Optimizing Qwen Image for edge devices
BF16 and image generation models
Metal FlashAttention 2.0: pushing forward on-device inference & training on Apple silicon