Improved Evaluations - Githubissues

fixie-ai / ultravox

MIT License

618 stars 24 forks source link

Closed farzadab closed 1 month ago

farzadab commented 1 month ago

This PR does the following:

Bugfix: AnyInstruct was not using val/train splits (as it had somehow gotten removed from the last PR)
Faster evaluations by using DDP multiprocessing for text generation
Validation loss added by default to training

farzadab commented 1 month ago

Small training run to check for regressions seems to have run fine: https://wandb.ai/fixie/ultravox/runs/q3rh3yxv