Core ML: drop the fp32 variant, fp16 is equivalent in task terms dec040b verified msluszniak commited on 9 days ago
Remove MLX artifacts: slower than Core ML on device ea0d0d1 verified msluszniak commited on 9 days ago