Hugging Face's logo

THUdyh
/
Oryx-7B

Text Generation
Safetensors
English
Chinese
oryx_qwen
conversational
Model card Files Community
1
Oryx-7B
File size: 4,629 Bytes 
6f87ed1
 
 
 
 
 
 
8e21d27
 
 
6f87ed1
 
 
 
 
 
 
 
 
70f5dff
6f87ed1
d407444
6f87ed1
 
 
1ceac3f
 
6f87ed1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7fa65bb
74d075b
6f87ed1
 
 
 
 
 
 
 
 
 
 
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
 
---
license: apache-2.0
datasets:
- THUdyh/Oryx-SFT-Data
base_model:
- Qwen/Qwen2-7B-Instruct
pipeline_tag: text-generation
language:
- en
- zh
---
# Oryx-7B

## Model Summary

The Oryx models are 7/34B parameter models trained on [Oryx-SFT-Data](https://huggingface.co/datasets/THUdyh/Oryx-SFT-Data), based on Qwen2 language model with a context window of 32K tokens.

Oryx offers an on-demand solution to seamlessly and efficiently process visual inputs with arbitrary spatial sizes and temporal lengths.

- **Repository:** https://github.com/Oryx-mllm/Oryx
- **Languages:** English, Chinese
- **Paper:** https://arxiv.org/abs/2409.12961

## Use

We provide a simple generation process for using our model. For more details, please refer to our [Github Repo](https://github.com/liuzuyan/oryx)

```
from oryx.model.builder import load_pretrained_model
from oryx.mm_utils import get_model_name_from_path, process_images, tokenizer_image_token
from oryx.constants import IMAGE_TOKEN_INDEX, DEFAULT_IMAGE_TOKEN, DEFAULT_IM_START_TOKEN, DEFAULT_IM_END_TOKEN, IGNORE_INDEX
from oryx.conversation import conv_templates, SeparatorStyle
from PIL import Image
import requests
import copy
import torch
import sys
import warnings
from decord import VideoReader, cpu
import numpy as np

def load_video(self, video_path, max_frames_num,fps=1,force_sample=False):
    if max_frames_num == 0:
        return np.zeros((1, 336, 336, 3))
    vr = VideoReader(video_path, ctx=cpu(0),num_threads=1)
    total_frame_num = len(vr)
    video_time = total_frame_num / vr.get_avg_fps()
    fps = round(vr.get_avg_fps()/fps)
    frame_idx = [i for i in range(0, len(vr), fps)]
    frame_time = [i/fps for i in frame_idx]
    if len(frame_idx) > max_frames_num or force_sample:
        sample_fps = max_frames_num
        uniform_sampled_frames = np.linspace(0, total_frame_num - 1, sample_fps, dtype=int)
        frame_idx = uniform_sampled_frames.tolist()
        frame_time = [i/vr.get_avg_fps() for i in frame_idx]
    frame_time = ",".join([f"{i:.2f}s" for i in frame_time])
    spare_frames = vr.get_batch(frame_idx).asnumpy()
    # import pdb;pdb.set_trace()
    return spare_frames,frame_time,video_time
pretrained = "THUdyh/Oryx-7B"
model_name = "oryx_qwen"
device = "cuda"
device_map = "auto"
tokenizer, model, image_processor, max_length = load_pretrained_model(pretrained, None, model_name, device_map=device_map)
model.eval()
video_path = ""
max_frames_num = "64"
video,frame_time,video_time = load_video(video_path, max_frames_num, 1, force_sample=True)
video = image_processor.preprocess(video, return_tensors="pt")["pixel_values"].cuda().bfloat16()
video = [video]
video_data = (video, video)
input_data = (video_data, (384, 384), "video")
conv_template = "qwen_1_5"
question = DEFAULT_IMAGE_TOKEN + "\nPlease describe this video in detail."
conv = copy.deepcopy(conv_templates[conv_template])
conv.append_message(conv.roles[0], question)
conv.append_message(conv.roles[1], None)
prompt_question = conv.get_prompt()
input_ids = tokenizer_image_token(prompt_question, tokenizer, IMAGE_TOKEN_INDEX, return_tensors="pt").unsqueeze(0).to(device)
output_ids = model.generate(
    inputs=input_ids,
    images=input_data[0][0],
    images_highres=input_data[0][1],
    modalities=video_data[2],
    do_sample=False,
    temperature=0,
    max_new_tokens=128,
    use_cache=True,
)

text_outputs = tokenizer.batch_decode(cont, skip_special_tokens=True)
print(text_outputs)
```


### Results

#### General Video Benchmark

<img src="https://cdn-uploads.huggingface.co/production/uploads/652965773a416e1f2173443b/hKfOK0u3OXly_u4hgGLDB.png" alt="image/png" style="zoom: 33%;" />

#### Long-Form Video Understanding

<img src="https://cdn-uploads.huggingface.co/production/uploads/652965773a416e1f2173443b/Xweq9f4OWkqeVc_FZIMuO.png" alt="image/png" style="zoom:33%;" />

#### Common Image Benchmark

<img src="https://cdn-uploads.huggingface.co/production/uploads/652965773a416e1f2173443b/ybfroSA9WaKXtJbP_9cLR.png" alt="image/png" style="zoom:33%;" />

#### 3D Spatial Understanding

<img src="https://cdn-uploads.huggingface.co/production/uploads/652965773a416e1f2173443b/5v8ACRzAoKS0FbcVBXZhT.png" alt="image/png" style="zoom:33%;" />



### Model Architecture

- **Architecture:** Pre-trained [Oryx-ViT](https://huggingface.co/THUdyh/Oryx-ViT) + Qwen2-7B
- **Init Model:** [Oryx-7B-Image](https://huggingface.co/THUdyh/Oryx-7B-Image)
- **Data:** a mixture of 1.2M image/video data
- **Precision:** BFloat16

#### Hardware & Software

- **Hardware:** 64 * NVIDIA Tesla A100
- **Orchestration:** HuggingFace Trainer
- **Code:** Pytorch

## Citation