Fixes mismatched number of output arguments (#315)

martinarroyo · web-flow · commit 5a05e75d833f · 2026-01-21T10:48:25.000-08:00
In #306 the return of WanTimeTextImageEmbedding was modified to include a 5th element. This refactors the call in WanVACEModel to comply with the new interface
diff --git a/src/maxdiffusion/models/wan/transformers/transformer_wan_vace.py b/src/maxdiffusion/models/wan/transformers/transformer_wan_vace.py
@@ -474,6 +474,7 @@ def __call__(
         timestep_proj,
         encoder_hidden_states,
         encoder_hidden_states_image,
+        _,
     ) = self.condition_embedder(  # We will need to mask out the text embedding.
         timestep, encoder_hidden_states, encoder_hidden_states_image
     )

Original file line number	Diff line number	Diff line change
`@@ -474,6 +474,7 @@ def __call__(`
`474`	`474`	`timestep_proj,`
`475`	`475`	`encoder_hidden_states,`
`476`	`476`	`encoder_hidden_states_image,`
	`477`	`+ _,`
`477`	`478`	`) = self.condition_embedder( # We will need to mask out the text embedding.`
`478`	`479`	`timestep, encoder_hidden_states, encoder_hidden_states_image`
`479`	`480`	`)`