Hi there,
I am trying to download some information from some runs I’ve already performed. The information is logged as a table of variables that were used to train the models.
I’ve used the api to get the list of runs that are relevant for me, then in a loop I’m looking at the artifacts for each run.
When I look at the logged tables with run.summary['training_vars']
it shoes me what looks like actual paths of where the artifact/table is stored:
{'_latest_artifact_path': 'wandb-client-artifact://ntblxo22qc1984gme2whsu6zxhody5jxplzg1jvpt9g979avqt6qjrh80nuhmmtggalakd932412d6suzipk26pxjdc30mvnw96qjutxpgzf7phtdq4jmigj66otqhwl:latest/training_vars.table.json', '_type': 'table-file', 'artifact_path': 'wandb-client-artifact://6ldyof2zj3f4xtjr2td2oqa89h5uft7nwjn1a3bhzujrkkd5duo08wiilj44wycd61a5h6pjyc10z0xdpxe0ntxxyup5qjes36yp1ayhb9rpgvffzub2g7m382xbh5ir/training_vars.table.json', 'ncols': 1, 'nrows': 10000, 'path': 'media/table/training_vars_0_df2c1ab6a95963ca2dd2.table.json', 'sha256': 'df2c1ab6a95963ca2dd22d8e915540f4f46630a13b797cd95d98737695fa92bc', 'size': 124652}
Is there anyway I can instruct the api to download the table directly from those paths? That would save me having to join together the entity, project, and alias name. It also seems obvious to do it this way as the paths are right there in the run summary.
Thanks