fix T2V
This commit is contained in:
@@ -713,8 +713,8 @@ class CogVideoSampler:
|
||||
freenoise=context_options["freenoise"] if context_options is not None else None,
|
||||
controlnet=controlnet,
|
||||
tora=tora_trajectory if tora_trajectory is not None else None,
|
||||
image_cond_start_percent=image_cond_start_percent,
|
||||
image_cond_end_percent=image_cond_end_percent
|
||||
image_cond_start_percent=image_cond_start_percent if image_cond_latents is not None else 0.0,
|
||||
image_cond_end_percent=image_cond_end_percent if image_cond_latents is not None else 1.0,
|
||||
)
|
||||
if not model["cpu_offloading"] and model["manual_offloading"]:
|
||||
pipe.transformer.to(offload_device)
|
||||
|
||||
Reference in New Issue
Block a user