Rulin Shao comments

Results 7 comments of


                                            Rulin Shao

Can't reproduce UniT results with gradient accumulation

I could load the saved checkpoint and resume training, the NaN doesn't seem to appear in the same iteration, instead, it appears every 16900 iterations. I.e., I resumed the training...

How to convert the predicted bounding boxes to actual predicted text file.

same question

Any reference results for rag evaluation?

Hi @hkvision ! Thanks for your interest in using our repo! We did not evaluate on qwen2-1.5B, but we had 34.6% EM with Llama2-7B using DPR-Wiki as the datastore. See...

Any reference results for rag evaluation?

In our paper, we used 5-shot evaluation, which may be important for models that sometimes do not follow instructions. We used 3 documents, 5 shot for RAG evaluation. 12.02 seems...

Any reference results for rag evaluation?

Yes, you're right. Good luck!

Version of numpy

Hi! I was using `numpy==1.26.4`, would you like to try with this version?

Version of numpy

Closing this issue, but let me know if you have further questions!