Weights and Biases for Logtables

[LukasFehring]

• PyExperimenter version: Any
• Database provider: Any
• Python version: Any supported
• Operating System: Any

Description

I would like functionality that syncs Weights and Biases with log tables for data visualization. Alternatively, another dashboard would do.

Proposed Approach

The function receives the name of a logtable. It then joins the logtable with the main experiment table on the experimentid and filters for finished experiments. The function that loops over each of these experiments

1. Extract keyfields and call wandb.init with their information
2. Go through the logtables and call wandb.log for every entry
3. Lastly call wandb.log for the main tables resultfields