xinyun/ComfyUI-CogVideoXWrapper
1
0
Fork 0
mirror of https://git.datalinker.icu/kijai/ComfyUI-CogVideoXWrapper.git synced 2025-12-09 04:44:22 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix FasterCache

Browse Source
This commit is contained in:
kijai 2024-11-12 07:39:00 +02:00
parent 9681c83c1d
commit 0a121dba53
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
custom_cogvideox_transformer_3d.py
View File

@ -630,11 +630,11 @@ class CogVideoXTransformer3DModel(ModelMixin, ConfigMixin, PeftAdapterMixin):
# - However, for CogVideoX-5b-I2V also takes concatenated input image latents (number of input channels is twice the output channels) # - However, for CogVideoX-5b-I2V also takes concatenated input image latents (number of input channels is twice the output channels)
if p_t is None: if p_t is None:
output = hidden_states.reshape(batch_size, num_frames, height // p, width // p, -1, p, p) output = hidden_states.reshape(1, num_frames, height // p, width // p, -1, p, p)
output = output.permute(0, 1, 4, 2, 5, 3, 6).flatten(5, 6).flatten(3, 4) output = output.permute(0, 1, 4, 2, 5, 3, 6).flatten(5, 6).flatten(3, 4)
else: else:
output = hidden_states.reshape( output = hidden_states.reshape(
batch_size, (num_frames + p_t - 1) // p_t, height // p, width // p, -1, p_t, p, p 1, (num_frames + p_t - 1) // p_t, height // p, width // p, -1, p_t, p, p
) )
output = output.permute(0, 1, 5, 4, 2, 6, 3, 7).flatten(6, 7).flatten(4, 5).flatten(1, 2) output = output.permute(0, 1, 5, 4, 2, 6, 3, 7).flatten(6, 7).flatten(4, 5).flatten(1, 2)