Add 'validation_split_percentage' and 'evaluation_strategy' parameters for Trainers

OptimalScale / LMFlow

An Extensible Toolkit for Finetuning and Inference of Large Foundation Models. Large Models for All.

https://optimalscale.github.io/LMFlow/

Apache License 2.0

8.11k stars 819 forks source link

Add 'validation_split_percentage' and 'evaluation_strategy' parameters for Trainers #823

Open smirn0v opened 1 month ago

smirn0v commented 1 month ago

Is your feature request related to a problem? Please describe. Currently 'validation_split_percentage' is set to 0 and changing it to any other value will have no effect cause 'evaluation_strategy' not set at all and by default 'no'

Describe the solution you'd like Please add 'validation_split_percentage' and 'evaluation_strategy' parameters to control training process.

research4pan commented 1 month ago

Thanks for your contribution to LMFlow! We will schedule this feature in our next version 😄