LLM/diffusers
1
0
Fork 0
mirror of https://github.com/huggingface/diffusers.git synced 2025-12-06 12:34:13 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix spatial compression ratio error for AutoEncoderKLWan doing tiled encode (#12753)

Browse Source 
fix spatial compression ratio compute error for AutoEncoderKLWan

Co-authored-by: lirui.926 <lirui.926@bytedance.com>
This commit is contained in:
Jiang
2025-12-05 02:57:13 +08:00
committed by GitHub
parent 2842c14c5f
commit cd00ba685b
1 changed files with 13 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
src/diffusers/models/autoencoders/autoencoder_kl_wan.py
View File

@@ -1259,14 +1259,20 @@ class AutoencoderKLWan(ModelMixin, AutoencoderMixin, ConfigMixin, FromOriginalMo
`torch.Tensor`:
The latent representation of the encoded videos.
"""
_, _, num_frames, height, width = x.shape
latent_height = height // self.spatial_compression_ratio
latent_width = width // self.spatial_compression_ratio
tile_latent_min_height = self.tile_sample_min_height // self.spatial_compression_ratio
tile_latent_min_width = self.tile_sample_min_width // self.spatial_compression_ratio
tile_latent_stride_height = self.tile_sample_stride_height // self.spatial_compression_ratio
tile_latent_stride_width = self.tile_sample_stride_width // self.spatial_compression_ratio
_, _, num_frames, height, width = x.shape
encode_spatial_compression_ratio = self.spatial_compression_ratio
if self.config.patch_size is not None:
assert encode_spatial_compression_ratio % self.config.patch_size == 0
encode_spatial_compression_ratio = self.spatial_compression_ratio // self.config.patch_size
latent_height = height // encode_spatial_compression_ratio
latent_width = width // encode_spatial_compression_ratio
tile_latent_min_height = self.tile_sample_min_height // encode_spatial_compression_ratio
tile_latent_min_width = self.tile_sample_min_width // encode_spatial_compression_ratio
tile_latent_stride_height = self.tile_sample_stride_height // encode_spatial_compression_ratio
tile_latent_stride_width = self.tile_sample_stride_width // encode_spatial_compression_ratio
blend_height = tile_latent_min_height - tile_latent_stride_height
blend_width = tile_latent_min_width - tile_latent_stride_width