Why cannot using the same dataset run on different models?

[wuQi-666]

Can the C-Eval dataset be used to run on the qwen-7b-Chat dataset, but there are no running results on Baichun2-13B-Chat?

对Baichuan2-13B-Chat的评测一直显示如上图所示，查看显卡内存也没有变化

[Leymore]

Please try to run Baichun2-13B-Chat with an addition --debug option to show more logs.

You may also need to check your transformers version as Baichun2-13B-Chat cannot run properly with transformers==4.34.0

[wuQi-666]

In addition, debug using the command line, as follows: CUDA_VISIBLE_DEVICES=2 python run.py --models hf_baichuan2_13b_chat --datasets ceval_gen_5f30c7 --debug, the final run result is empty.

Some of them reported errors, as shown in the following figure:

[Leymore]

The prediction is missing due to inference stage failure. I find this totally understandable.

[wuQi-666]

how can i fix it ?

[Leymore]

Please show the log in inference stage?

[wuQi-666]

I'm very sorry, the following error was caused by myself. When I just ran, the graphics card was not enough, which caused the above inference stage error. It is indeed possible to run correctly after changing the Transformers version. Thank you.