The time cost for training this instructor?

[afalf]

Hi, thanks for your great work. I do not see the time cost for training this model on the paper. Could you public the time cost for training the different size models(base, large, xl)?

[Harry-hash]

Hi, thanks a lot for your interest in INSTRUCTOR!

It takes around 6, 12 and 30 hours to train base, large and xl-sized models respectively.

Feel free to leave any further question or comment here!

[afalf]

Hi, thanks a lot for your interest in INSTRUCTOR!

It takes around 6, 12 and 30 hours to train base, large and xl-sized models respectively.

Feel free to leave any further question or comment here!

Thanks a lot for your reply. But I tried to run the train.py following README, and find that it needs about 300 hours to train the gtr-base model under 4 A100 GPUs.... What should I do to check it?

[Harry-hash]

Hi, thanks a lot for your question!

We have prepared much more training data here. It may not be necessary to complete all the training steps. In our experiments, we stopped tuning INSTRUCTOR-base at around 40K steps.

Hope this helps! Feel free to leave any further question or comment here!

[afalf]

Hi, thanks a lot for your question!

We have prepared much more training data here. It may not be necessary to complete all the training steps. In our experiments, we stopped tuning INSTRUCTOR-base at around 40K steps.

Hope this helps! Feel free to leave any further question or comment here!

Okay, thanks a lot!

[Harry-hash]

Feel free to reopen this issue and add any following comments!

[yangjianxin1]

what is the batch size to train the instructor? Is 32 enough?