Add More Wandb Logging for Unsupervised Learning
Improve logged information in unsupervised experiments (to wandb).
Our benchmark experiments that rely on pretrained models and only evaluate inference, already log a lot of statistics to Weights and Biases (wandb). We use these logs to update our benchmark tables and track changes in Monty's performance.
However, for the unsupervised learning benchmarks, we don't have good wandb logging set up and analysis of results needs to happen manually, based on the saved .csv file (see instructions here). It would improve workflows significantly to also have the wandb logging handler send the necessary statistics to wandb during unsupervised learning and inference experiments.
For further details and discussion, see this Discourse thread
Help Us Make This Page Better
All our docs are open-source. If something is wrong or unclear, submit a PR to fix it!
Updated 9 days ago