Questions about parameters, and what is the the setting that is the MOST COSTLY? #87
Unanswered
AIhasArrived
asked this question in
Q&A
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hello there,
I opend DLAS UI and I see a bunch of paramters that I am not sure what they mean yet:
FP16: this obviousely reduces precision and makes the training faster.
WandB: what is this?
Gradient Checkpointing?
TB Logger?
What is the minimum number of train and valid files that you were able to get good results with?
What learning speed worked out best for you?
Same questions for all other parameters.
What are the settings that were the most constly but the are the most effective to get good results?
Thank you
Beta Was this translation helpful? Give feedback.
All reactions