Upload folder using huggingface_hub

model_index.json

@@ -9,6 +9,10 @@
       "transformers",
       "Qwen3VLForConditionalGeneration"
     ],
+    "processor": [
+      "processor",
+      "Qwen3VLProcessor"
+    ],
     "transformer": [
       "modeling_nucleusmoe",
       "NucleusMoEImageTransformer2DModel"

pipeline_nucleusmoe.py

@@ -153,36 +153,17 @@ class NucleusMoEImagePipeline(DiffusionPipeline):
     """
 
     model_cpu_offload_seq = "text_encoder->transformer->vae"
-    _optional_components = ["processor"]
     _callback_tensor_inputs = ["latents", "prompt_embeds"]
 
-    @classmethod
-    def from_pretrained(cls, pretrained_model_name_or_path, **kwargs):
-        if "processor" not in kwargs:
-            kwargs["processor"] = Qwen3VLProcessor.from_pretrained(
-                pretrained_model_name_or_path, subfolder="text_encoder"
-            )
-        return super().from_pretrained(pretrained_model_name_or_path, **kwargs)
-
     def __init__(
         self,
         transformer: NucleusMoEImageTransformer2DModel,
         scheduler: FlowMatchEulerDiscreteScheduler,
         vae: AutoencoderKLQwenImage,
         text_encoder: Qwen3VLForConditionalGeneration,
-        processor: Qwen3VLProcessor | None = None,
+        processor: Qwen3VLProcessor,
     ):
         super().__init__()
-        if processor is None:
-            processor_path = (
-                getattr(text_encoder, "name_or_path", None)
-                or getattr(getattr(text_encoder, "config", None), "_name_or_path", None)
-            )
-            if processor_path is None:
-                raise ValueError(
-                    "Could not infer a processor path from `text_encoder`; pass `processor=` explicitly."
-                )
-            processor = Qwen3VLProcessor.from_pretrained(processor_path)
         self.register_modules(
             transformer=transformer,
             scheduler=scheduler,

processor/preprocessor_config.json

@@ -0,0 +1,21 @@
+{
+    "size": {
+        "longest_edge": 16777216,
+        "shortest_edge": 65536
+    },
+    "patch_size": 16,
+    "temporal_patch_size": 2,
+    "merge_size": 2,
+    "image_mean": [
+        0.5,
+        0.5,
+        0.5
+    ],
+    "image_std": [
+        0.5,
+        0.5,
+        0.5
+    ],
+    "processor_class": "Qwen3VLProcessor",
+    "image_processor_type": "Qwen2VLImageProcessorFast"
+}