AI/diffusers
1
0
Fork 0
mirror of https://github.com/huggingface/diffusers.git synced 2026-01-29 07:22:12 +03:00
Code Activity

support cogvideox

Browse Source
This commit is contained in:
Aryan
2025-05-15 21:41:45 +02:00
parent 82fa9df1c7
commit 367fdef96d
4 changed files with 37 additions and 33 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
src/diffusers/pipelines/cogvideo/pipeline_cogvideox.py
View File

@@ -718,14 +718,15 @@ class CogVideoXPipeline(DiffusionPipeline, CogVideoXLoraLoaderMixin):
timestep = t.expand(latent_model_input.shape[0])
# predict noise model_output
noise_pred = self.transformer(
hidden_states=latent_model_input,
encoder_hidden_states=prompt_embeds,
timestep=timestep,
image_rotary_emb=image_rotary_emb,
attention_kwargs=attention_kwargs,
return_dict=False,
)[0]
with self.transformer.cache_context("cond_uncond"):
noise_pred = self.transformer(
hidden_states=latent_model_input,
encoder_hidden_states=prompt_embeds,
timestep=timestep,
image_rotary_emb=image_rotary_emb,
attention_kwargs=attention_kwargs,
return_dict=False,
)[0]
noise_pred = noise_pred.float()
# perform guidance

17
src/diffusers/pipelines/cogvideo/pipeline_cogvideox_fun_control.py
View File

@@ -784,14 +784,15 @@ class CogVideoXFunControlPipeline(DiffusionPipeline, CogVideoXLoraLoaderMixin):
timestep = t.expand(latent_model_input.shape[0])
# predict noise model_output
noise_pred = self.transformer(
hidden_states=latent_model_input,
encoder_hidden_states=prompt_embeds,
timestep=timestep,
image_rotary_emb=image_rotary_emb,
attention_kwargs=attention_kwargs,
return_dict=False,
)[0]
with self.transformer.cache_context("cond_uncond"):
noise_pred = self.transformer(
hidden_states=latent_model_input,
encoder_hidden_states=prompt_embeds,
timestep=timestep,
image_rotary_emb=image_rotary_emb,
attention_kwargs=attention_kwargs,
return_dict=False,
)[0]
noise_pred = noise_pred.float()
# perform guidance

19
src/diffusers/pipelines/cogvideo/pipeline_cogvideox_image2video.py
View File

@@ -831,15 +831,16 @@ class CogVideoXImageToVideoPipeline(DiffusionPipeline, CogVideoXLoraLoaderMixin)
timestep = t.expand(latent_model_input.shape[0])
# predict noise model_output
noise_pred = self.transformer(
hidden_states=latent_model_input,
encoder_hidden_states=prompt_embeds,
timestep=timestep,
ofs=ofs_emb,
image_rotary_emb=image_rotary_emb,
attention_kwargs=attention_kwargs,
return_dict=False,
)[0]
with self.transformer.cache_context("cond_uncond"):
noise_pred = self.transformer(
hidden_states=latent_model_input,
encoder_hidden_states=prompt_embeds,
timestep=timestep,
ofs=ofs_emb,
image_rotary_emb=image_rotary_emb,
attention_kwargs=attention_kwargs,
return_dict=False,
)[0]
noise_pred = noise_pred.float()
# perform guidance

17
src/diffusers/pipelines/cogvideo/pipeline_cogvideox_video2video.py
View File

@@ -799,14 +799,15 @@ class CogVideoXVideoToVideoPipeline(DiffusionPipeline, CogVideoXLoraLoaderMixin)
timestep = t.expand(latent_model_input.shape[0])
# predict noise model_output
noise_pred = self.transformer(
hidden_states=latent_model_input,
encoder_hidden_states=prompt_embeds,
timestep=timestep,
image_rotary_emb=image_rotary_emb,
attention_kwargs=attention_kwargs,
return_dict=False,
)[0]
with self.transformer.cache_context("cond_uncond"):
noise_pred = self.transformer(
hidden_states=latent_model_input,
encoder_hidden_states=prompt_embeds,
timestep=timestep,
image_rotary_emb=image_rotary_emb,
attention_kwargs=attention_kwargs,
return_dict=False,
)[0]
noise_pred = noise_pred.float()
# perform guidance