nodes/kijai-ComfyUI-WanVideoWrapper
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Update nodes_sampler.py

Browse Source
This commit is contained in:
kijai
2025-10-30 17:55:55 +02:00
parent ba1beba982 9e0b3afe4e
commit 95391f403d
1 changed files with 15 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
nodes_sampler.py
+15 -3
View File
@@ -2256,7 +2256,11 @@ class WanVideoSampler:
timesteps = [torch.tensor([t], device=device) for t in timesteps]
timesteps = [timestep_transform(t, shift=shift, num_timesteps=1000) for t in timesteps]
else:
sample_scheduler, timesteps,_,_ = get_scheduler(scheduler, total_steps, start_step, end_step, shift, device, transformer.dim, flowedit_args, denoise_strength, sigmas=sigmas)
if isinstance(scheduler, dict):
sample_scheduler = copy.deepcopy(scheduler["sample_scheduler"])
timesteps = scheduler["timesteps"]
else:
sample_scheduler, timesteps,_,_ = get_scheduler(scheduler, total_steps, start_step, end_step, shift, device, transformer.dim, flowedit_args, denoise_strength, sigmas=sigmas)
timesteps = [torch.tensor([float(t)], device=device) for t in timesteps] + [torch.tensor([0.], device=device)]
# sample videos
@@ -2567,7 +2571,11 @@ class WanVideoSampler:
if s2v_pose is not None:
s2v_pose_slice = pose_cond_list[r].to(device)
sample_scheduler, timesteps,_,_ = get_scheduler(scheduler, total_steps, start_step, end_step, shift, device, transformer.dim, flowedit_args, denoise_strength, sigmas=sigmas)
if isinstance(scheduler, dict):
sample_scheduler = copy.deepcopy(scheduler["sample_scheduler"])
timesteps = scheduler["timesteps"]
else:
sample_scheduler, timesteps,_,_ = get_scheduler(scheduler, total_steps, start_step, end_step, shift, device, transformer.dim, flowedit_args, denoise_strength, sigmas=sigmas)
latent = noise.to(device)
for i, t in enumerate(tqdm(timesteps, desc=f"Sampling audio indices {left_idx}-{right_idx}", position=0)):
@@ -2784,7 +2792,11 @@ class WanVideoSampler:
thresholds = thresholds.reshape(-1, 1, 1, 1, 1).to(device)
masks = (1-noise_mask.repeat(len(timesteps), 1, 1, 1, 1).to(device)) > thresholds
sample_scheduler, timesteps,_,_ = get_scheduler(scheduler, total_steps, start_step, end_step, shift, device, transformer.dim, flowedit_args, denoise_strength, sigmas=sigmas)
if isinstance(scheduler, dict):
sample_scheduler = copy.deepcopy(scheduler["sample_scheduler"])
timesteps = scheduler["timesteps"]
else:
sample_scheduler, timesteps,_,_ = get_scheduler(scheduler, total_steps, start_step, end_step, shift, device, transformer.dim, flowedit_args, denoise_strength, sigmas=sigmas)
# sample videos
latent = noise