Not able to get any result from the fine tuning job

Question

Parth Shah 0

We are facing the issue with fine tuning of the model. We are trying to achieve the model customization on our data.

We are following https://learn.microsoft.com/en-us/azure/ai-services/openai/tutorials/fine-tune?tabs=command-line to fine tune the model. Job is not showing any progress and stuck on the steps

We tried to execute it multiple times but results are same.
We need help on this as it is impacting our business.

YutongTie-MSFT 53,951 Reputation points

2025-05-01T19:46:32.5833333+00:00

Hello @Parth Shah Thanks for reporting this issue, what is the status showing for you? Is that keeping showing as not started?
Manas Mohanty 3,210 Reputation points Microsoft External Staff

2025-05-02T11:37:37.05+00:00

Hi Parth Shah

Sorry for the inconvenience.

The suggested step would be to cancel the finetuning job and then check format and size of Jsonl file before submitting a new finetune job.

Cancel through Rest api

Here are suggested steps to check

Fileformat should be UTF-9 BOM

Token Size should be under Batch Quota limit of model,

Please increase your TPM of Batch model deployment.

Switch to other available region.

Please test with sample Json file to find out formatting issue or backend issue

Adjust your hypermeters like batch_size, learning_rate_multiplier, n_epochs etc as mentioned here.

Debugging

Please check your finetuning logs for further leads

Reference thread

Hope it helps.

Thank you.

Your answer

YutongTie-MSFT 53,951 Reputation points

2025-05-01T19:46:32.5833333+00:00

Hello @Parth Shah Thanks for reporting this issue, what is the status showing for you? Is that keeping showing as not started?
Manas Mohanty 3,210 Reputation points Microsoft External Staff

2025-05-02T11:37:37.05+00:00

Hi Parth Shah

Sorry for the inconvenience.

The suggested step would be to cancel the finetuning job and then check format and size of Jsonl file before submitting a new finetune job.

Cancel through Rest api

Here are suggested steps to check

Fileformat should be UTF-9 BOM

Token Size should be under Batch Quota limit of model,

Please increase your TPM of Batch model deployment.

Switch to other available region.

Please test with sample Json file to find out formatting issue or backend issue

Adjust your hypermeters like batch_size, learning_rate_multiplier, n_epochs etc as mentioned here.

Debugging

Please check your finetuning logs for further leads

Reference thread

Hope it helps.

Thank you.