AI/diffusers
1
0
Fork 0
mirror of https://github.com/huggingface/diffusers.git synced 2026-01-27 17:22:53 +03:00
Code Activity

fix copies and add pack latents to i2v

Browse Source
This commit is contained in:
Pham Hong Vinh
2026-01-12 11:30:31 +07:00
parent 837fd85c76
commit 96fbcd8301
1 changed files with 8 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
src/diffusers/pipelines/ltx2/pipeline_ltx2_image2video.py
View File

@@ -689,6 +689,11 @@ class LTX2ImageToVideoPipeline(DiffusionPipeline, FromSingleFileMixin, LTXVideoL
conditioning_mask = self._pack_latents(
conditioning_mask, self.transformer_spatial_patch_size, self.transformer_temporal_patch_size
).squeeze(-1)
if latents.ndim == 5:
# latents are of shape [B, C, F, H, W], need to be packed
latents = self._pack_latents(
latents, self.transformer_spatial_patch_size, self.transformer_temporal_patch_size
)
if latents.ndim != 3 or latents.shape[:2] != conditioning_mask.shape:
raise ValueError(
f"Provided `latents` tensor has shape {latents.shape}, but the expected shape is {conditioning_mask.shape + (num_channels_latents,)}."
@@ -754,6 +759,9 @@ class LTX2ImageToVideoPipeline(DiffusionPipeline, FromSingleFileMixin, LTXVideoL
latent_length = round(duration_s * latents_per_second)
if latents is not None:
if latents.ndim == 4:
# latents are of shape [B, C, L, M], need to be packed
latents = self._pack_audio_latents(latents)
return latents.to(device=device, dtype=dtype), latent_length
# TODO: confirm whether this logic is correct