Weights and Biases with RLLIB error

How severe does this issue affect your experience of using Ray?

  • Medium: It contributes to significant difficulty to complete my task, but I can work around it.


I am on Ray 2.4

when I try to run the official example for WandB integration I get:

 File "/opt/anaconda3/lib/python3.8/multiprocessing/reduction.py", line 60, in dump
    ForkingPickler(file, protocol).dump(obj)
  File "/opt/anaconda3/lib/python3.8/site-packages/ray/air/integrations/wandb.py", line 367, in __reduce__
    raise RuntimeError("_WandbLoggingProcess is not pickleable.")
RuntimeError: _WandbLoggingProcess is not pickleable.

I’ve also tried to hookup WandB with RLLIB and got the same error. Here is simple example replicating the error:

I could not reproduce the bug with the Colab given. Could you please retry it? If it still doesn’t work, try with the Ray nightly wheel, and if that doesn’t work, please file a bug report on the Github issues.

Hello Ok, so here is my report of the situation with RLLIB-WandB.

  1. On Ray 2.4 → Everything works. The Callbacks and the API

  2. On Ray < 2.4 → Only the API works .There is a fix, shared by someone on the Ray Slack that makes the Callbacks work in case you are using Ray<2.4

Thank you very much!