Repository navigation
Add mlx-smolvla (MLX-native SmolVLA runtime for Apple Silicon) - #8
Merged
Merged
Conversation
Owner
|
Thx, daniiarabdiev, |
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Adds mlx-smolvla, my publicly available MLX-native SmolVLA runtime, to Toolkits using the project-link format. Its scope is Apple Silicon checkpoint inference, LeRobot-protocol serving, and preview LoRA training with PyTorch parity checks; training remains a research preview and task-level hardware success is not claimed.
https://github.com/daniiarabdiev/mlx-smolvla
Prior art: tokimoa/smolvla-mlx (Hugging Face, 2026-07-29) is an earlier independent inference-only SmolVLA-to-MLX port; tokimoa also published pi0-mlx and groot-n1.7-mlx. The SmolVLM implementation in mlx-vlm was an architecture reference, with no runtime dependency. mlx-smolvla is not a fork of either.