amdgpu: route gfx11 and newer to MIGraphX regardless of model - #68
Open
aditya-dl wants to merge 1 commit into
Open
amdgpu: route gfx11 and newer to MIGraphX regardless of model#68aditya-dl wants to merge 1 commit into
aditya-dl wants to merge 1 commit into
Conversation
The Auto heuristic split gfx110x/gfx1103 by model type: LLMs went to MIGraphX, everything else to DirectML. That regressed Navi31, where certification models had been running on MIGraphX and started landing on DirectML instead, with at least one of them failing there. Route everything from gfx11 up to MIGraphX. DirectML is now only for the three cases that actually need it: WebNN callers, Medusa (gfx117x), and anything older than gfx11. No arch distinguishes LLM from non-LLM any more, so nothing reads kLlmModelArch. Left in place, with a note, for the next arch that needs the split; kArchModelBackend still works for pinning a specific model on a specific arch. Policy unit test updated and green (47/47).
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Port of #67 to
main. Identical change; the release branch could not be merged directly because of unrelated commits in its history.Problem
The Auto heuristic split gfx110x/gfx1103 by model type: LLMs to MIGraphX, everything else to DirectML. That regressed Navi31, where certification models had been running on MIGraphX and started landing on DirectML, with at least one failing there.
Change
DirectML is now only for the three cases that need it: WebNN callers, Medusa, and anything older than gfx11. Explicit profiles and
kArchModelBackendare unchanged.No arch distinguishes LLM from non-LLM any more, so nothing reads
kLlmModelArch. Left in place with a note for the next arch that needs the split.