Hi, I’ve been using WandB in conjunction with the Hugging Face trainer for visualizing training and validation loss and accuracy during model fine-tuning. Generally, WandB creates a new run for each model fine-tuning, but I’ve encountered situations where this doesn’t consistently happen.
To address this, I decided to initialize WandB using the following line of code:
However, I noticed that this approach sometimes leads to the removal of all datapoints from that run in the WandB panel.
I would like to seek advice on how to ensure that WandB initializes the panel plot for each model and, at the same time, preserves the datapoints associated with that run. Are there any best practices or considerations that I might be overlooking?
I appreciate any help