Update README.md
Browse files
README.md
CHANGED
|
@@ -12,4 +12,15 @@ pipeline_tag: text-generation
|
|
| 12 |
|
| 13 |
## 使用
|
| 14 |
|
| 15 |
-
参数加入 `kv_cache_dtype="fp8"`
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
|
| 13 |
## 使用
|
| 14 |
|
| 15 |
+
参数加入 `kv_cache_dtype="fp8"`
|
| 16 |
+
|
| 17 |
+
## 评估
|
| 18 |
+
|
| 19 |
+
使用 [lm-evaluation-harness](https://github.com/EleutherAI/lm-evaluation-harness/tree/7ad7c5b9d0f1c35c048af0ce8b197ebc2021dbd3) + vLLM serve 进行评估：
|
| 20 |
+
|
| 21 |
+
|项目|Qwen2-7B-Instruct|Qwen2-7B-Instruct-FP8-CN|Recovery|此项目|Recovery|
|
| 22 |
+
|---|---|---|---|---|---|
|
| 23 |
+
|ceval-valid|81.87|**81.65**|**99.73%**|81.35|99.36%|
|
| 24 |
+
|cmmlu|81.78|**81.26**|**99.36%**|81.19|99.28%|
|
| 25 |
+
|agieval_logiqa_zh (5 shots)|47.63|48.54|101.91%|46.54|97.71%|
|
| 26 |
+
|**平均**|**70.43**|**70.48**|**100.07%**|69.69|98.95%|
|