ml-explore/mlx

[Feature Request] Fallback for custom kernels

已关闭

#3,240 创建于 2026年3月11日

 (4 条评论) (0 个反应) (0 位负责人)C++ (1,794 个派生)batch import
enhancementgood first issue

仓库指标

星标
 (26,216 个星标)
PR 合并指标
 (平均合并 6天 17小时) (30 天内合并 75 个 PR)

描述

We have a specific use-case whereby we need to export MLX callables from Linux. Currently, this prevents us from using any operations that require custom kernels (e.g. grid sample) because otherwise, MLX complains that no Metal GPU was found.

Ideally, we would want to keep the fast kernel so that we can export without a Metal GPU, but execute on a Metal GPU.

贡献者指南