Repository navigation
How to re-run only failed mini-batches in AML pipeline with prompt flow component ? #3102
Megha Patil (meghapatilcode)
started this conversation in
General
Replies: 1 comment
|
Alain LI (@alainli0928) do we have such support for parallel component? |
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
We are using prompt flow as a parallel component in our AML pipelines. Prompt flow is run for each input row from a JSONL file in mini-batches. The mini-batch workloads are set to fail after reaching the "mini batch max retries". The AML pipeline is however set to not fail despite failed mini-batches by keeping the default value for "mini batch error threshold" as -1. We are unable to find a way to re-run the pipeline for only the failed mini-batches. Also, we are unable to find a direct way to isolate and download the records for which mini-batch workloads have failed. How can we re-run the AML pipeline such that only the failed mini-batches are run again?
All reactions