kanshan/diffusers
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge d76b744ac3 into 6bf668c4d2

Browse Source
This commit is contained in:
Sayak Paul
2025-11-27 12:24:02 +03:00
committed by GitHub
parent 6bf668c4d2 d76b744ac3
commit e44f8b6c04
3 changed files with 9 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/source/en/api/cache.md
+1 -1
View File
@@ -29,7 +29,7 @@ Cache methods speedup diffusion transformers by storing and reusing intermediate
[[autodoc]] apply_faster_cache [[autodoc]] apply_faster_cache
### FirstBlockCacheConfig ## FirstBlockCacheConfig
[[autodoc]] FirstBlockCacheConfig [[autodoc]] FirstBlockCacheConfig
docs/source/en/optimization/cache.md
+5 -1
View File
@@ -66,4 +66,8 @@ config = FasterCacheConfig(
tensor_format="BFCHW", tensor_format="BFCHW",
) )
pipeline.transformer.enable_cache(config) pipeline.transformer.enable_cache(config)
``` ```
## FirstBlockCache
[FirstBlock Cache](https://huggingface.co/docs/diffusers/main/en/api/cache#diffusers.FirstBlockCacheConfig) builds on the ideas of [TeaCache](https://huggingface.co/papers/2411.19108). It is much simpler to implement generically for a wide range of models and has been integrated first for experimental purposes.
src/diffusers/models/cache_utils.py
+3 -1
View File
@@ -41,9 +41,11 @@ class CacheMixin:
Enable caching techniques on the model. Enable caching techniques on the model.
Args: Args:
config (`Union[PyramidAttentionBroadcastConfig]`): config (`Union[PyramidAttentionBroadcastConfig, FasterCacheConfig, FirstBlockCacheConfig]`):
The configuration for applying the caching technique. Currently supported caching techniques are: The configuration for applying the caching technique. Currently supported caching techniques are:
- [`~hooks.PyramidAttentionBroadcastConfig`] - [`~hooks.PyramidAttentionBroadcastConfig`]
- [`~hooks.FasterCacheConfig`]
- [`~hooks.FirstBlockCacheConfig`]
Example: Example: