n0p/mlx-video
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix timestep_conditioning logic in infer_vae_decoder_config to ensure consistent behavior based on has_timestep flag.

Browse Source
This commit is contained in:
Prince Canuma
2026-03-11 18:30:29 +01:00
parent 207c223354
commit d1fa47722b
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
mlx_video/models/ltx/convert.py
View File

@@ -494,7 +494,7 @@ def infer_vae_decoder_config(weights: Dict[str, mx.array], variant: str) -> dict
"num_res_blocks": 2, "num_res_blocks": 2,
"out_ch": 2, "out_ch": 2,
"resolution": 256, "resolution": 256,
"timestep_conditioning": has_timestep if has_timestep else (variant == "dev"), "timestep_conditioning": has_timestep,
"z_channels": 8, "z_channels": 8,
} }
return config return config