Michael IV
Michael IV

Reputation: 11424

Vertex AI API. Fine tuning the chat model

I am trying to launch Vertex AI fine tuning. I have followed the tutorial step by step:

  1. Created storage
  2. Uploaded dataset in .jsonl format
  3. Configured the model details etc.

One entry in my dataset file (jsonl) looks as follows:

{"messages": [{"author": "user", "content": "Tell a joke about flovers"}, {"author": "assistant", "content": "Roses are red, Violets are blue. I have a gun. Get in the van."}]}

When I hit "start tuning" button nothing happens. The manual says I should see the process in the pipelines window, but I see nothing there. I thought maybe my dataset is corrupted and tried also to use the sample dataset provided with the tutorial.Same thing- tuning process is not starting. What am I missing here?

enter image description here

Update:

I was able to fix the problem above.All I had to do was to enable all the related APIs. It is weird that GCP doesn't provide any kind of warning during fine tuning startup. enter image description here

Now, I have the fine tune pipeline job running but failing with the following error:

enter image description here

Upvotes: 3

Views: 827

Answers (1)

kiran mathew
kiran mathew

Reputation: 2343

Posting the above comments as an answer to help community members in their research.

Issue 1:When trying to launch Vertex AI fine tuning chat model by clicking the "start tuning" button nothing happens.

Solution: The above issue was solved after enabling API’s.On the Vertex AI main page there is a blue button "enable all recommended APIs". Once it clicked,the fine tuning setup started working.

Issue 2: Error message:code=RESOURCE_EXHAUSTED

Solution: This type of error generally happens when the requested resource is not in that region or reaches the maximum quota. We can check the regions available of the resource in the Quotas page within your project. go to "IAM & Admin" > "Quotas" then go find the resource at the search bar.

For more information refer to these link1,link2 and link3.

Posting the answer as community wiki for the benefit of the community that might encounter this use case in the future.

Feel free to edit this answer for additional information.

Upvotes: 1

Related Questions