Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

wandb.errors.RequireError: require() unsupported requirement: report-editing #162

Closed
asmith26 opened this issue Jan 3, 2023 · 10 comments
Closed
Labels
bug Something isn't working

Comments

@asmith26
Copy link

asmith26 commented Jan 3, 2023

🐛 Describe the bug

Hi there,

I've successfully ran python examples/ppo_sentiments.py, so I though I'd try Use Ray Tune to launch hyperparameter sweep but I'm getting an error I can't seem to fix:

$ python -m trlx.sweep --config configs/sweeps/ppo_sweep.yml examples/ppo_sentiments.py

wandb: WARNING require() unsupported requirement: report-editing
wandb: ERROR Supported wandb.require() features can be found at: https://wandb.me/library-require
Traceback (most recent call last):
  File "~/miniconda3/envs/CarperAI/lib/python3.10/runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "~/miniconda3/envs/CarperAI/lib/python3.10/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "~/git/CarperAI/trlx/trlx/sweep.py", line 15, in <module>
    from trlx.ray_tune.wandb import log_trials, create_report
  File "~/git/CarperAI/trlx/trlx/ray_tune/wandb.py", line 11, in <module>
    wandb.require("report-editing")
  File "~/miniconda3/envs/CarperAI/lib/python3.10/site-packages/wandb/sdk/wandb_require.py", line 89, in require
    f.apply()
  File "~/miniconda3/envs/CarperAI/lib/python3.10/site-packages/wandb/sdk/wandb_require.py", line 68, in apply
    raise RequireError(last_message)
wandb.errors.RequireError: require() unsupported requirement: report-editing

Many thanks for any help! :)

Which trlX version are you using?

Installing from main (i.e. pip install -e .

Additional system and package information

wandb=0.13.7; python=3.10.8; Ubuntu

@asmith26 asmith26 added the bug Something isn't working label Jan 3, 2023
@LouisCastricato
Copy link
Contributor

@ayulockin

@ayulockin
Copy link
Contributor

Will make a PR to fix it. Thanks for reporting @asmith26 :)

@ayulockin
Copy link
Contributor

Hey @asmith26 can you try to comment out this line in case you had cloned the repo and working from it.

@jon-tow
Copy link
Collaborator

jon-tow commented Jan 3, 2023

@ayulockin Commenting out the linked line and updating some of the configs seems to have resolved the issue at the following branch: jon-tow@b99c9ce

  • A manual report for the PPO sentiments example, here.
  • Unfortunately, it seems that the automated report does not render the parallel coordinates plot properly, see here. Is this intended behavior and how it was produced previously? Thanks!

@ayulockin
Copy link
Contributor

Hey @jon-tow the parallel coordinate plot is rendering properly for me.

image

Am I missing something?

@jon-tow
Copy link
Collaborator

jon-tow commented Jan 4, 2023

Sorry, I accidentally deleted the automated report from:

def create_report(project_name, param_space, tune_config, trial_path, best_config=None):

See the updated link here, and more specifically, the following plot:
Screenshot 2023-01-04 at 03 36 50

@LouisCastricato
Copy link
Contributor

Oh that's kinda weird

@ayulockin
Copy link
Contributor

@jon-tow when I try to edit the report manually I can totally see the values being logged and thus I was able to manually fix it. Not sure why it's not rendering properly in the first place. Sharing with the team internally. I will let you all know when I learn something.

Can you let me know which browser and which wandb version were you using?

@jon-tow
Copy link
Collaborator

jon-tow commented Jan 4, 2023

wandb==0.13.7
Chrome Version 108.0.5359.124 (arm64)

Thank you!

LouisCastricato pushed a commit that referenced this issue Jan 10, 2023
* check wandb version

* Update wandb.py

* add wandb version + mean/reward

* fix api changes

* Update ILQL sweep hparams

* Run `isort`

* Remove `return`s from examples to avoid RayTune erros

* Skip `isort` on wandb reports import

* Remove erroring `tqdm` call in `BaseRLTrainer.evaluate`

Co-authored-by: jon-tow <jonathantow1@gmail.com>
@LouisCastricato
Copy link
Contributor

This is now resolved.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug Something isn't working
Projects
None yet
Development

No branches or pull requests

4 participants