File size: 9,979 Bytes
ba1d61a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
import os
import json
import argparse
from tqdm import tqdm
import time
from PIL import Image
import numpy as np
import torch
import torchvision.transforms as T
from decord import VideoReader, cpu
from torchvision.transforms.functional import InterpolationMode
from transformers import AutoModel, AutoTokenizer


IMAGENET_MEAN = (0.485, 0.456, 0.406)
IMAGENET_STD = (0.229, 0.224, 0.225)

DEFAULT_IMAGE_SIZE = 448
DEFAULT_VIDEO_SEGMENTS = 8
DEFAULT_MAX_PATCHES_PER_FRAME = 1  
DEFAULT_MAX_PATCHES_PER_IMAGE = 6  

def build_transform(input_size):
    return T.Compose([
        T.Lambda(lambda img: img.convert('RGB') if img.mode != 'RGB' else img),
        T.Resize((input_size, input_size), interpolation=InterpolationMode.BICUBIC),
        T.ToTensor(),
        T.Normalize(mean=IMAGENET_MEAN, std=IMAGENET_STD)
    ])

def find_closest_aspect_ratio(aspect_ratio, target_ratios, width, height, image_size):
    best_ratio_diff = float('inf')
    best_ratio = (1, 1)
    area = width * height
    for ratio in target_ratios:
        target_aspect_ratio = ratio[0] / ratio[1]
        ratio_diff = abs(aspect_ratio - target_aspect_ratio)
        if ratio_diff < best_ratio_diff:
            best_ratio_diff = ratio_diff
            best_ratio = ratio
        elif ratio_diff == best_ratio_diff:
            if area > 0.5 * image_size * image_size * ratio[0] * ratio[1]:
                best_ratio = ratio
    return best_ratio

def dynamic_preprocess(image, min_num=1, max_num=12, image_size=448, use_thumbnail=False):
    orig_width, orig_height = image.size
    aspect_ratio = orig_width / orig_height

    target_ratios = set(
        (i, j)
        for n in range(min_num, max_num + 1)
        for i in range(1, n + 1)
        for j in range(1, n + 1)
        if i * j <= max_num and i * j >= min_num
    )
    target_ratios = sorted(target_ratios, key=lambda x: x[0] * x[1])
    target_aspect_ratio = find_closest_aspect_ratio(
        aspect_ratio, target_ratios, orig_width, orig_height, image_size
    )

    target_width = image_size * target_aspect_ratio[0]
    target_height = image_size * target_aspect_ratio[1]
    blocks = target_aspect_ratio[0] * target_aspect_ratio[1]

    resized_img = image.resize((target_width, target_height))
    processed_images = []
    for i in range(blocks):
        box = (
            (i % (target_width // image_size)) * image_size,
            (i // (target_width // image_size)) * image_size,
            ((i % (target_width // image_size)) + 1) * image_size,
            ((i // (target_width // image_size)) + 1) * image_size
        )
        processed_images.append(resized_img.crop(box))
    assert len(processed_images) == blocks
    if use_thumbnail and len(processed_images) != 1:
        processed_images.append(image.resize((image_size, image_size)))
    return processed_images

def get_frame_indices(bound, fps, max_frame, first_idx=0, num_segments=32):
    if bound:
        start, end = bound[0], bound[1]
    else:
        start, end = -100000, 100000
    start_idx = max(first_idx, round(start * fps))
    end_idx = min(round(end * fps), max_frame)
    seg_size = float(end_idx - start_idx) / num_segments
    frame_indices = np.array([
        int(start_idx + (seg_size / 2) + np.round(seg_size * idx))
        for idx in range(num_segments)
    ])
    return frame_indices

def load_video(video_path, bound=None, input_size=448, max_num=1, num_segments=32):
    vr = VideoReader(video_path, ctx=cpu(0), num_threads=1)
    max_frame = len(vr) - 1
    fps = float(vr.get_avg_fps())

    pixel_values_list, num_patches_list = [], []
    transform = build_transform(input_size=input_size)
    frame_indices = get_frame_indices(bound, fps, max_frame, first_idx=0, num_segments=num_segments)
    
    valid_indices = [i for i in frame_indices if i < len(vr)]
    if not valid_indices:
        raise ValueError(f"No valid frames could be sampled from video {video_path}.")

    frames = vr.get_batch(valid_indices).asnumpy()

    for frame_np in frames:
        img = Image.fromarray(frame_np).convert('RGB')
        tiles = dynamic_preprocess(img, image_size=input_size, use_thumbnail=True, max_num=max_num)
        pixel_values = torch.stack([transform(tile) for tile in tiles])
        num_patches_list.append(pixel_values.shape[0])
        pixel_values_list.append(pixel_values)
        
    pixel_values = torch.cat(pixel_values_list)
    return pixel_values, num_patches_list

def get_media_type(file_path: str) -> str:
    ext = os.path.splitext(file_path)[1].lower()
    if ext in ['.mp4', '.avi', '.mov', '.mkv', '.webm']:
        return 'video'
    elif ext in ['.jpg', '.jpeg', '.png', '.bmp', '.gif', '.webp']:
        return 'image'
    else:
        raise ValueError(f"Unsupported file format: {ext} in file {file_path}")

def process_file(dataset_json_path: str, model, tokenizer, result_suffix: str):
    json_filename = os.path.basename(dataset_json_path)
    result_json_path = os.path.join(
        os.path.dirname(dataset_json_path),
        f"{os.path.splitext(json_filename)[0]}{result_suffix}"
    )

    if os.path.exists(result_json_path):
        print(f"Result file '{os.path.basename(result_json_path)}' already exists. Skipping.")
        return

    try:
        with open(dataset_json_path, 'r', encoding='utf-8') as f:
            data = json.load(f)
    except (json.JSONDecodeError, FileNotFoundError) as e:
        print(f"Failed to read or parse JSON file {dataset_json_path}: {e}")
        return

    generation_config = dict(num_beams=1, max_new_tokens=2048, do_sample=False)
    device = next(model.parameters()).device
    all_results = []
    
    base_path = os.path.dirname(dataset_json_path)

    for item in tqdm(data, desc=f"  Inferring on {json_filename}"):
        start_time = time.time()
        model_output = "N/A"
        
        try:
            prompt_text = item['conversations'][0]['value']
            ground_truth = item['conversations'][1]['value']
            media_path_key = 'image' if 'image' in item else 'video'
            media_relative_path = item.get(media_path_key)
            if not media_relative_path:
                raise ValueError("JSON item is missing 'image' or 'video' key.")
            
            media_full_path = os.path.join(base_path, media_relative_path)
            if not os.path.exists(media_full_path):
                raise FileNotFoundError(f"Media file not found: {media_full_path}")
            
            media_type = get_media_type(media_full_path)
            clean_prompt = prompt_text.replace("<image>", "").replace("<video>", "").strip()

            pixel_values, num_patches_list, question = None, None, None

            if media_type == 'image':
                image = Image.open(media_full_path).convert('RGB')
                transform = build_transform(input_size=DEFAULT_IMAGE_SIZE)
                patches = dynamic_preprocess(image, image_size=DEFAULT_IMAGE_SIZE, use_thumbnail=True, max_num=DEFAULT_MAX_PATCHES_PER_IMAGE)
                pixel_values = torch.stack([transform(p) for p in patches])
                num_patches_list = [len(patches)]
                question = f"<image>\n{clean_prompt}"
            
            elif media_type == 'video':
                pixel_values, num_patches_list = load_video(
                    media_full_path,
                    num_segments=DEFAULT_VIDEO_SEGMENTS,
                    max_num=DEFAULT_MAX_PATCHES_PER_FRAME,
                    input_size=DEFAULT_IMAGE_SIZE
                )
                video_prefix = ''.join([f'Frame{i+1}: <image>\n' for i in range(len(num_patches_list))])
                question = f"{video_prefix}{clean_prompt}"

            pixel_values = pixel_values.to(torch.bfloat16).to(device)

            response = model.chat(
                tokenizer=tokenizer,
                pixel_values=pixel_values,
                question=question,
                generation_config=generation_config,
                num_patches_list=num_patches_list,
                history=None
            )
            model_output = response.strip()

        except Exception as e:
            model_output = f"ERROR: {str(e)}"
        
        end_time = time.time()
        all_results.append({
            "id": item.get('id', 'N/A'),
            "prompt": prompt_text,
            "model_output": model_output,
            "ground_truth": ground_truth,
            "processing_time_seconds": round(end_time - start_time, 2)
        })

    with open(result_json_path, 'w', encoding='utf-8') as f:
        json.dump(all_results, f, indent=4, ensure_ascii=False)

def main():
    parser = argparse.ArgumentParser(description="Batch inference with InternVL model on local JSON datasets.")
    parser.add_argument("--model-path", required=True, help="Full path to the local model directory.")
    parser.add_argument("--result-suffix", default="_result.json", help="Suffix for the generated result files.")
    args = parser.parse_args()

    try:
        torch.backends.cuda.matmul.allow_tf32 = True
        model = AutoModel.from_pretrained(
            args.model_path,
            torch_dtype=torch.bfloat16,
            low_cpu_mem_usage=True,
            trust_remote_code=True,
            device_map="auto"
        ).eval()
        tokenizer = AutoTokenizer.from_pretrained(args.model_path, trust_remote_code=True, use_fast=False)
    except Exception as e:
        print(f"Failed to load the model from {args.model_path}. Error: {e}")
        return
    
    current_dir = os.getcwd()
    source_json_files = [
        f for f in os.listdir(current_dir)
        if f.endswith('.json') and not f.endswith(args.result_suffix)
    ]

    if not source_json_files:
        print(f"\nNo source JSON files: {current_dir}")
    else:
        for json_filename in sorted(source_json_files):
            process_file(os.path.join(current_dir, json_filename), model, tokenizer, args.result_suffix)


if __name__ == "__main__":
    main()