search

zyang1580 / CoLLM

The implementation for the work "CoLLM: Integrating Collaborative Embeddings into Large Language Models for Recommendation".
BSD 3-Clause "New" or "Revised" License
47 stars 6 forks source link

ValueError: Input contains NaN. #11

Closed manjusaka-L closed 1 month ago

manjusaka-L commented 1 month ago

When I run the program on Amazon Book to 11 epochs, the loss becomes nan. All the settings are default. The GPU is A100 *1.

Can you give some suggestions?

manjusaka-L commented 1 month ago

BTW, can we access the pre-trained weights?

zyang1580 commented 1 month ago

BTW, can we access the pre-trained weights?

I will upload the pre-trained weights when I am free.

zyang1580 commented 1 month ago

When I run the program on Amazon Book to 11 epochs, the loss becomes nan. All the settings are default. The GPU is A100 *1.

Can you give some suggestions?

What learning rate are you using? You might try tuning it down.

manjusaka-L commented 1 month ago

Thanks for your suggestions!