如何跳过推理仅进行测评？
#626
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
{{ message }}
-
我想在一个非 GPU 的硬件上测试各个数据集上的推理精度，目前想到的方案是保存推理结果，再调用 OpenCompass 的接口进行测评。请问这样是否可行？数据保存的格式和测评步骤是否有参考资料？
Beta Was this translation helpful? Give feedback.
All reactions