fanzhihang

get 0.995 with 1024 * 1024

3

have changed the batch size to 1, any suggestion?

The big the better or what strategy?

Any plan to add unit tests?

4

Hi, I think this is a great project that gives many implementation and great guidance for number theory learning. But I think it would be better to add test cases...

### Is there an existing issue for this? - [X] I have searched the existing issues ### Current Behavior 从huggingface上下载lfs不稳定，但是清华云平台没有上传chatglm2-6b的tokenizer.model ### Expected Behavior 期望清华云平台上上传tokenizer.model，方便下载 ### Steps To Reproduce 直接复现即可 ###...