Build models expecting the operator support of onnxruntime >=1.26.0 - #235
Open
Christian Bourjau (cbourjau) wants to merge 6 commits into
Open
Christian Bourjau (cbourjau) wants to merge 6 commits into
Christian Bourjau (cbourjau) wants to merge 6 commits into
Conversation
Christian Bourjau (cbourjau)
requested a review
from Sam Gilmour (sgilm)
as a code owner
September 8, 2026 12:14
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
We deploy many workarounds for bugs and missing kernel support in onnxruntime. Currently, we build models targeting onnxruntime >= 1.20.1. We couldn't update this for a long time due to various regressions upstream. However, onnxruntime >=1.26.0 is finally in a good enough shape that we can rely on it. Note that upstream is already on 1.29.2, but there are no fixes in those newer versions relevant to these workarounds, so we can target 1.26.0 without missing anything for now.
The possibly most impactful removed workaround is for a bug in basic arithmetic operations on zero-sized arrays, which could lead to segfaults in onnxruntime. We mitigated that using an unsqueeze-op-squeeze pattern, which significantly increased graph size and even runtime in some circumstances.
Note that this long-overdue update is not an indication of how conservative we want to be about this in the future. Ideally, we will update the minimally expected onnxruntime version within about a month after a new relevant onnxruntime version becomes available on conda-forge.