ai-models/Qwen3-VL-8B-Thinking
2
0
Fork 0
mirror of https://www.modelscope.cn/Qwen/Qwen3-VL-8B-Thinking.git synced 2026-04-02 13:02:53 +08:00
Code Issues Packages Projects Releases Wiki Activity

Update README.md

Browse Source
This commit is contained in:
Cherrytest
2025-10-14 12:35:54 +00:00
parent 0bd00a954f
commit f3e6023a32
1 changed files with 25 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

27
README.md
View File

@ -58,10 +58,10 @@ This is the weight repository for Qwen3-VL-8B-Thinking.
**Multimodal performance**
![](https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen3-VL/table_thinking_vl_8b.jpg)
![](https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen3-VL/qwen3vl_4b_8b_vl_thinking.jpg)
**Pure text performance**
![](https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen3-VL/table_thinking_text_8b.jpg)
![](https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen3-VL/qwen3vl_4b_8b_text_thinking.jpg)
## Quickstart
@ -128,6 +128,29 @@ output_text = processor.batch_decode(
print(output_text)
```
### Generation Hyperparameters
#### VL
```bash
export greedy='false'
export top_p=0.95
export top_k=20
export repetition_penalty=1.0
export presence_penalty=0.0
export temperature=1.0
export out_seq_length=40960
```
#### Text
```bash
export greedy='false'
export top_p=0.95
export top_k=20
export repetition_penalty=1.0
export presence_penalty=1.5
export temperature=1.0
export out_seq_length=32768 (for aime, lcb, and gpqa, it is recommended to set to 81920)
```
## Citation