During the execution of a bulk test in Azure prompt flow, the test is failing within 0.3 seconds and the only message that appears states, "Variant_0 is the default run generated for the entire flow when none of the LLM nodes have additional variants." This seems to occur across all the evaluation metrics I have tested.
I understand from the nature of the message that it might have something to do with the requirement of adding additional variants to the LLM nodes. Yet, I am currently unaware of where to precisely place these variants during bulk test and how to effectively solve this issue. Does anyone have any advice?
Could anyone please explain what this error message is suggesting, and suggest a method to handle it? Any illustrations or code samples would be greatly appreciated.
Thanks in advance for your help.
- I tried different evaluation methods, tried with and without evaluation
- Changed data format (csv and json)
In all the cases, I kept getting the same error.
The error message suggests that you need to add additional variants to the LLM nodes during bulk test.
To add variants to your nodes, you can click on show variants on your specific node.
Then you clone the variant and make changes as per your requirement.
Once you have multiple variants you can select them in the basic settings of your batch run.
For more details regarding variants, you can check this documentation.