fix wan i2v num_videos batching, vace latent trim, config scale factors, processor config, modular dtype

src/diffusers/modular_pipelines/wan/before_denoise.py

@@ -252,7 +252,7 @@ def __call__(self, components: WanModularPipeline, state: PipelineState) -> Pipe
         self.check_inputs(components, block_state)
 
         block_state.batch_size = block_state.prompt_embeds.shape[0]
-        block_state.dtype = block_state.prompt_embeds.dtype
+        block_state.dtype = components.transformer.dtype
 
         _, seq_len, _ = block_state.prompt_embeds.shape
         block_state.prompt_embeds = block_state.prompt_embeds.repeat(1, block_state.num_videos_per_prompt, 1)

src/diffusers/modular_pipelines/wan/modular_pipeline.py

@@ -70,14 +70,14 @@ def patch_size_spatial(self):
     def vae_scale_factor_spatial(self):
         vae_scale_factor = 8
         if hasattr(self, "vae") and self.vae is not None:
-            vae_scale_factor = 2 ** len(self.vae.temperal_downsample)
+            vae_scale_factor = self.vae.config.scale_factor_spatial
         return vae_scale_factor
 
     @property
     def vae_scale_factor_temporal(self):
         vae_scale_factor = 4
         if hasattr(self, "vae") and self.vae is not None:
-            vae_scale_factor = 2 ** sum(self.vae.temperal_downsample)
+            vae_scale_factor = self.vae.config.scale_factor_temporal
         return vae_scale_factor
 
     @property

src/diffusers/pipelines/wan/image_processor.py

@@ -68,7 +68,17 @@ def __init__(
         do_convert_grayscale: bool = False,
         fill_color: str | float | tuple[float, ...] | None = 0,
     ):
-        super().__init__()
+        super().__init__(
+            do_resize=do_resize,
+            vae_scale_factor=vae_scale_factor,
+            vae_latent_channels=vae_latent_channels,
+            resample=resample,
+            reducing_gap=reducing_gap,
+            do_normalize=do_normalize,
+            do_binarize=do_binarize,
+            do_convert_rgb=do_convert_rgb,
+            do_convert_grayscale=do_convert_grayscale,
+        )
         if do_convert_rgb and do_convert_grayscale:
             raise ValueError(
                 "`do_convert_rgb` and `do_convert_grayscale` can not both be set to `True`,"

src/diffusers/pipelines/wan/pipeline_wan_i2v.py

@@ -351,8 +351,10 @@ def check_inputs(
             raise ValueError(
                 "Provide either `image` or `prompt_embeds`. Cannot leave both `image` and `image_embeds` undefined."
             )
-        if image is not None and not isinstance(image, torch.Tensor) and not isinstance(image, PIL.Image.Image):
-            raise ValueError(f"`image` has to be of type `torch.Tensor` or `PIL.Image.Image` but is {type(image)}")
+        if image is not None and not isinstance(image, (torch.Tensor, PIL.Image.Image, list, tuple)):
+            raise ValueError(
+                f"`image` has to be of type `torch.Tensor`, `PIL.Image.Image`, or a list of them but is {type(image)}"
+            )
         if height % 16 != 0 or width % 16 != 0:
             raise ValueError(f"`height` and `width` have to be divisible by 16 but are {height} and {width}.")
 
@@ -453,7 +455,7 @@ def prepare_latents(
             latent_condition = torch.cat(latent_condition)
         else:
             latent_condition = retrieve_latents(self.vae.encode(video_condition), sample_mode="argmax")
-            latent_condition = latent_condition.repeat(batch_size, 1, 1, 1, 1)
+            latent_condition = latent_condition.repeat_interleave(batch_size // latent_condition.shape[0], dim=0)
 
         latent_condition = latent_condition.to(dtype)
         latent_condition = (latent_condition - latents_mean) * latents_std
@@ -698,7 +700,12 @@ def __call__(
                     image_embeds = self.encode_image(image, device)
                 else:
                     image_embeds = self.encode_image([image, last_image], device)
-            image_embeds = image_embeds.repeat(batch_size, 1, 1)
+            if last_image is None:
+                image_embeds = image_embeds.repeat_interleave(
+                    batch_size * num_videos_per_prompt // image_embeds.shape[0], dim=0
+                )
+            else:
+                image_embeds = image_embeds.repeat(batch_size, 1, 1)
             image_embeds = image_embeds.to(transformer_dtype)
 
         # 4. Prepare timesteps

src/diffusers/pipelines/wan/pipeline_wan_vace.py

@@ -196,8 +196,8 @@ def __init__(
             scheduler=scheduler,
         )
         self.register_to_config(boundary_ratio=boundary_ratio)
-        self.vae_scale_factor_temporal = 2 ** sum(self.vae.temperal_downsample) if getattr(self, "vae", None) else 4
-        self.vae_scale_factor_spatial = 2 ** len(self.vae.temperal_downsample) if getattr(self, "vae", None) else 8
+        self.vae_scale_factor_temporal = self.vae.config.scale_factor_temporal if getattr(self, "vae", None) else 4
+        self.vae_scale_factor_spatial = self.vae.config.scale_factor_spatial if getattr(self, "vae", None) else 8
         self.video_processor = VideoProcessor(vae_scale_factor=self.vae_scale_factor_spatial)
 
     # Copied from diffusers.pipelines.wan.pipeline_wan.WanPipeline._get_t5_prompt_embeds
@@ -1020,8 +1020,8 @@ def __call__(
 
         self._current_timestep = None
 
+        latents = latents[:, :, num_reference_images:]
         if not output_type == "latent":
-            latents = latents[:, :, num_reference_images:]
             latents = latents.to(vae_dtype)
             latents_mean = (
                 torch.tensor(self.vae.config.latents_mean)

src/diffusers/pipelines/wan/pipeline_wan_video2video.py

@@ -214,8 +214,8 @@ def __init__(
             scheduler=scheduler,
         )
 
-        self.vae_scale_factor_temporal = 2 ** sum(self.vae.temperal_downsample) if getattr(self, "vae", None) else 4
-        self.vae_scale_factor_spatial = 2 ** len(self.vae.temperal_downsample) if getattr(self, "vae", None) else 8
+        self.vae_scale_factor_temporal = self.vae.config.scale_factor_temporal if getattr(self, "vae", None) else 4
+        self.vae_scale_factor_spatial = self.vae.config.scale_factor_spatial if getattr(self, "vae", None) else 8
         self.video_processor = VideoProcessor(vae_scale_factor=self.vae_scale_factor_spatial)
 
     # Copied from diffusers.pipelines.wan.pipeline_wan.WanPipeline._get_t5_prompt_embeds