fix: extend f32 matmul precision to ROCm for Qwen-Image, Krea2 and Boogu#1772
Merged
Conversation
Same failure mechanism as the Z-Image fix: DiT activations exceed the f16 range, and on HIP for RDNA3.x ggml routes large-batch quantized matmuls to the hipBLAS path, which converts activations to f16 and saturates them to inf, ending in NaN latents and blank images. Widen the existing Vulkan-only force_prec_f32 gates on the attention out and FFN out projections to also cover ROCm. The shared FeedForward change also applies to the other DiTs that use it. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Same failure mechanism as the Z-Image fix, so changing just in case (haven't verified that it breaks, but there's a gate for Vulkan there, so probably safe to do it for ROCm as well)