vllm - 💡(How to fix) Fix [CI Failure]: mi325_1: Spec Decode Draft Model [2 comments, 1 participants]

TL;DR

Investigate the mi325_1: Spec Decode Draft Model test group to identify the cause of the failing tests on AMD hardware.

Guidance

• Review the test results from the latest nightly build (#7791) to understand the specific failures in the mi325_1: Spec Decode Draft Model test group.
• Check the hardware status table for any patterns or correlations between hardware and test failures.
• Analyze the sync_ready_tickets.py script to ensure it is correctly managing the test group and updating the status accordingly.
• Investigate the test history to see if there are any common factors or changes that may have contributed to the failures, given the break frequency of 3 pass-fail flips in the last 60 days.

Notes

The provided information lacks specific details about the test failures, making it challenging to provide a more targeted solution. Further investigation into the test results and hardware configuration is necessary to determine the root cause.

Recommendation

Apply workaround: Investigate and debug the mi325_1: Spec Decode Draft Model test group to identify the cause of the failures, as there is no clear indication of a fixed version or straightforward solution.