feat(wan): Add LoRA with improved quantization pipeline

mlx_video/lora/__init__.py

@@ -0,0 +1,25 @@
+"""LoRA support for mlx-video."""
+
+from mlx_video.lora.apply import (
+    LoRALinear,
+    apply_lora_to_linear,
+    apply_loras_to_model,
+    apply_loras_to_weights,
+)
+from mlx_video.lora.loader import (
+    load_lora_weights,
+    load_multiple_loras,
+)
+from mlx_video.lora.types import AppliedLoRA, LoRAConfig, LoRAWeights
+
+__all__ = [
+    "LoRAConfig",
+    "LoRAWeights",
+    "AppliedLoRA",
+    "load_lora_weights",
+    "load_multiple_loras",
+    "apply_lora_to_linear",
+    "apply_loras_to_weights",
+    "apply_loras_to_model",
+    "LoRALinear",
+]