search

tianyi-lab / Cherry_LLM

[NAACL'24] Self-data filtering of LLM instruction-tuning data using a novel perplexity-based difficulty score, without using any other models
306 stars 21 forks source link

batch? #16

Closed sevenHQ closed 11 months ago

sevenHQ commented 11 months ago

How batch?

MingLiiii commented 11 months ago

Thanks for your interest! Due to the limitations of our computing resources, the GPU memory is not big enough to compute the loss of more than 1 sample, so we did not write the batch situation. However, I think it won't be so difficult to do so, we recommend you can do the customization on your own needs~