[Bug][MetaScheduler] InternalError: Check failed: original_producers.size() == 1u (0 vs. 1)

[twmht]

Thanks for participating in the TVM community! We use https://discuss.tvm.ai for any general usage questions and discussions. The issue tracker is used for actionable items such as feature proposals discussion, roadmaps, and bug tracking. You are always welcomed to post on the forum first 😸

Issues that are inactive for a period of time may get closed. We adopt this policy so that we won't lose track of actionable issues that may fall at the bottom of the pile. Feel free to reopen a new one if you feel there is an additional problem that needs attention when an old one gets closed.

Expected behavior

Optimize Resnet50 with meta scheduler

Actual behavior

What actually happened

Environment

TVM main branch commit: 95a2fff

Steps to reproduce

https://gist.github.com/twmht/d719f7e4b3ccaa8b269b7dd7475e023d

https://gist.github.com/twmht/1dc0c41a80478c2de6f215824a7b8e14

Triage

Please refer to the list of label tags here to find the relevant tags and add them below in a bullet format (example below).

• needs-triage

[krishnab30]

Hi @twmht , good day to you.

I tried executing your script and I too, get the following error :

InternalError: Check failed: original_producers.size() == 1u (0 vs. 1) :

I traced the error to its source and it seems to be from the ToMixedPrecision() Pass. Has something to do with float16 handling by the metascheduler. Were you able to solve this? Or atleast find out the exact nature and source of this bug?

Can you please let me know if you were able to find a workaround for this? TIA and have a nice day!

Regards,

Krishna

[krishnab30]

I also tried your script with other classifiers like mobilenetV2 and VGG16. I get the same InternalError in all the cases. The problem seems to be with the comaptibility between the metascheduler API and the 'float-16' datatype.

Just wanted to inform you that this error persists across multiple classifier models. It would be of great benefit if you could take a look at this issue again and provide your insights.
Thanks, have a nice day.