File size: 8,044 Bytes
1688f96
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
#
# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
# Copyright 2025 The Qwen Team and The HuggingFace Inc. team. All rights reserved.
# Adapted from transformers/models/qwen2_5_vl/processing_qwen2_5_vl.py

# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

from typing import List, Optional, Union, Tuple
from transformers.utils import logging
from transformers.models.qwen2_5_vl.processing_qwen2_5_vl import Qwen2_5_VLProcessor, Qwen2_5_VLProcessorKwargs
from transformers.feature_extraction_utils import BatchFeature
from transformers.image_utils import ImageInput
from transformers.processing_utils import ImagesKwargs, ProcessingKwargs, ProcessorMixin, Unpack, VideosKwargs
from transformers.tokenization_utils_base import PreTokenizedInput, TextInput
from transformers.video_utils import VideoInput

logger = logging.get_logger(__name__)


class OpenPanguVLProcessor(Qwen2_5_VLProcessor):
    tokenizer_class = ("AutoTokenizer")

    def __init__(self, image_processor=None, tokenizer=None, video_processor=None, chat_template=None, **kwargs):
        self.tokenizer = tokenizer

        self.image_token = "[unused19]"
        self.video_token = "[unused32]"
        self.vision_start_token = "[unused18]"
        self.vision_end_token = "[unused20]"

        self.image_token_id = (
            self.tokenizer.image_token_id
            if getattr(self.tokenizer, "image_token_id", None)
            else self.tokenizer.convert_tokens_to_ids(self.image_token)
        )
        self.video_token_id = (
            self.tokenizer.video_token_id
            if getattr(self.tokenizer, "video_token_id", None)
            else self.tokenizer.convert_tokens_to_ids(self.video_token)
        )
        self.vision_start_token_id = (
            self.tokenizer.vision_start_token_id
            if getattr(self.tokenizer, "vision_start_token_id", None)
            else self.tokenizer.convert_tokens_to_ids(self.vision_start_token)
        )
        self.vision_end_token_id = (
            self.tokenizer.vision_end_token_id
            if getattr(self.tokenizer, "vision_end_token_id", None)
            else self.tokenizer.convert_tokens_to_ids(self.vision_end_token)
        )

        self.image_processor = image_processor
        self.video_processor = video_processor
        self.chat_template = chat_template

    def __call__(
        self,
        images: ImageInput = None,
        text: Union[TextInput, PreTokenizedInput, List[TextInput], List[PreTokenizedInput]] = None,
        videos: VideoInput = None,
        **kwargs: Unpack[Qwen2_5_VLProcessorKwargs],
    ) -> BatchFeature:
        output_kwargs = self._merge_kwargs(
            Qwen2_5_VLProcessorKwargs,
            tokenizer_init_kwargs=self.tokenizer.init_kwargs,
            **kwargs,
        )
        if not isinstance(text, list):
            text = [text]
        image_inputs = {}
        videos_inputs = {}
        if images is not None:
            image_inputs = self.image_processor(images=images, **output_kwargs["images_kwargs"])
            self._process_vision_placeholders(
                text=text,
                vision_token=self.image_token,
                grid_thw=image_inputs["image_grid_thw"],
                merge_size=self.image_processor.merge_size,
                vision_start_token=self.vision_start_token,
                vision_end_token=self.vision_end_token,
            )
        if videos is not None:
            videos_inputs = self.video_processor(videos=videos, **output_kwargs["videos_kwargs"])
            self._process_vision_placeholders(
                text=text,
                vision_token=self.video_token,
                grid_thw=videos_inputs["video_grid_thw"],
                merge_size=self.video_processor.merge_size,
                vision_start_token=self.vision_start_token,
                vision_end_token=self.vision_end_token,
            )
        return_tensors = output_kwargs["text_kwargs"].pop("return_tensors", None)
        return_mm_token_type_ids = output_kwargs["text_kwargs"].pop("return_mm_token_type_ids", None)
        text_inputs = self.tokenizer(text, **output_kwargs["text_kwargs"])
        self._check_special_mm_tokens(text, text_inputs, modalities=["image", "video"])

        if return_mm_token_type_ids:
            array_ids = np.array(text_inputs["input_ids"])
            mm_token_type_ids = np.zeros_like(text_inputs["input_ids"])
            mm_token_type_ids[array_ids == self.image_token_id] = 1
            text_inputs["mm_token_type_ids"] = mm_token_type_ids.tolist()

        return BatchFeature(data={**text_inputs, **image_inputs, **videos_inputs}, tensor_type=return_tensors)

    @staticmethod
    def _process_vision_placeholders(
        text: List[str],
        vision_token: str,
        grid_thw: List[Tuple[int, int, int]],
        merge_size: int,
        vision_start_token: str,
        vision_end_token: str,
    ) -> None:
        """
        Replace placeholder tokens (e.g., <image> or <video>) in text with structured vision token sequences.
        For images:
            - `grid_thw[i] = (1, H, W)` → treated as a single time slice.
        For videos:
            - `grid_thw[i] = (T, H, W)` → repeated `T` times.

        Args:
            text (`List[str]`):
                List of input text strings that may contain placeholder tokens.
            vision_token (`str`):
                The vision token to replace (e.g., "<|image|>" or "<|video|>").
            grid_thw (`List[Tuple[int, int, int]]`):
                List of (T, H, W) grid dimensions for each media item.
            merge_size (`int`):
                Spatial merging factor used to compute token count per frame.
            vision_start_token (`str`):
                Special token marking the start of a vision sequence.
            vision_end_token (`str`):
                Special token marking the end of a vision sequence.

        Returns:
            `None`. The `text` is modified in-place.
        """
        index = 0
        for i in range(len(text)):
            while vision_token in text[i]:
                if index >= len(grid_thw):
                    raise ValueError(
                        f"Found more vision tokens than entries in 'grid_thw'. "
                        f"Expected one (T, H, W) tuple per '{vision_token}' token."
                    )
                grid_t, grid_h, grid_w = grid_thw[index]
                # Calculate the sequence length per time slice based on the grid size and merge length.
                seq_length_per_time = (grid_h * grid_w) // (merge_size ** 2)
                # Prepare a placeholder string that includes start and end tokens,
                # and then calculate the number of media tokens to replace.
                placeholder_string = (
                    vision_start_token
                    + ("<|vision_placeholder|>" * seq_length_per_time)
                    + vision_end_token
                )
                if grid_t > 1:
                    # For videos only, repeat the placeholder string for each time slice.
                    placeholder_string *= grid_t
                placeholder_string = placeholder_string.removeprefix(vision_start_token)
                placeholder_string = placeholder_string.removesuffix(vision_end_token)
                text[i] = text[i].replace(vision_token, placeholder_string, 1)
                index += 1
            text[i] = text[i].replace("<|vision_placeholder|>", vision_token)


__all__ = ["OpenPanguVLProcessor"]