You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently, there is no visually indication to a user that the pipeline is executed and which step it is currently running. Even when monitoring and tracing are enabled, the setup requires additional setup time as well as technical in-depth knowledge, while log information is presented in a highly technical format that isn't intuitive for quick status assessment.
There have been some similar requests regarding progress bars in the past, but they are related to specific components or aspects in pipelines (e.g. batch processing, see #2580). Some users also want to have the option to disable progress bars globally (#8782, #5098).
Describe the solution you'd like
A progress bar displayed when running pipeline.run() could
provide immediate quick, visual and real-time feedback
provide a time estimation and an overall status
complement monitoring and tracing
increase overall usability
From my point-of-view, I think it would be useful to have this information displayed when pipeline.run():
percentage completion
processing step indicator (step name, step number of total number of steps)
Could be integrated directly into the pipeline.run() logic, integrated as parameter to be turned on (True) or off (False) or as env var
Additional context
Hugging Face extensively uses progress bars with tqdm across its ecosystem for model downloading, training, inference, and data processing operations, demonstrating their value in workflows.
Is your feature request related to a problem? Please describe.
As a follow-up of this discussion about the
SuperComponent
abstraction with @sjrl, I think it would be great to have a progress bar displayed when executing a pipeline withpipeline.run()
.Currently, there is no visually indication to a user that the pipeline is executed and which step it is currently running. Even when monitoring and tracing are enabled, the setup requires additional setup time as well as technical in-depth knowledge, while log information is presented in a highly technical format that isn't intuitive for quick status assessment.
There have been some similar requests regarding progress bars in the past, but they are related to specific components or aspects in pipelines (e.g. batch processing, see #2580). Some users also want to have the option to disable progress bars globally (#8782, #5098).
Describe the solution you'd like
A progress bar displayed when running
pipeline.run()
couldFrom my point-of-view, I think it would be useful to have this information displayed when
pipeline.run()
:Describe alternatives you've considered
Could be integrated directly into the
pipeline.run()
logic, integrated as parameter to be turned on (True
) or off (False
) or as env varAdditional context
Hugging Face extensively uses progress bars with
tqdm
across its ecosystem for model downloading, training, inference, and data processing operations, demonstrating their value in workflows.Example HF progress bar:
https://stackoverflow.com/questions/74404985/huggingface-transformer-trainer-tqdm-progress-bar-not-moving-at-all-in-jupyter-n
The text was updated successfully, but these errors were encountered: