fix T2V
This commit is contained in:
@@ -891,7 +891,7 @@ class HunyuanVideoPipeline(DiffusionPipeline):
|
||||
)
|
||||
|
||||
# compute the previous noisy sample x_t -> x_t-1
|
||||
if i2v_condition_type == "token_replace":
|
||||
if image_cond_latents is not None and i2v_condition_type == "token_replace":
|
||||
latents = self.scheduler.step(
|
||||
noise_pred[:, :, 1:, :, :], t, latents[:, :, 1:, :, :], **extra_step_kwargs, return_dict=False
|
||||
)[0]
|
||||
|
||||
Reference in New Issue
Block a user