Open 0xfabioo opened 1 month ago
Thanks for the feedback! We are routing this to the appropriate team for follow-up. cc @Azure/azure-ml-sdk @azureml-github.
Following, is there any updates here?
Curious about the status of this as well. Currently running into this issue when attempting to run a Phi-3 training job.
Describe the bug Attempting to log more than 200 parameters in a job fails with the following error message:
This limitation is not documented at the provided link (https://aka.ms/azure-machine-learning-limits), nor could I find any reference to it elsewhere.
This limitation appears to originate from Azure Machine Learning (AML) itself, as the same code works correctly when using a local MLflow instance.
To Reproduce Steps to reproduce the behavior:
Expected behavior It should be possible to log more than 200 parameters in a single job. If this limit is intentional, it should be clearly documented. Additionally, a limit of 200 parameters is quite low and can be reached quickly with the current state-of-the-art network architectures.
Screenshots If applicable, add screenshots to help explain your problem.
Additional context Add any other context about the problem here.