search

Thytu / SMIT

SMIT: A Simple Modality Integration Tool
MIT License
15 stars 3 forks source link

WandB projects and runs grouping should be reworked #2

Open Thytu opened 5 months ago

Thytu commented 5 months ago

In order to allow user to better find and filter their runs, each run should be named based on the model used.

One option could be {decoder_name}_{speech_encoder_name} (i.e abacaj/phi-2-super_facebook/hubert-large-ls960-ft).

However this format does not allow user to filter based on dataset/tasks used to train their models.

Thytu commented 5 months ago

I just saw that wandb provides group & job_type parameters that could be used.

Maybe instead of having two project (SMIT-Training & SMIT-Pretraining) we could have only one project named SMIT containing runs separated into groups (one per encoder/decoder duos) with the job_type being set either as training or pretraining.