Wandb.init resume can't find previous run

W&B Help
1

I’m attempting to resume a run with

wandb.init(project="backgammon", id="5ii86afx", resume="must")

but I get the following error:

You provided an invalid value for the resume argument. The value ‘must’ is not a valid option for resuming a run (5ii86afx) that has not been initialized. Please check your inputs and try again with a valid run ID. If you are trying to start a new run, please omit the resume argument or use resume='allow'.

This happens from either a python script or a Jupyter notebook. There’s not any other useful information in the debug logs. This run definitely exists though. I’ve triple-checked the ID, I can go to the web page, and accessing it via the API works:

api = wandb.Api()
run = api.run("gcmn/backgammon/5ii86afx")

I have resumed this run before.

The issue started happening just after I accessed and edited the run via the API, so I’m wondering if something got corrupted. I edited it in a Jupyter notebook like this:

api = wandb.Api()
run = api.run("gcmn/backgammon/5ii86afx")
run.summary["session"] = 21
run.update()

Side note: I’m futzing around editing the run in the API in the first place because wandb doesn’t have good a way to add new metrics after a training run and when I did so it messed up summary metrics. Basically, I have the same use case as Log custom metrics for a run outside of the training loop and am already using a custom x-axis, but it’s still pretty janky.

2

Hello Geoffrey:
Thank you very much for reporting this issue to us. I have tried out your workflow above and was able to recreate your problem. I’m very sorry about this bug. I will escalated it to our engineer team and keep you posted on any update related!
Best,
W&B