issue-tracking icon indicating copy to clipboard operation
issue-tracking copied to clipboard

How to tweak contents of the 'Reproduce'

Open kirilllzaitsev opened this issue 10 months ago • 4 comments

Before Asking:

  • [+] I have searched the Issue Tracker.
  • [+] I have searched the Documentation.

What is your question related to?

  • [ ] Comet Python SDK
  • [+] Comet UI
  • [ ] Third Party Integrations (Huggingface, TensorboardX, Pytorch Lightning etc.)

What is your question?

I haven't found any docs on how to configure the contents of the Reproduce tab in the experiment UI. For example, on the screenshot below I have proper git information (masked out), which I set up thanks to a warning in the log. There is no warning on how to provide the Run information. image

Could you provide a solution on how to get the Run information?

Also, I think it could be beneficial to display the necessary information to set up UI elements straight in the UI.

Code

exp = comet_ml.Experiment(
        api_key=os.environ["COMET_API_KEY"],
        auto_output_logging="simple",
        auto_metric_logging=True,
        auto_param_logging=True,
        log_env_details=True,
        log_env_host=False,
        log_env_gpu=True,
        log_env_cpu=True,
        log_code=False,
        parse_args=True,
        display_summary_level=0,
)

What have you tried?

Searching Google for the ways to configure 'Reproduce'

kirilllzaitsev avatar Mar 25 '24 18:03 kirilllzaitsev

@kirilllzaitsev , how are you running your code? Is it in a Python script, notebook, or other environment? Is the code run by a job scheduler?

dsblank avatar Mar 26 '24 11:03 dsblank

@dsblank the code is executed as a script via bash command

kirilllzaitsev avatar Mar 26 '24 11:03 kirilllzaitsev

Could you send me a link to the Comet experiment? It is ok if it is private, or you can create a "shareable" link. You can email me at [email protected]

Screenshot from 2024-03-26 08-26-53

dsblank avatar Mar 26 '24 12:03 dsblank

You can change/set the Run area of the Reproduce dialog with:

api = API()
exp = api.get_experiment(WORKSPACE, PROJECT, EXPERIMENT)
exp.set_command(["shell.sh", "python", "arg1", "--flag", "value"])

which would give you:

Screenshot from 2024-05-10 15-53-56

dsblank avatar May 10 '24 20:05 dsblank