mirror of
https://github.com/QwenLM/Qwen.git
synced 2026-05-21 00:45:48 +08:00
fix format problems in evaluation code; update ceval extraction rules
This commit is contained in:
@@ -34,6 +34,19 @@ pip install thefuzz
|
||||
python evaluate_chat_mmlu.py -d data/mmlu/data/
|
||||
```
|
||||
|
||||
- CMMLU
|
||||
|
||||
```Shell
|
||||
wget https://huggingface.co/datasets/haonan-li/cmmlu/resolve/main/cmmlu_v1_0_1.zip
|
||||
mkdir data/cmmlu
|
||||
mv cmmlu_v1_0_1.zip data/cmmlu
|
||||
cd data/cmmlu; unzip cmmlu_v1_0_1.zip
|
||||
cd ../../
|
||||
|
||||
# Qwen-7B
|
||||
python evaluate_cmmlu.py -d data/cmmlu/
|
||||
```
|
||||
|
||||
- HumanEval
|
||||
|
||||
Get the HumanEval.jsonl file from [here](https://github.com/openai/human-eval/tree/master/data)
|
||||
|
||||
Reference in New Issue
Block a user