cache-dit 0.2.26__py3-none-any.whl → 0.2.28__py3-none-any.whl

cache_dit/cache_factory/cache_interface.py

@@ -1,23 +1,18 @@
 from typing import Any, Tuple, List
 from diffusers import DiffusionPipeline
-from cache_dit.cache_factory.forward_pattern import ForwardPattern
 from cache_dit.cache_factory.cache_types import CacheType
-from cache_dit.cache_factory.cache_adapters import BlockAdapter
-from cache_dit.cache_factory.cache_adapters import UnifiedCacheAdapter
+from cache_dit.cache_factory.block_adapters import BlockAdapter
+from cache_dit.cache_factory.block_adapters import BlockAdapterRegistry
+from cache_dit.cache_factory.cache_adapters import CachedAdapter
 
 from cache_dit.logger import init_logger
 
 logger = init_logger(__name__)
 
 
-def supported_pipelines() -> Tuple[int, List[str]]:
-    return UnifiedCacheAdapter.supported_pipelines()
-
-
 def enable_cache(
-    # BlockAdapter & forward pattern
+    # DiffusionPipeline or BlockAdapter
     pipe_or_adapter: DiffusionPipeline | BlockAdapter | Any,
-    forward_pattern: ForwardPattern = ForwardPattern.Pattern_0,
     # Cache context kwargs
     Fn_compute_blocks: int = 8,
     Bn_compute_blocks: int = 0,
@@ -26,7 +21,7 @@ def enable_cache(
     max_continuous_cached_steps: int = -1,
     residual_diff_threshold: float = 0.08,
     # Cache CFG or not
-    do_separate_cfg: bool = False,
+    enable_spearate_cfg: bool = False,
     cfg_compute_first: bool = False,
     cfg_diff_compute_separate: bool = True,
     # Hybird TaylorSeer
@@ -34,7 +29,7 @@ def enable_cache(
     enable_encoder_taylorseer: bool = False,
     taylorseer_cache_type: str = "residual",
     taylorseer_order: int = 2,
-    **other_cache_kwargs,
+    **other_cache_context_kwargs,
 ) -> DiffusionPipeline | Any:
     r"""
     Unified Cache API for  almost Any Diffusion Transformers (with Transformer Blocks
@@ -48,9 +43,6 @@ def enable_cache(
             The standard Diffusion Pipeline or custom BlockAdapter (from cache-dit or user-defined).
             For example: cache_dit.enable_cache(FluxPipeline(...)). Please check https://github.com/vipshop/cache-dit/blob/main/docs/BlockAdapter.md
             for the usgae of BlockAdapter.
-        forward_pattern (`ForwardPattern`, *required*, defaults to `ForwardPattern.Pattern_0`):
-            The forward pattern of Transformer block, please check https://github.com/vipshop/cache-dit/tree/main?tab=readme-ov-file#forward-pattern-matching
-            for more details.
         Fn_compute_blocks (`int`, *required*, defaults to 8):
             Specifies that `DBCache` uses the **first n** Transformer blocks to fit the information
             at time step t, enabling the calculation of a more stable L1 diff and delivering more
@@ -72,9 +64,9 @@ def enable_cache(
         residual_diff_threshold (`float`, *required*, defaults to 0.08):
             he value of residual diff threshold, a higher value leads to faster performance at the
             cost of lower precision.
-        do_separate_cfg (`bool`, *required*,  defaults to False):
+        enable_spearate_cfg (`bool`, *required*,  defaults to False):
             Whether to do separate cfg or not, such as Wan 2.1, Qwen-Image. For model that fused CFG
-            and non-CFG into single forward step, should set do_separate_cfg as False, for example:
+            and non-CFG into single forward step, should set enable_spearate_cfg as False, for example:
             CogVideoX, HunyuanVideo, Mochi, etc.
         cfg_compute_first (`bool`, *required*,  defaults to False):
             Compute cfg forward first or not, default False, namely, 0, 2, 4, ..., -> non-CFG step;
@@ -97,7 +89,7 @@ def enable_cache(
             The order of taylorseer, higher values of n_derivatives will lead to longer computation time,
             but may improve precision significantly.
         other_cache_kwargs: (`dict`, *optional*, defaults to {})
-            Other cache context kwargs, please check https://github.com/vipshop/cache-dit/blob/main/src/cache_dit/cache_factory/cache_context.py
+            Other cache context kwargs, please check https://github.com/vipshop/cache-dit/blob/main/src/cache_dit/cache_factory/cache_contexts/cache_context.py
             for more details.
 
     Examples:
@@ -111,7 +103,11 @@ def enable_cache(
     """
 
     # Collect cache context kwargs
-    cache_context_kwargs = other_cache_kwargs.copy()
+    cache_context_kwargs = other_cache_context_kwargs.copy()
+    if cache_type := cache_context_kwargs.get("cache_type", None):
+        if cache_type == CacheType.NONE:
+            return pipe_or_adapter
+
     cache_context_kwargs["cache_type"] = CacheType.DBCache
     cache_context_kwargs["Fn_compute_blocks"] = Fn_compute_blocks
     cache_context_kwargs["Bn_compute_blocks"] = Bn_compute_blocks
@@ -121,7 +117,7 @@ def enable_cache(
         max_continuous_cached_steps
     )
     cache_context_kwargs["residual_diff_threshold"] = residual_diff_threshold
-    cache_context_kwargs["do_separate_cfg"] = do_separate_cfg
+    cache_context_kwargs["enable_spearate_cfg"] = enable_spearate_cfg
     cache_context_kwargs["cfg_compute_first"] = cfg_compute_first
     cache_context_kwargs["cfg_diff_compute_separate"] = (
         cfg_diff_compute_separate
@@ -131,31 +127,35 @@ def enable_cache(
         enable_encoder_taylorseer
     )
     cache_context_kwargs["taylorseer_cache_type"] = taylorseer_cache_type
-    if "taylorseer_kwargs" in cache_context_kwargs:
-        cache_context_kwargs["taylorseer_kwargs"][
-            "n_derivatives"
-        ] = taylorseer_order
-    else:
-        cache_context_kwargs["taylorseer_kwargs"] = {
-            "n_derivatives": taylorseer_order
-        }
+    cache_context_kwargs["taylorseer_order"] = taylorseer_order
 
     if isinstance(pipe_or_adapter, BlockAdapter):
-        return UnifiedCacheAdapter.apply(
+        return CachedAdapter.apply(
             pipe=None,
             block_adapter=pipe_or_adapter,
-            forward_pattern=forward_pattern,
             **cache_context_kwargs,
         )
     elif isinstance(pipe_or_adapter, DiffusionPipeline):
-        return UnifiedCacheAdapter.apply(
+        return CachedAdapter.apply(
             pipe=pipe_or_adapter,
             block_adapter=None,
-            forward_pattern=forward_pattern,
             **cache_context_kwargs,
         )
     else:
         raise ValueError(
+            f"type: {type(pipe_or_adapter)} is not valid, "
             "Please pass DiffusionPipeline or BlockAdapter"
             "for the 1's position param: pipe_or_adapter"
         )
+
+
+def supported_pipelines(
+    **kwargs,
+) -> Tuple[int, List[str]]:
+    return BlockAdapterRegistry.supported_pipelines(**kwargs)
+
+
+def get_adapter(
+    pipe: DiffusionPipeline | str | Any,
+) -> BlockAdapter:
+    return BlockAdapterRegistry.get_adapter(pipe)

cache_dit/cache_factory/patch_functors/functor_chroma.py

@@ -30,6 +30,9 @@ class ChromaPatchFunctor(PatchFunctor):
         blocks: torch.nn.ModuleList = None,
         **kwargs,
     ) -> ChromaTransformer2DModel:
+        if getattr(transformer, "_is_patched", False):
+            return transformer
+
         if blocks is None:
             blocks = transformer.single_transformer_blocks
 

cache_dit/cache_factory/patch_functors/functor_flux.py

@@ -30,6 +30,10 @@ class FluxPatchFunctor(PatchFunctor):
         blocks: torch.nn.ModuleList = None,
         **kwargs,
     ) -> FluxTransformer2DModel:
+
+        if getattr(transformer, "_is_patched", False):
+            return transformer
+
         if blocks is None:
             blocks = transformer.single_transformer_blocks
 

cache_dit/quantize/quantize_ao.py

@@ -179,6 +179,7 @@ def quantize_ao(
     force_empty_cache()
 
     logger.info(
+        f"Quantized        Method: {quant_type:>5}\n"
         f"Quantized Linear Layers: {num_quant_linear:>5}\n"
         f"Skipped   Linear Layers: {num_skip_linear:>5}\n"
         f"Total     Linear Layers: {num_linear_layers:>5}\n"

cache_dit/utils.py

@@ -7,7 +7,6 @@ from diffusers import DiffusionPipeline
 
 from typing import Dict, Any
 from cache_dit.logger import init_logger
-from cache_dit.cache_factory import CacheType
 
 
 logger = init_logger(__name__)
@@ -30,27 +29,32 @@ class CacheStats:
 
 
 def summary(
-    pipe_or_transformer: DiffusionPipeline | torch.nn.Module | Any,
+    pipe_or_module: DiffusionPipeline | torch.nn.Module | Any,
     details: bool = False,
     logging: bool = True,
 ) -> CacheStats:
     cache_stats = CacheStats()
-    cls_name = pipe_or_transformer.__class__.__name__
-    if not isinstance(pipe_or_transformer, torch.nn.Module):
-        assert hasattr(pipe_or_transformer, "transformer")
-        transformer = pipe_or_transformer.transformer
+
+    if not isinstance(pipe_or_module, torch.nn.Module):
+        assert hasattr(pipe_or_module, "transformer")
+        module = pipe_or_module.transformer
+        cls_name = module.__class__.__name__
     else:
-        transformer = pipe_or_transformer
+        module = pipe_or_module
+
+    cls_name = module.__class__.__name__
+    if isinstance(module, torch.nn.ModuleList):
+        cls_name = module[0].__class__.__name__
 
-    if hasattr(transformer, "_cache_context_kwargs"):
-        cache_options = transformer._cache_context_kwargs
+    if hasattr(module, "_cache_context_kwargs"):
+        cache_options = module._cache_context_kwargs
         cache_stats.cache_options = cache_options
         if logging:
             print(f"\n🤗Cache Options: {cls_name}\n\n{cache_options}")
 
-    if hasattr(transformer, "_cached_steps"):
-        cached_steps: list[int] = transformer._cached_steps
-        residual_diffs: dict[str, float] = dict(transformer._residual_diffs)
+    if hasattr(module, "_cached_steps"):
+        cached_steps: list[int] = module._cached_steps
+        residual_diffs: dict[str, float] = dict(module._residual_diffs)
         cache_stats.cached_steps = cached_steps
         cache_stats.residual_diffs = residual_diffs
 
@@ -91,11 +95,9 @@ def summary(
                     compact=True,
                 )
 
-    if hasattr(transformer, "_cfg_cached_steps"):
-        cfg_cached_steps: list[int] = transformer._cfg_cached_steps
-        cfg_residual_diffs: dict[str, float] = dict(
-            transformer._cfg_residual_diffs
-        )
+    if hasattr(module, "_cfg_cached_steps"):
+        cfg_cached_steps: list[int] = module._cfg_cached_steps
+        cfg_residual_diffs: dict[str, float] = dict(module._cfg_residual_diffs)
         cache_stats.cfg_cached_steps = cfg_cached_steps
         cache_stats.cfg_residual_diffs = cfg_residual_diffs
 
@@ -153,9 +155,15 @@ def strify(
         cache_options = stats.cache_options
         cached_steps = len(stats.cached_steps)
     elif isinstance(pipe_or_stats, dict):
+        from cache_dit.cache_factory import CacheType
+
         # Assume cache_context_kwargs
         cache_options = pipe_or_stats
         cached_steps = None
+        cache_type = cache_options.get("cache_type", CacheType.NONE)
+
+        if cache_type == CacheType.NONE:
+            return "NONE"
     else:
         raise ValueError(
             "Please set pipe_or_stats param as one of: "
@@ -165,17 +173,9 @@ def strify(
     if not cache_options:
         return "NONE"
 
-    if cache_options.get("cache_type", None) != CacheType.DBCache:
-        return "NONE"
-
     def get_taylorseer_order():
         taylorseer_order = 0
-        if "taylorseer_kwargs" in cache_options:
-            if "n_derivatives" in cache_options["taylorseer_kwargs"]:
-                taylorseer_order = cache_options["taylorseer_kwargs"][
-                    "n_derivatives"
-                ]
-        elif "taylorseer_order" in cache_options:
+        if "taylorseer_order" in cache_options:
             taylorseer_order = cache_options["taylorseer_order"]
         return taylorseer_order
 

{cache_dit-0.2.26.dist-info → cache_dit-0.2.28.dist-info}/METADATA

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: cache_dit
-Version: 0.2.26
-Summary: 🤗 CacheDiT: An Unified and Training-free Cache Acceleration Toolbox for Diffusion Transformers
+Version: 0.2.28
+Summary: 🤗 A Unified and Training-free Cache Acceleration Toolbox for Diffusion Transformers
 Author: DefTruth, vipshop.com, etc.
 Maintainer: DefTruth, vipshop.com, etc
 Project-URL: Repository, https://github.com/vipshop/cache-dit.git
@@ -44,7 +44,7 @@ Dynamic: requires-python
   <img src=https://github.com/vipshop/cache-dit/raw/main/assets/cache-dit-logo.png height="120">
 
   <p align="center">
-    An <b>Unified</b> and Training-free <b>Cache Acceleration</b> Toolbox for <b>Diffusion Transformers</b> <br>
+    A <b>Unified</b> and Training-free <b>Cache Acceleration</b> Toolbox for <b>Diffusion Transformers</b> <br>
     ♥️ <b>Cache Acceleration</b> with <b>One-line</b> Code ~ ♥️
   </p>
   <div align='center'>
@@ -59,23 +59,24 @@ Dynamic: requires-python
   🔥<b><a href="#unified">Unified Cache APIs</a> | <a href="#dbcache">DBCache</a> | <a href="#taylorseer">Hybrid TaylorSeer</a> | <a href="#cfg">Hybrid Cache CFG</a></b>🔥
   </p>
   <p align="center">
-  🎉Now, <b>cache-dit</b> covers <b>All</b> mainstream <b>DiT-based</b> Diffusers' Pipelines</b>🎉<br>
+  🎉Now, <b>cache-dit</b> covers <b>100%</b> Diffusers' <b>DiT-based</b> Pipelines🎉<br>
   🔥<b><a href="#supported">Qwen-Image</a> | <a href="#supported">FLUX.1</a> | <a href="#supported">Wan 2.1/2.2</a> | <a href="#supported"> ... </a> | <a href="#supported">CogVideoX</a></b>🔥
   </p>
 </div>
 
 ## 🔥News  
 
-- [2025-08-29] 🔥</b>Covers <b>All</b> Diffusers' <b>DiT-based</b> Pipelines via **[BlockAdapter](#unified) + [Pattern Matching](#unified).**
-- [2025-08-26] 🎉[**Wan2.2**](https://github.com/Wan-Video) **1.8x⚡️** speedup with `cache-dit + compile`! Please check the [example](./examples/run_wan_2.2.py).
-- [2025-08-19] 🔥[**Qwen-Image-Edit**](https://github.com/QwenLM/Qwen-Image) **2x⚡️** speedup! Check the example at [run_qwen_image_edit.py](./examples/run_qwen_image_edit.py).
+- [2025-09-03] 🎉[**Wan2.2-MoE**](https://github.com/Wan-Video) **2.4x⚡️** speedup! Please refer to [run_wan_2.2.py](./examples/pipeline/run_wan_2.2.py) as an example.
+- [2025-08-19] 🔥[**Qwen-Image-Edit**](https://github.com/QwenLM/Qwen-Image) **2x⚡️** speedup! Check the example: [run_qwen_image_edit.py](./examples/pipeline/run_qwen_image_edit.py).
 - [2025-08-12] 🎉First caching mechanism in [QwenLM/Qwen-Image](https://github.com/QwenLM/Qwen-Image) with **[cache-dit](https://github.com/vipshop/cache-dit)**, check this [PR](https://github.com/QwenLM/Qwen-Image/pull/61). 
-- [2025-08-11] 🔥[**Qwen-Image**](https://github.com/QwenLM/Qwen-Image) **1.8x⚡️** speedup! Please refer [run_qwen_image.py](./examples/run_qwen_image.py) as an example.
+- [2025-08-11] 🔥[**Qwen-Image**](https://github.com/QwenLM/Qwen-Image) **1.8x⚡️** speedup! Please refer to [run_qwen_image.py](./examples/pipeline/run_qwen_image.py) as an example.
 
 <details>
 <summary> Previous News </summary>  
-  
-- [2025-08-10] 🔥[**FLUX.1-Kontext-dev**](https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev) is supported! Please refer [run_flux_kontext.py](./examples/run_flux_kontext.py) as an example.
+
+- [2025-09-01] 📚[**Hybird Forward Pattern**](#unified) is supported! Please check [FLUX.1-dev](./examples/run_flux_adapter.py) as an example.
+- [2025-08-29] 🔥</b>Covers <b>100%</b> Diffusers' <b>DiT-based</b> Pipelines: **[BlockAdapter](#unified) + [Pattern Matching](#unified).**  
+- [2025-08-10] 🔥[**FLUX.1-Kontext-dev**](https://huggingface.co/black-forest-labs/FLUX.1-Kontext-dev) is supported! Please refer [run_flux_kontext.py](./examples/pipeline/run_flux_kontext.py) as an example.
 - [2025-07-18] 🎉First caching mechanism in [🤗huggingface/flux-fast](https://github.com/huggingface/flux-fast) with **[cache-dit](https://github.com/vipshop/cache-dit)**, check the [PR](https://github.com/huggingface/flux-fast/pull/13). 
 - [2025-07-13] **[🤗flux-faster](https://github.com/xlite-dev/flux-faster)** is released! **3.3x** speedup for FLUX.1 on NVIDIA L20 with **[cache-dit](https://github.com/vipshop/cache-dit)**.
 
@@ -88,6 +89,11 @@ Dynamic: requires-python
 - [⚙️Installation](#️installation)
 - [🔥Supported Models](#supported)
 - [🎉Unified Cache APIs](#unified)
+  - [📚Forward Pattern Matching](#unified)
+  - [🎉Cache with One-line Code](#unified)
+  - [🔥Automatic Block Adapter](#unified)
+  - [📚Hybird Forward Pattern](#unified)
+  - [🤖Cache Acceleration Stats](#unified)
 - [⚡️Dual Block Cache](#dbcache)
 - [🔥Hybrid TaylorSeer](#taylorseer)
 - [⚡️Hybrid Cache CFG](#cfg)
@@ -176,11 +182,11 @@ Currently, **cache-dit** library supports almost **Any** Diffusion Transformers
 
 Currently, for any **Diffusion** models with **Transformer Blocks** that match the specific **Input/Output patterns**, we can use the **Unified Cache APIs** from **cache-dit**, namely, the `cache_dit.enable_cache(...)` API. The **Unified Cache APIs** are currently in the experimental phase; please stay tuned for updates. The supported patterns are listed as follows:
 
-![](https://github.com/vipshop/cache-dit/raw/main/assets/patterns.png)
+![](https://github.com/vipshop/cache-dit/raw/main/assets/patterns-v1.png)
 
 ### ♥️Cache Acceleration with One-line Code
 
-In most cases, you only need to call **one-line** of code, that is `cache_dit.enable_cache(...)`. After this API is called, you just need to call the pipe as normal. The `pipe` param can be **any** Diffusion Pipeline. Please refer to [Qwen-Image](./examples/run_qwen_image.py) as an example. 
+In most cases, you only need to call **one-line** of code, that is `cache_dit.enable_cache(...)`. After this API is called, you just need to call the pipe as normal. The `pipe` param can be **any** Diffusion Pipeline. Please refer to [Qwen-Image](./examples/pipeline/run_qwen_image.py) as an example. 
 
 ```python
 import cache_dit
@@ -198,17 +204,20 @@ output = pipe(...)
 
 ### 🔥Automatic Block Adapter
 
-But in some cases, you may have a **modified** Diffusion Pipeline or Transformer that is not located in the diffusers library or not officially supported by **cache-dit** at this time. The **BlockAdapter** can help you solve this problems. Please refer to [Qwen-Image w/ BlockAdapter](./examples/run_qwen_image_adapter.py) as an example.
+But in some cases, you may have a **modified** Diffusion Pipeline or Transformer that is not located in the diffusers library or not officially supported by **cache-dit** at this time. The **BlockAdapter** can help you solve this problems. Please refer to [🔥Qwen-Image w/ BlockAdapter](./examples/adapter/run_qwen_image_adapter.py) as an example.
 
 ```python
 from cache_dit import ForwardPattern, BlockAdapter
 
-# Use BlockAdapter with `auto` mode.
+# Use 🔥BlockAdapter with `auto` mode.
 cache_dit.enable_cache(
-    BlockAdapter(pipe=pipe, auto=True), # Qwen-Image, etc.  
-    # Check `📚Forward Pattern Matching` documentation and hack the code of
-    # of Qwen-Image, you will find that it has satisfied `FORWARD_PATTERN_1`.
-    forward_pattern=ForwardPattern.Pattern_1,  
+    BlockAdapter(
+        # Any DiffusionPipeline, Qwen-Image, etc.  
+        pipe=pipe, auto=True,
+        # Check `📚Forward Pattern Matching` documentation and hack the code of
+        # of Qwen-Image, you will find that it has satisfied `FORWARD_PATTERN_1`.
+        forward_pattern=ForwardPattern.Pattern_1,
+    ),   
 )
 
 # Or, manually setup transformer configurations.
@@ -218,12 +227,39 @@ cache_dit.enable_cache(
         transformer=pipe.transformer,
         blocks=pipe.transformer.transformer_blocks,
         blocks_name="transformer_blocks",
+        forward_pattern=ForwardPattern.Pattern_1,
     ), 
-    forward_pattern=ForwardPattern.Pattern_1,
 )
 ```
 For such situations, **BlockAdapter** can help you quickly apply various cache acceleration features to your own Diffusion Pipelines and Transformers. Please check the [📚BlockAdapter.md](./docs/BlockAdapter.md) for more details.
 
+### 📚Hybird Forward Pattern
+
+Sometimes, a Transformer class will contain more than one transformer `blocks`. For example, **FLUX.1** (HiDream, Chroma, etc) contains transformer_blocks and single_transformer_blocks (with different forward patterns). The **BlockAdapter** can also help you solve this problem. Please refer to [📚FLUX.1](./examples/adapter/run_flux_adapter.py) as an example.
+
+```python
+# For diffusers <= 0.34.0, FLUX.1 transformer_blocks and 
+# single_transformer_blocks have different forward patterns.
+cache_dit.enable_cache(
+    BlockAdapter(
+        pipe=pipe, # FLUX.1, etc.
+        transformer=pipe.transformer,
+        blocks=[
+            pipe.transformer.transformer_blocks,
+            pipe.transformer.single_transformer_blocks,
+        ],
+        blocks_name=[
+            "transformer_blocks",
+            "single_transformer_blocks",
+        ],
+        forward_pattern=[
+            ForwardPattern.Pattern_1,
+            ForwardPattern.Pattern_3,
+        ],
+    ),
+)
+```
+
 ### 🤖Cache Acceleration Stats Summary
 
 After finishing each inference of `pipe(...)`, you can call the `cache_dit.summary()` API on pipe to get the details of the **Cache Acceleration Stats** for the current inference. 
@@ -347,7 +383,7 @@ cache_dit.enable_cache(
 
 <div id="cfg"></div>
 
-cache-dit supports caching for **CFG (classifier-free guidance)**. For models that fuse CFG and non-CFG into a single forward step, or models that do not include CFG (classifier-free guidance) in the forward step, please set `do_separate_cfg` param to **False (default)**. Otherwise, set it to True. For examples:
+cache-dit supports caching for **CFG (classifier-free guidance)**. For models that fuse CFG and non-CFG into a single forward step, or models that do not include CFG (classifier-free guidance) in the forward step, please set `enable_spearate_cfg` param to **False (default)**. Otherwise, set it to True. For examples:
 
 ```python
 cache_dit.enable_cache(
@@ -355,10 +391,10 @@ cache_dit.enable_cache(
     ...,
     # CFG: classifier free guidance or not
     # For model that fused CFG and non-CFG into single forward step,
-    # should set do_separate_cfg as False. For example, set it as True 
+    # should set enable_spearate_cfg as False. For example, set it as True 
     # for Wan 2.1/Qwen-Image and set it as False for FLUX.1, HunyuanVideo, 
     # CogVideoX, Mochi, LTXVideo, Allegro, CogView3Plus, EasyAnimate, SD3, etc.
-    do_separate_cfg=True, # Wan 2.1, Qwen-Image, CogView4, Cosmos, SkyReelsV2, etc.
+    enable_spearate_cfg=True, # Wan 2.1, Qwen-Image, CogView4, Cosmos, SkyReelsV2, etc.
     # Compute cfg forward first or not, default False, namely, 
     # 0, 2, 4, ..., -> non-CFG step; 1, 3, 5, ... -> CFG step.
     cfg_compute_first=False,
@@ -433,7 +469,7 @@ The **cache-dit** codebase is adapted from FBCache. Special thanks to their exce
 
 ```BibTeX
 @misc{cache-dit@2025,
-  title={cache-dit: An Unified and Training-free Cache Acceleration Toolbox for Diffusion Transformers},
+  title={cache-dit: A Unified and Training-free Cache Acceleration Toolbox for Diffusion Transformers},
   url={https://github.com/vipshop/cache-dit.git},
   note={Open-source software available at https://github.com/vipshop/cache-dit.git},
   author={vipshop.com},

cache_dit-0.2.28.dist-info/RECORD

@@ -0,0 +1,47 @@
+cache_dit/__init__.py,sha256=V4jCkTic4XvWojCUqYcjlvxiNM2DjGQbOLk6R-tAx2A,1191
+cache_dit/_version.py,sha256=hCl1MKhh249NDbigjeJY-mrKYvjPFbJ7oklAepBQrto,706
+cache_dit/logger.py,sha256=0zsu42hN-3-rgGC_C29ms1IvVpV4_b4_SwJCKSenxBE,4304
+cache_dit/utils.py,sha256=pb5298XKmaZDoHwyteYRhixAG_0DGWrvvaObeShIhOM,7146
+cache_dit/cache_factory/.gitignore,sha256=5Cb-qT9wsTUoMJ7vACDF7ZcLpAXhi5v-xdcWSRit988,23
+cache_dit/cache_factory/__init__.py,sha256=M8q9furJOq2AZcLHRuCXZCjR9fNSELoEYdsCofIjMAo,1037
+cache_dit/cache_factory/cache_adapters.py,sha256=q7MxY44qw90h449Gr8W5iJjSwXPJR-YIyRmu_KJnQo0,13284
+cache_dit/cache_factory/cache_interface.py,sha256=2jcuTZ4D_P0M5pSw0z3BMPalobYen3YO1yKvRjaQjdQ,8332
+cache_dit/cache_factory/cache_types.py,sha256=FIFa6ZBfvvSMMHyBBhvarvgg2Y2wbRgITcG_uGylGe0,991
+cache_dit/cache_factory/forward_pattern.py,sha256=FumlCuZ-TSmSYH0hGBHctSJ-oGLCftdZjLygqhsmdR4,2258
+cache_dit/cache_factory/utils.py,sha256=XkVM9AXcB9zYq8-S8QKAsGz80r3tA6U3lBNGDGeHOe4,1871
+cache_dit/cache_factory/block_adapters/__init__.py,sha256=UFuHxNR7Y0RZoCl97wn0u2d_2rj8PzNsWfzgda5AoKM,17395
+cache_dit/cache_factory/block_adapters/block_adapters.py,sha256=WsqGUDSDU_5-pIXwDqAK_k4a-4jgbFGoLCoF6kAjLt8,19198
+cache_dit/cache_factory/block_adapters/block_registers.py,sha256=99ouWioxldlZJYQWhcUkOu94f8vO9V9QGzVNhKWtyO4,2005
+cache_dit/cache_factory/cache_blocks/__init__.py,sha256=OWjnpJxA8EJVoRzuyb5miuiRphUFj831-bbtWsTDjnM,2750
+cache_dit/cache_factory/cache_blocks/pattern_0_1_2.py,sha256=ElMps6_7uI74tSF9GDR_dEI0bZEhdzcepM29xFWnYo8,428
+cache_dit/cache_factory/cache_blocks/pattern_3_4_5.py,sha256=CtBr6nvtAW8SAeEwPwiwWtPgrmwyb5ukb-j3IwFULJU,9953
+cache_dit/cache_factory/cache_blocks/pattern_base.py,sha256=XSDy3hsaKbAZPGZY92YgGA0qLgjQyIX8irQkb2R5T2c,20331
+cache_dit/cache_factory/cache_blocks/utils.py,sha256=wfreGvtK22hDnXuw0z0hUw-9ywu91FnExfPkP8ZzlkA,891
+cache_dit/cache_factory/cache_contexts/__init__.py,sha256=rqnJ5__zqnpVHK5A1OqWILpNh5Ss-0ZDTGgtxZMKGGo,250
+cache_dit/cache_factory/cache_contexts/cache_context.py,sha256=N88WLdd4KE9DuMWmpX8URcF55E2zWNwcKMxgVYkxMJY,13691
+cache_dit/cache_factory/cache_contexts/cache_manager.py,sha256=wSghuX93KmCxE4pFEVKuyrO0Jt5STu_x4CxypS2EdxI,34276
+cache_dit/cache_factory/cache_contexts/taylorseer.py,sha256=etSUIZzDvqW3ScKCbccTPcFaSmxV1T-xAXdk-p3e3wk,3802
+cache_dit/cache_factory/patch_functors/__init__.py,sha256=yK05iONMGILsTZ83ynrUUJtiJKJ_FDjxmVIzRLy416s,252
+cache_dit/cache_factory/patch_functors/functor_base.py,sha256=Ahk0fTfrHgNdEl-9JSkACvfyyv9G-Ei5OSz7XBIlX5o,357
+cache_dit/cache_factory/patch_functors/functor_chroma.py,sha256=IFCuFU8HCnohM9Qpij7oU_UG1T8Ok8uTI3B9Nw7BHfY,10029
+cache_dit/cache_factory/patch_functors/functor_flux.py,sha256=3Q8x-PEeDiUtuv-FnQ2qEzo4qbpsYOPB9uf7YridE88,9538
+cache_dit/compile/__init__.py,sha256=FcTVzCeyypl-mxlc59_ehHL3lBNiDAFsXuRoJ-5Cfi0,56
+cache_dit/compile/utils.py,sha256=nN2OIrSdwRR5zGxJinKDqb07pXpvTNTF3g_OgLkeeBU,3858
+cache_dit/custom_ops/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+cache_dit/custom_ops/triton_taylorseer.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+cache_dit/metrics/__init__.py,sha256=RaUhl5dieF40RqnizGzR30qoJJ9dyMUEADwgwMaMQrE,575
+cache_dit/metrics/config.py,sha256=ieOgD9ayz722RjVzk24bSIqS2D6o7TZjGk8KeXV-OLQ,551
+cache_dit/metrics/fid.py,sha256=9Ivtazl6mW0Bon2VXa-Ia5Xj2ewxRD3V1Qkd69zYM3Y,17066
+cache_dit/metrics/inception.py,sha256=pBVe2X6ylLPIXTG4-GWDM9DWnCviMJbJ45R3ulhktR0,12759
+cache_dit/metrics/lpips.py,sha256=I2qCNi6qJh5TRsaIsdxO0WoRX1DN7U_H3zS0oCSahYM,1032
+cache_dit/metrics/metrics.py,sha256=8jvM1sF-nDxUuwCRy44QEoo4dYVLCQVh1QyAMs4eaQY,27840
+cache_dit/quantize/__init__.py,sha256=kWYoMAyZgBXu9BJlZjTQ0dRffW9GqeeY9_iTkXrb70A,59
+cache_dit/quantize/quantize_ao.py,sha256=mGspqYgQtenl3QnKPtsSYsSD7LbVX93f1M940bhXKLU,6066
+cache_dit/quantize/quantize_interface.py,sha256=2s_R7xPSKuJeFpEGeLwRxnq_CqJcBG3a3lzyW5wh-UM,1241
+cache_dit/quantize/quantize_svdq.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+cache_dit-0.2.28.dist-info/licenses/LICENSE,sha256=Dqb07Ik2dV41s9nIdMUbiRWEfDqo7-dQeRiY7kPO8PE,3769
+cache_dit-0.2.28.dist-info/METADATA,sha256=03FPh4nIDfjSFMfkDz-sWr2g3l30UsQek8VjQ6TPn8g,23204
+cache_dit-0.2.28.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+cache_dit-0.2.28.dist-info/entry_points.txt,sha256=FX2gysXaZx6NeK1iCLMcIdP8Q4_qikkIHtEmi3oWn8o,65
+cache_dit-0.2.28.dist-info/top_level.txt,sha256=ZJDydonLEhujzz0FOkVbO-BqfzO9d_VqRHmZU-3MOZo,10
+cache_dit-0.2.28.dist-info/RECORD,,