File size: 1,565 Bytes
d6b3c68
 
 
205b025
91abf6f
 
028d758
91abf6f
d6b3c68
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
567dab9
e4f0bfd
567dab9
d6b3c68
 
567dab9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d6b3c68
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
---
pipeline_tag: visual-question-answering
---

## MiniCPM-Llama3-V 2.5 int4
This is the int4 quantized version of [MiniCPM-Llama3-V 2.5](https://huggingface.co/openbmb/MiniCPM-Llama3-V-2_5).   
Running with int4 version would use lower GPU memory (about 9GB).


## Usage
Inference using Huggingface transformers on NVIDIA GPUs. Requirements tested on python 3.10:
```
Pillow==10.1.0
torch==2.1.2
torchvision==0.16.2
transformers==4.40.0
sentencepiece==0.1.99
accelerate==0.30.1
bitsandbytes==0.43.1
```

```python
# test.py
import torch
from PIL import Image
from transformers import AutoModel, AutoTokenizer

model = AutoModel.from_pretrained('openbmb/MiniCPM-Llama3-V-2_5-int4', trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained('openbmb/MiniCPM-Llama3-V-2_5-int4', trust_remote_code=True)
model.eval()

image = Image.open('xx.jpg').convert('RGB')
question = 'What is in the image?'
msgs = [{'role': 'user', 'content': question}]

res = model.chat(
    image=image,
    msgs=msgs,
    tokenizer=tokenizer,
    sampling=True, # if sampling=False, beam_search will be used by default
    temperature=0.7,
    # system_prompt='' # pass system_prompt if needed
)
print(res)

## if you want to use streaming, please make sure sampling=True and stream=True
## the model.chat will return a generator
res = model.chat(
    image=image,
    msgs=msgs,
    tokenizer=tokenizer,
    sampling=True,
    temperature=0.7,
    stream=True
)

generated_text = ""
for new_text in res:
    generated_text += new_text
    print(new_text, flush=True, end='')
```