Based on the sample code, I want to delete some of my checkpoints. However, an error occurred.
File c:\Users\XXXX\miniconda3\envs\py10\lib\site-packages\wandb\apis\public\files.py:204, in File.delete(self)
203 if self._server_accepts_project_id_for_delete_file():
→ 204 variable_values[“projectId”] = self.run._project_internal_id
205 project_id_variable_fragment = “, $projectId: Int”
File c:\Users\XXXX\miniconda3\envs\py10\lib\site-packages\wandb\apis\attrs.py:51, in Attrs.getattr(self, name)
50 else:
—> 51 raise AttributeError(f"{repr(self)!r} object has no attribute {name!r}")
AttributeError: ‘<Run XXXX (finished)>’ object has no attribute ‘_project_internal_id’
During handling of the above exception, another exception occurred:
Can anyone help me? Thanks
Yeah their entire api stuff is just broken for no reason. They probs discourage using it cos its extra requests to their servers. Weirdly using wandb.ai website seems fine other than it being painfully slow. In the end I had to get my students to do it - apparently wandb’s python api uses some deprecated graphQL stuff that made it so you can only delete your data properly over the website? (although permanently deleting actual files seems impossible). Got very pissed off and the phd students ended up pushing me to use mlop.ai which was free and seem a lot more respecting of my data. They got all my data off wandb’s servers using nothing but my api key, which was kinda insane. Will probably be using mlop for non llm stuff going forward considering it’s foss too GitHub - mlop-ai/mlop: Next Generation Experimental Tracking for Machine Learning Operations