, motion: torch.Tensor) -> torch.Tensor:
"""
Encode motion features to discrete tokens.
Args:
motion: Motion features of shape (batch, time, 263)
Returns:
Token indices of shape (batch, time // 8)
"""
return self.motion_tokenizer.encode(motion)
def forward(
self,
input_ids: Optional[torch.LongTensor] = None,
attention_mask: Optional[torch.FloatTensor] = None,
past_key_values: Optional[Tuple[Tuple[torch.Tensor]]] = None,
labels: Optional[torch.LongTensor] = None,
use_cache: Optional[bool] = None,
output_attentions: Optional[bool] = None,
output_hidden_states: Optional[bool] = None,
return_dict: Optional[bool] = None,
**kwargs
):
"""
Forward pass through the language model.
For motion-to-text generation, use the `generate_text` method instead.
"""
return self.language_model(
input_ids=input_ids,
attention_mask=attention_mask,
past_key_values=past_key_values,
labels=labels,
use_cache=use_cache,
output_attentions=output_attentions,
output_hidden_states=output_hidden_states,
return_dict=return_dict,
)
def prepare_inputs_for_generation(self, input_ids, past_key_values=None, **kwargs):
"""Prepare inputs for text generation."""
return self.language_model.prepare_inputs_for_generation(
input_ids, past_key_values=past_key_values, **kwargs
)
@torch.no_grad()
def generate_text(
self,
motion_tokens: torch.Tensor,
max_new_tokens: int = 128,
num_beams: int = 4,
temperature: float = 0.7,
top_p: float = 0.9,
do_sample: bool = True,
**kwargs
) -> List[str]:
"""
Generate text descriptions from motion tokens.
Args:
motion_tokens: Motion token indices of shape (batch, seq_len)
max_new_tokens: Maximum number of new tokens to generate
num_beams: Number of beams for beam search
temperature: Sampling temperature
top_p: Top-p sampling parameter
do_sample: Whether to use sampling
Returns:
List of generated text strings
"""
device = motion_tokens.device
batch_size = motion_tokens.shape[0]
# Offset motion tokens (they come after text tokens)
motion_offset = self.config.text_vocab_size
input_ids = motion_tokens + motion_offset
# Add BOS token at the start
bos_tokens = torch.full(
(batch_size, 1),
self.config.bos_token_id,
dtype=torch.long,
device=device
)
input_ids = torch.cat([bos_tokens, input_ids], dim=1)
# Generate
outputs = self.language_model.generate(
input_ids=input_ids,
max_new_tokens=max_new_tokens,
num_beams=num_beams,
temperature=temperature,
top_p=top_p,
do_sample=do_sample,
pad_token_id=self.config.pad_token_id,
eos_token_id=self.config.eos_token_id,
**kwargs
)
# Decode only the generated part
generated_ids = outputs[:, input_ids.shape[1]:]
# Note: Actual text decoding requires a tokenizer
# Return raw generated IDs for now
return generated_ids
# Register for AutoClass
GeoMotionGPTConfig.register_for_auto_class()
GeoMotionGPTForCausalLM.register_for_auto_class("AutoModelForCausalLM")
|