File size: 20,050 Bytes
8063d5b |
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 |
"""To minimize modifications to pre-existing code,
we store all multimodal informations as image embeddings.
"""
import os
from dataclasses import dataclass
from typing import Any, Dict, List, Optional, Union, cast
import torch
from transformers import PreTrainedTokenizer
from transformers.feature_extraction_utils import BatchFeature
from transformers.processing_utils import (
AllKwargsForChatTemplate,
ProcessingKwargs,
ProcessorMixin,
Unpack,
render_jinja_template,
)
from transformers.tokenization_utils_base import PreTokenizedInput, TextInput
JINJA_TEMPLATE = """
{%- set image_count = namespace(value=0) %}\
{%- set content_ns_sys = namespace(content='') %}
{%- if messages[0].role == \'system\' %}\
{%- if messages[0].content is string %}
{%- set content_ns_sys.content = messages[0].content %}
{%- elif 'type' in messages[0].content[0] %}
{%- set content_ns_sys.content = messages[0].content[0].text %}
{%- endif %}
{%- endif %}
{%- if tools %}\
{{- \'<|im_start|>system\\n\' }}\
{%- if messages[0].role == \'system\' %}\
{{- content_ns_sys.content ~ \'\\n\\n\' }}\
{%- endif %}\
{{- "# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>" }}\
{%- for tool in tools %}\
{{- "\\n" }}\
{{- tool | tojson }}\
{%- endfor %}\
{{- "\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\"name\\": <function-name>, \\"arguments\\": <args-json-object>}\\n</tool_call><|im_end|>\\n" }}\
{%- else %}\
{%- if messages[0].role == \'system\' %}\
{{- \'<|im_start|>system\\n\' ~ content_ns_sys.content ~ \'<|im_end|>\\n\' }}\
{%- endif %}\
{%- endif %}\
{%- set ns = namespace(multi_step_tool=true, last_query_index=messages|length - 1) %}\
{%- for message in messages[::-1] %}\
{%- set index = (messages|length - 1) - loop.index0 %}\
{%- if ns.multi_step_tool and message.role == "user" and (message.content is string or 'text' in message.content[0]) %}\
{%- set text_content = message.content if message.content is string else message.content[0].text %}\
{%- if not text_content.startswith('<tool_response>') and not text_content.endswith('</tool_response>') %}\
{%- set ns.multi_step_tool = false %}\
{%- set ns.last_query_index = index %}\
{%- endif %}\
{%- endif %}\
{%- endfor %}\
{%- for message in messages %}\
{%- set content_ns = namespace(content='') %}
{%- if message.content is string %}\
{%- set content_ns.content = message.content %}\
{%- elif 'type' in message.content[0] %}\
{%- for m in message.content %}\
{%- if m.type == "text" %}\
{%- set content_ns.content = content_ns.content ~ m.text %}\
{%- endif %}\
{%- endfor %}\
{%- endif %}\
{%- set content = content_ns.content %}\
{%- if (message.role == "user") or (message.role == "system" and not loop.first) %}\
{{- \'<|im_start|>\' ~ message.role ~ \'\\n\' ~ content ~ \'<|im_end|>\' ~ \'\\n\' }}\
{%- elif message.role == "assistant" %}\
{%- set reasoning_content = \'\' %}\
{%- if message.reasoning_content is string %}\
{%- set reasoning_content = message.reasoning_content %}\
{%- else %}\
{%- if \'</think>\' in content %}\
{%- set reasoning_content = content.split(\'</think>\')[0].rstrip(\'\\n\').split(\'<think>\')[-1].lstrip(\'\\n\') %}\
{%- set content = content.split(\'</think>\')[-1].lstrip(\'\\n\') %}\
{%- endif %}\
{%- endif %}\
{%- if loop.index0 > ns.last_query_index %}\
{%- if loop.last or (not loop.last and reasoning_content) %}\
{{- \'<|im_start|>\' ~ message.role ~ \'\\n<think>\\n\' ~ reasoning_content.strip(\'\\n\') ~ \'\\n</think>\\n\\n\' ~ content.lstrip(\'\\n\') }}\
{%- else %}\
{{- \'<|im_start|>\' ~ message.role ~ \'\\n\' ~ content }}\
{%- endif %}\
{%- else %}\
{{- \'<|im_start|>\' ~ message.role ~ \'\\n\' ~ content }}\
{%- endif %}\
{%- if message.tool_calls %}\
{%- for tool_call in message.tool_calls %}\
{%- if (loop.first and content) or (not loop.first) %}\
{{- \'\\n\' }}\
{%- endif %}\
{%- if tool_call.function %}\
{%- set tool_call = tool_call.function %}\
{%- endif %}\
{{- \'<tool_call>\\n{"name": "\' }}\
{{- tool_call.name }}\
{{- \'", "arguments": \' }}\
{%- if tool_call.arguments is string %}\
{{- tool_call.arguments }}\
{%- else %}\
{{- tool_call.arguments | tojson }}\
{%- endif %}\
{{- \'}\\n</tool_call>\' }}\
{%- endfor %}\
{%- endif %}\
{{- \'<|im_end|>\\n\' }}\
{%- elif message.role == "tool" %}\
{%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}\
{{- \'<|im_start|>user\' }}\
{%- endif %}\
{{- \'\\n<tool_response>\\n\' }}\
{{- content }}\
{{- \'\\n</tool_response>\' }}\
{%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}\
{{- \'<|im_end|>\\n\' }}\
{%- endif %}\
{%- endif %}\
{%- endfor %}\
{%- if add_generation_prompt %}\
{{- \'<|im_start|>assistant\\n\' }}\
{%- if enable_thinking is defined and enable_thinking is false %}\
{{- \'<think>\\n\\n</think>\\n\\n\' }}\
{%- endif %}\
{%- endif %}
"""
@dataclass
class ProtKwargs: ...
class DockGenProcessorKwargs(ProcessingKwargs):
images_kwargs: ProtKwargs = ProtKwargs()
_defaults = {
"text_kwargs": {
"padding": "longest",
"return_tensors": "pt",
},
"images_kwargs": {},
"common_kwargs": {
"return_tensors": "pt",
},
}
class DockGenProcessor(ProcessorMixin):
"""
A tokenizer for the Qwen-3 model that handles multimodal inputs.
The input of the tokenizer is expected to be a list of dictionaries,
where each dictionaries containing a 'text' keya are separated by a special token: ""<|image_pad|>".
"""
attributes = ["tokenizer"]
tokenizer_class = "AutoTokenizer"
def __init__(
self,
tokenizer: Optional[PreTrainedTokenizer] = None,
image_token: str = "<|image_pad|>",
**kwargs: Unpack[DockGenProcessorKwargs],
) -> None:
if tokenizer is None:
raise ValueError("You need to specify a `tokenizer`.")
self.image_token = (
tokenizer.image_token if hasattr(tokenizer, "image_token") else image_token
)
self.image_token_id = (
tokenizer.image_token_id
if getattr(tokenizer, "image_token_id", None)
else tokenizer.convert_tokens_to_ids(self.image_token)
)
super().__init__(tokenizer)
self.chat_template = (
kwargs["chat_template"] if "chat_template" in kwargs else JINJA_TEMPLATE
)
self.pad_token = self.tokenizer.pad_token
self.pad_token_id = self.tokenizer.pad_token_id
self.eos_token = self.tokenizer.eos_token
self.eos_token_id = self.tokenizer.eos_token_id
def __call__(
self,
text: Union[
TextInput, PreTokenizedInput, list[TextInput], list[PreTokenizedInput]
] = None,
images: Optional[List[torch.Tensor] | List[str] | str] = None,
**kwargs: Unpack[DockGenProcessorKwargs],
) -> BatchFeature:
if isinstance(images, list) and len(images) == 1 and isinstance(images[0], str):
# If images is a list with a single string, convert it to a string
images = images[0]
if isinstance(images, list) and (len(images) == 0 or images[0] is None):
# If images is an empty list, set it to None
images = None
if isinstance(images, str):
# If images are provided as a single path, split it on |
images = images.split("|")
if images == []:
images = None
output_kwargs = self._merge_kwargs(
DockGenProcessorKwargs,
tokenizer_init_kwargs=self.tokenizer.init_kwargs,
**kwargs,
)
return_tensors = output_kwargs["common_kwargs"]["return_tensors"]
if images is not None:
if isinstance(images[0], str):
cast(List[str], images)
# If images are provided as paths, load them
images = [torch.load(img) for img in images]
if len(images) == 1:
image_inputs = {
"pixel_values": images[0].unsqueeze(0) # type: ignore
} # SHAPE: [BATCH_SIZE, MAX_N_MM, EMB_SIZE]
else:
# If images are two dimensional tensors, concatenate them
if images[0].ndim == 2: # type: ignore
image_inputs = {
"pixel_values": torch.concatenate(
[im for im in images if im is not None], dim=0
)
}
else:
assert images[0].ndim == 1 # type: ignore
image_inputs = {
"pixel_values": torch.stack(
[im for im in images if im is not None], dim=0
)
}
else:
image_inputs = {}
if isinstance(text, str):
text = [text]
elif not isinstance(text, list) and not isinstance(text[0], str):
raise ValueError(
"Invalid input text. Please provide a string, or a list of strings"
)
# try to expand inputs in processing if we have the necessary parts
prompt_strings = text
return_tensors = output_kwargs["text_kwargs"].pop("return_tensors", None)
text_inputs = self.tokenizer(prompt_strings, **output_kwargs["text_kwargs"])
self._check_special_mm_tokens(prompt_strings, text_inputs, modalities=["image"])
return BatchFeature(
data={
**text_inputs,
**image_inputs,
},
tensor_type=return_tensors,
)
# Need to override apply_chat_template
def apply_chat_template(
self,
conversation: List[List[Dict[str, Any]]] | List[Dict[str, Any]],
chat_template: Optional[str] = None,
**kwargs: Unpack[AllKwargsForChatTemplate],
) -> Any:
if chat_template is None:
if isinstance(self.chat_template, dict) and "default" in self.chat_template:
chat_template = self.chat_template["default"]
elif isinstance(self.chat_template, dict):
raise ValueError(
'The processor has multiple chat templates but none of them are named "default". You need to specify'
" which one to use by passing the `chat_template` argument. Available templates are: "
f"{', '.join(self.chat_template.keys())}"
)
elif self.chat_template is not None:
chat_template = self.chat_template
else:
raise ValueError(
"Cannot use apply_chat_template because this processor does not have a chat template."
)
else:
if (
isinstance(self.chat_template, dict)
and chat_template in self.chat_template
):
# It's the name of a template, not a full template string
chat_template = self.chat_template[chat_template]
else:
# It's a template string, render it directly
chat_template = chat_template
if kwargs.get("continue_final_message", False):
if kwargs.get("add_generation_prompt", False):
raise ValueError(
"continue_final_message and add_generation_prompt are not compatible. Use continue_final_message when you want the model to continue the final message, and add_generation_prompt when you want to add a header that will prompt it to start a new assistant message instead."
)
if kwargs.get("return_assistant_tokens_mask", False):
raise ValueError(
"continue_final_message is not compatible with return_assistant_tokens_mask."
)
# Fill sets of kwargs that should be used by different parts of template
processed_kwargs: Dict[str, Dict[str, Any]] = {
"mm_load_kwargs": {},
"template_kwargs": {},
}
for kwarg_type in processed_kwargs:
for key in AllKwargsForChatTemplate.__annotations__[
kwarg_type
].__annotations__.keys():
kwarg_type_defaults = AllKwargsForChatTemplate.__annotations__[
kwarg_type
]
default_value = getattr(kwarg_type_defaults, key, None)
value = kwargs.pop(key, default_value)
if value is not None and not isinstance(value, dict):
processed_kwargs[kwarg_type][key] = value
# Pass unprocessed custom kwargs
processed_kwargs["template_kwargs"].update(kwargs)
conversations: List[List[Dict[str, Any]]] = []
if isinstance(conversation, list) and isinstance(conversation[0], list):
conversations = cast(List[List[Dict[str, Any]]], conversation)
is_batched = True
elif isinstance(conversation, list) and isinstance(conversation[0], dict):
conversations = [cast(List[Dict[str, Any]], conversation)]
is_batched = False
else:
raise ValueError("Wrong conversation format.")
tokenize = processed_kwargs["template_kwargs"].pop("tokenize", False)
return_dict = processed_kwargs["template_kwargs"].pop("return_dict", False)
# mm_load_kwargs = processed_kwargs["mm_load_kwargs"]
batch_images_path: List[List[str]] = []
batch_images: List[torch.Tensor] = []
for conv in conversations:
images: List[torch.Tensor] = []
images_path: List[str] = []
for message in conv:
visuals: List[Dict[str, Any]] = [
content
for content in message["content"]
if content["type"] in ["image"]
]
image_fnames: List[str] = [
vision_info["path"]
for vision_info in visuals
if vision_info["type"] == "image"
]
for fname in image_fnames:
data_path = os.environ.get("DATA_PATH", "")
images.append(torch.load(os.path.join(data_path, fname)))
images_path.append(os.path.join(data_path, fname))
if images:
batch_images.append(torch.stack(images, dim=0))
batch_images_path.append(images_path)
prompt, generation_indices = render_jinja_template(
conversations=conversations,
chat_template=chat_template,
**processed_kwargs[
"template_kwargs"
], # different flags such as `return_assistant_mask`
**self.tokenizer.special_tokens_map, # tokenizer special tokens are used by some templates
)
if not is_batched:
prompt = prompt[0]
batch_images_path = batch_images_path[0] if batch_images_path else [] # type: ignore
if tokenize:
# Tokenizer's `apply_chat_template` never adds special tokens when tokenizing
# But processor's `apply_chat_template` didn't have an option to tokenize, so users had to format the prompt
# and pass it to the processor. Users thus never worried about special tokens relying on processor handling
# everything internally. The below line is to keep BC for that and be able to work with model that have
# special tokens in the template (consistent with tokenizers). We dont want to raise warning, it will flood command line
# without actionable solution for users
single_prompt = prompt[0] if is_batched else prompt
if self.tokenizer.bos_token is not None and single_prompt.startswith(
self.tokenizer.bos_token
):
kwargs["add_special_tokens"] = False
out = self(
text=prompt,
images=batch_images if batch_images else None,
**kwargs,
)
if return_dict:
if processed_kwargs["template_kwargs"].get(
"return_assistant_tokens_mask", False
):
assistant_masks = []
input_ids = out["input_ids"]
for i in range(len(input_ids)):
current_mask = [0] * len(input_ids[i])
for (
assistant_start_char,
assistant_end_char,
) in generation_indices[i]:
start_token = out.char_to_token(i, assistant_start_char)
end_token = out.char_to_token(i, assistant_end_char - 1)
if start_token is None:
# start_token is out of bounds maybe due to truncation.
break
for token_id in range(
start_token,
end_token + 1 if end_token else len(input_ids[i]),
):
current_mask[token_id] = 1
assistant_masks.append(current_mask)
out["assistant_masks"] = assistant_masks
out.convert_to_tensors(
tensor_type=kwargs.get("return_tensors", None)
)
return out
else:
return out["input_ids"]
if return_dict:
return {
"text": prompt,
"mm_data": batch_images_path,
}
return prompt
def batch_decode(self, *args: Any, **kwargs: Any) -> Any:
"""
This method forwards all its arguments to LlamaTokenizerFast's [`~PreTrainedTokenizer.batch_decode`]. Please
refer to the docstring of this method for more information.
"""
return self.tokenizer.batch_decode(*args, **kwargs)
def decode(self, *args: Any, **kwargs: Any) -> Any:
"""
This method forwards all its arguments to LlamaTokenizerFast's [`~PreTrainedTokenizer.decode`]. Please refer to
the docstring of this method for more information.
"""
return self.tokenizer.decode(*args, **kwargs)
|