search

OpenBMB / MiniCPM-V

MiniCPM-V 2.6: A GPT-4V Level MLLM for Single Image, Multi Image and Video on Your Phone
Apache License 2.0
12.74k stars 893 forks source link

[BUG] <title> Multi turns conversation 's output is not complete! #663

Open FantasticZihao opened 2 weeks ago

FantasticZihao commented 2 weeks ago

是否已有关于该错误的issue或讨论？ | Is there an existing issue / discussion for this?

该问题是否在FAQ中有解答？ | Is there an existing answer for this in FAQ?

当前行为 | Current Behavior

When I used the Minicpm-llama3-v2.5 to make multi turns conversation for one image, I found the model's output is not complete.

期望行为 | Expected Behavior

I appreciate the model's complete answer when doing multi turns conversation based on a single image for one dataset.

复现方法 | Steps To Reproduce

No response

运行环境 | Environment

- OS:
- Python:
- Transformers:
- PyTorch:
- CUDA (`python -c 'import torch; print(torch.version.cuda)'`):

备注 | Anything else?

No response

FantasticZihao commented 2 weeks ago

Thanks for your excellent work! I am eager to hearing your reply.

LDLINGLINGLING commented 2 weeks ago

hello ,please show me your example.

FantasticZihao commented 2 weeks ago

hello ,please show me your example.

code 代码 captions captions As is shown, sometimes the answer is not complete.

FantasticZihao commented 1 week ago

hello ,please show me your example.

Hello, can you give me some suggestions?

LDLINGLINGLING commented 1 week ago

hello，Do the following problems occur frequently? I think you can try some different sampling arguments。