Artifacts in log_table function

implementations
1

Hello all,

I’m using pytorch lightning with the Wandb logger. During the training I log table with a log_table function that I have overwritten, but the function remains almost the same :

    def log_table(self, log_name: str, columns: List[str], data: List[List[str]], step: Optional[int] = None):
        if self._save_tables:
            super(WandbLogger, self).log_table(key=log_name, columns=columns, data=data, step=step)

So when I log tables I obtain Wandb artifacts also, but I was wondering if there was a way to log them without these artifacts ?
(I am using pytorch-lightning 1.6.1 and wandb 0.13.1)

2

Hi

What kind of artifacts do you mean here? Do you mean the files being written to the wandb folder? I’m not sure, but I think this can’t be avoided as W&B handles it this way for upload.

If you want to clean up the local logs, you can use this command here:

wandb sync --clean ./
3

Hello,

Thank you for your response. I was speeking about artifacts as mentioned here : Storage - Documentation

artifacts | ~/.cache/wandb | the `WANDB_CACHE_DIR` environment variable

Because in wandb doc (Log Tables - Documentation) it seems that tables are just logged not as artifacts by default