File size: 3,898 Bytes
5b15eaf
 
 
 
 
d69cfa6
5b15eaf
d69cfa6
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5b15eaf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
16ad53d
 
 
 
 
 
 
 
 
 
 
 
5b15eaf
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
---
datasets:
- shenxq/OneVision
- shenxq/VideoChat2
base_model:
- Vision-CAIR/LongVU_Qwen2_7B_img
model-index:
- name: llava-onevision-qwen-7b-ov
  results:
  - task:
      type: multimodal
    dataset:
      name: EgoSchema
      type: egoschema
    metrics:
    - type: accuracy
      value: 67.6
      name: accuracy
      verified: true
  - task:
      type: multimodal
    dataset:
      name: MLVU
      type: mlvu
    metrics:
    - type: accuracy
      value: 65.4
      name: accuracy
      verified: true
  - task:
      type: multimodal
    dataset:
      name: MVBench
      type: mvbench
    metrics:
    - type: accuracy
      value: 66.9
      name: accuracy
      verified: true
  - task:
      type: multimodal
    dataset:
      name: VideoMME
      type: videomme
    metrics:
    - type: accuracy
      value: 60.6
      name: accuracy
      verified: true
---
# LongVU

Play with the model on the [HF demo](https://huggingface.co/spaces/Vision-CAIR/LongVU).

<div align="left">
    <a href='https://vision-cair.github.io/LongVU'><img src="https://longvu.s3.amazonaws.com/assets/demo.gif" alt="Demo GIF" style="width: 100%; max-width: 650px;"></a>
</div>

# Use

We provide the simple generation process for using our model. For more details, you could refer to [Github](https://github.com/Vision-CAIR/LongVU)

```python
# git clone https://github.com/Vision-CAIR/LongVU
import numpy as np
import torch
from longvu.builder import load_pretrained_model
from longvu.constants import (
    DEFAULT_IMAGE_TOKEN,
    IMAGE_TOKEN_INDEX,
)
from longvu.conversation import conv_templates, SeparatorStyle
from longvu.mm_datautils import (
    KeywordsStoppingCriteria,
    process_images,
    tokenizer_image_token,
)
from decord import cpu, VideoReader

tokenizer, model, image_processor, context_len = load_pretrained_model(
    "./checkpoints/longvu_qwen", None, "cambrian_qwen",
)

model.eval()
video_path = "./examples/video1.mp4"
qs = "Describe this video in detail"

vr = VideoReader(video_path, ctx=cpu(0), num_threads=1)
fps = float(vr.get_avg_fps())
frame_indices = np.array([i for i in range(0, len(vr), round(fps),)])
video = []
for frame_index in frame_indices:
    img = vr[frame_index].asnumpy()
    video.append(img)
video = np.stack(video)
image_sizes = [video[0].shape[:2]]
video = process_images(video, image_processor, model.config)
video = [item.unsqueeze(0) for item in video]

qs = DEFAULT_IMAGE_TOKEN + "\n" + qs
conv = conv_templates["qwen"].copy()
conv.append_message(conv.roles[0], qs)
conv.append_message(conv.roles[1], None)
prompt = conv.get_prompt()

input_ids = tokenizer_image_token(prompt, tokenizer, IMAGE_TOKEN_INDEX, return_tensors="pt").unsqueeze(0).to(model.device)
stop_str = conv.sep if conv.sep_style != SeparatorStyle.TWO else conv.sep2
keywords = [stop_str]
stopping_criteria = KeywordsStoppingCriteria(keywords, tokenizer, input_ids)
with torch.inference_mode():
    output_ids = model.generate(
        input_ids,
        images=video,
        image_sizes=image_sizes,
        do_sample=False,
        temperature=0.2,
        max_new_tokens=128,
        use_cache=True,
        stopping_criteria=[stopping_criteria],
    )
pred = tokenizer.batch_decode(output_ids, skip_special_tokens=True)[0].strip()
```

```
@misc{shen2024longvuspatiotemporaladaptivecompression,
      title={LongVU: Spatiotemporal Adaptive Compression for Long Video-Language Understanding}, 
      author={Xiaoqian Shen and Yunyang Xiong and Changsheng Zhao and Lemeng Wu and Jun Chen and Chenchen Zhu and Zechun Liu and Fanyi Xiao and Balakrishnan Varadarajan and Florian Bordes and Zhuang Liu and Hu Xu and Hyunwoo J. Kim and Bilge Soran and Raghuraman Krishnamoorthi and Mohamed Elhoseiny and Vikas Chandra},
      year={2024},
      eprint={2410.17434},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2410.17434}, 
}
```