File size: 8,077 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
178
179
180
181
182
183
184
185
186
187
188
189
190
#
# Copyright (c) 2025 Huawei Technologies Co., Ltd. All Rights Reserved.
# Copyright 2025 The HuggingFace Inc. team
# Copyright 2024 The Qwen team, Alibaba Group and the HuggingFace Inc. team. All rights reserved.
# Adapted from transformers/models/qwen2_vl/image_processing_qwen2_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
import torch
from transformers.image_processing_utils import (
    BatchFeature,
    get_size_dict,
)
from transformers.image_utils import (
    OPENAI_CLIP_MEAN,
    OPENAI_CLIP_STD,
    ChannelDimension,
    SizeDict,
    get_image_size,
)
from transformers.processing_utils import Unpack, VideosKwargs
from transformers.utils import TensorType
from transformers.utils.import_utils import requires
from transformers.video_processing_utils import BASE_VIDEO_PROCESSOR_DOCSTRING, BaseVideoProcessor
from transformers.video_utils import group_videos_by_shape, reorder_videos
from transformers.image_utils import PILImageResampling
from transformers.models.qwen2_vl.image_processing_qwen2_vl import smart_resize
from torchvision import transforms
from torchvision.transforms.v2 import functional as F


class OpenPanguVLVideoProcessorInitKwargs(VideosKwargs):
    min_pixels: Optional[int]
    max_pixels: Optional[int]
    patch_size: Optional[int]
    temporal_patch_size: Optional[int]
    merge_size: Optional[int]
    any_res_dynamic_video_pixels: bool
    any_res_min_video_total_pixels: Optional[int]
    any_res_max_video_total_pixels: Optional[int]
    any_res_min_frame_pixels: Optional[int]
    any_res_max_frame_pixels: Optional[int]


@requires(backends=("torchvision",))
class OpenPanguVLVideoProcessor(BaseVideoProcessor):

    resample = PILImageResampling.BICUBIC
    size = {"height": 448, "width": 448}
    do_resize = True
    do_rescale = True
    rescale_factor = 1 / 255
    do_normalize = True
    do_convert_rgb = True
    min_pixels = 56 * 56
    max_pixels = 28 * 28 * 1280
    patch_size = 14
    temporal_patch_size = 1
    merge_size = 2
    image_mean = OPENAI_CLIP_MEAN
    image_std = OPENAI_CLIP_STD
    any_res_dynamic_video_pixels = True
    any_res_min_video_total_pixels = 448 * 448 * 32
    any_res_max_video_total_pixels = 448 * 448 * 32
    any_res_min_frame_pixels = 56 * 56
    any_res_max_frame_pixels = 28 * 28 * 1280
    valid_kwargs = OpenPanguVLVideoProcessorInitKwargs
    model_input_names = ["pixel_values_videos", "video_grid_thw"]
    dtype = torch.bfloat16

    def __init__(self, **kwargs: Unpack[OpenPanguVLVideoProcessorInitKwargs]):
        super().__init__(**kwargs)

    def _preprocess(
        self,
        videos: List["torch.Tensor"],
        do_convert_rgb: bool,
        do_resize: bool,
        size: SizeDict,
        interpolation: Optional["F.InterpolationMode"],
        do_rescale: bool,
        rescale_factor: float,
        do_normalize: bool,
        image_mean: Optional[Union[float, List[float]]],
        image_std: Optional[Union[float, List[float]]],
        return_tensors: Optional[Union[str, TensorType]] = None,
        patch_size: Optional[int] = None,
        temporal_patch_size: Optional[int] = None,
        merge_size: Optional[int] = None,
        **kwargs,
    ):
        temporal_patch_size = OpenPanguVLVideoProcessor.temporal_patch_size
        # Recalculate the maximum and minimum resolution of a single frame
        num_frames = sum(video.shape[0] for video in videos)
        if not self.any_res_dynamic_video_pixels:
            self.min_pixels = self.any_res_min_frame_pixels
            self.max_pixels = self.any_res_max_frame_pixels
        else:
            # dynamic video pixels
            self.min_pixels = max(min(self.any_res_min_video_total_pixels // num_frames, \
                              self.any_res_max_frame_pixels), self.any_res_min_frame_pixels)
            self.max_pixels = max(min(self.any_res_max_video_total_pixels // num_frames, \
                              self.any_res_max_frame_pixels), self.any_res_min_frame_pixels)
        # Group videos by size for batched resizing
        grouped_videos, grouped_videos_index = group_videos_by_shape(videos)
        resized_videos_grouped = {}
        for shape, stacked_videos in grouped_videos.items():
            height, width = get_image_size(stacked_videos[0], channel_dim=ChannelDimension.FIRST)
            resized_height, resized_width = height, width
            if do_resize:
                resized_height, resized_width = smart_resize(
                    height,
                    width,
                    factor=patch_size * merge_size,
                    min_pixels=self.min_pixels,
                    max_pixels=self.max_pixels,
                )
                stacked_videos = F.resize(
                    stacked_videos, size=(resized_height, resized_width), interpolation=interpolation
                )
            resized_videos_grouped[shape] = stacked_videos 
        resized_videos = reorder_videos(resized_videos_grouped, grouped_videos_index)
        # Group videos by size for further processing
        # Needed in case do_resize is False, or resize returns videos with different sizes
        grouped_videos, grouped_videos_index = group_videos_by_shape(resized_videos)
        processed_videos_grouped = {}
        processed_video_grid_thw = {}
        for shape, stacked_videos in grouped_videos.items():
            resized_height, resized_width = get_image_size(stacked_videos[0], channel_dim=ChannelDimension.FIRST)

            # rescale and normalize
            stacked_videos = torch.mul(stacked_videos, rescale_factor)
            stacked_videos = transforms.Normalize(mean=image_mean, std=image_std)(stacked_videos)


            # Need to fill frames to cope with temporal_patch_size, avoid time block sticking
            stacked_videos = torch.repeat_interleave(stacked_videos, repeats=temporal_patch_size, dim=1)

            batch_size, grid_t, channel = stacked_videos.shape[:3]
            grid_t, grid_h, grid_w = (
                grid_t // temporal_patch_size,
                resized_height // patch_size,
                resized_width // patch_size,
            )

            stacked_videos = stacked_videos.view(
                batch_size,
                grid_t,
                temporal_patch_size,
                channel,
                grid_h // merge_size,
                merge_size,
                patch_size,
                grid_w // merge_size,
                merge_size,
                patch_size,
            )
            stacked_videos = stacked_videos.permute(0, 1, 4, 7, 5, 8, 3, 2, 6, 9) 
            processed_stacked_videos = stacked_videos.reshape(
                batch_size,
                grid_t * grid_h * grid_w,
                channel * temporal_patch_size * patch_size * patch_size,
            )

            processed_videos_grouped[shape] = processed_stacked_videos
            processed_video_grid_thw[shape] = [[grid_t, grid_h, grid_w]] * batch_size

        processed_videos = reorder_videos(processed_videos_grouped, grouped_videos_index)
        processed_video_grid_thw = reorder_videos(processed_video_grid_thw, grouped_videos_index)
        pixel_values_videos = torch.cat(processed_videos, dim=0).to(OpenPanguVLVideoProcessor.dtype)
        video_grid_thw = torch.tensor(processed_video_grid_thw)
        return BatchFeature(
            data={"pixel_values_videos": pixel_values_videos, "video_grid_thw": video_grid_thw},
            tensor_type=return_tensors,
        )


__all__ = ["OpenPanguVLVideoProcessor"]