Fine Tune Model Job Fails

Question

Hello,

I am trying to create a fine tune model for classification. I am creating the job from Python code with following args

create_args = {
    "training_file": training_id,
    "validation_file": validation_id,
    "model": "curie",
    "compute_classification_metrics": True,
    "classification_n_classes": 3
}

The job is submitted and getting failed. There are no logs to see the reason.

Interestingly using the same training and validation file and creating a job from the Azure OpenAI studio, the job succeeded but there is no option to put the args as below to get the F1 score, accuracy etc

  "compute_classification_metrics": True,

    "classification_n_classes": 3

User's image

Answer

To enable logs for the Azure OpenAI API, you need to set the "log_level" argument to "debug" when calling the API. This will enable detailed logs for the API, which can be useful for troubleshooting errors and issues.

For example, you can set the "log_level" argument as follows:

create_args = { "training_file": training_id, "validation_file": validation_id, "model": "curie", "compute_classification_metrics": True, "classification_n_classes": 3, "log_level": "debug" }

This will enable detailed logs for the API.

Answer

Hello,

I am having the same issue - Fine tuning fails with error "Fine failed during post processing"

Steps attempted:

1 - Copied notebook from here, added my api key and endpoint, ran it sequentially until fine tune job. Notebook was executed on Databricks Runtime 11.3

2 - Started a curie/babbage fine-tuning job from the portal with the notebook demo data, both failed.

3 - Ran the notebook linked above with option {log_level": "debug"} enabled. Status shows "succeeded" in Databricks, and failure in the portal (Double checked model names match)

Resource Details:

Pricing tier: Standard

Location: South Central US

Please advise

Share via

Fine Tune Model Job Fails

2 answers

Your answer