fix: cast bfloat16 tensors to float32 before numpy conversion in MLX pipeline#57
Closed
kswanjitsu wants to merge 1 commit into
Closed
fix: cast bfloat16 tensors to float32 before numpy conversion in MLX pipeline#57kswanjitsu wants to merge 1 commit into
kswanjitsu wants to merge 1 commit into
Conversation
…pipeline PyTorch CPU does not support `.numpy()` on bfloat16 tensors, causing a `TypeError: Got unsupported ScalarType BFloat16` when converting models like `privacy-filter-nemotron` (which uses bf16 weights) to MLX format. Cast the tensor to float32 before calling `.numpy()` when the dtype is bfloat16. The resulting MLX weights are stored in float32, which is fully supported on all Apple Silicon chips. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Owner
|
closing in the favor of this PR: #58 |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Problem
When loading models with bfloat16 weights (e.g.
OpenMed/privacy-filter-nemotron) through the MLX pipeline, conversion fails with:This happens in
_to_numpy()because PyTorch CPU does not support.numpy()on bfloat16 tensors.Fix
Cast bfloat16 tensors to float32 before calling
.numpy(). The resulting MLX weights are stored in float32, which is fully supported on all Apple Silicon chips and incurs no meaningful accuracy loss for token classification.Affected models
OpenMed/privacy-filter-nemotron(and any future bf16 models)Testing
Verified that
create_mlx_pipeline("OpenMed/privacy-filter-nemotron")completes weight conversion successfully after this fix.