cache-dit 1.0.8__py3-none-any.whl → 1.0.10__py3-none-any.whl

cache_dit/quantize/quantize_interface.py

@@ -7,37 +7,24 @@ logger = init_logger(__name__)
 
 def quantize(
     module: torch.nn.Module,
-    quant_type: str = "fp8_w8a8_dq",
+    quant_type: str = "float8_weight_only",
     backend: str = "ao",
     exclude_layers: List[str] = [
         "embedder",
         "embed",
     ],
     filter_fn: Optional[Callable] = None,
-    # only for fp8_w8a8_dq
-    per_row: bool = True,
     **kwargs,
 ) -> torch.nn.Module:
     assert isinstance(module, torch.nn.Module)
 
     if backend.lower() in ("ao", "torchao"):
-        from cache_dit.quantize.quantize_ao import quantize_ao
-
-        quant_type = quant_type.lower()
-        assert quant_type in (
-            "fp8_w8a8_dq",
-            "fp8_w8a16_wo",
-            "int8_w8a8_dq",
-            "int8_w8a16_wo",
-            "int4_w4a8_dq",
-            "int4_w4a4_dq",
-            "int4_w4a16_wo",
-        ), f"{quant_type} is not supported for torchao backend now!"
+        from cache_dit.quantize.backends.torchao import quantize_ao
 
         return quantize_ao(
             module,
             quant_type=quant_type,
-            per_row=per_row,
+            per_row=kwargs.pop("per_row", True),
             exclude_layers=exclude_layers,
             filter_fn=filter_fn,
             **kwargs,

cache_dit/utils.py

@@ -13,6 +13,7 @@ from cache_dit.cache_factory import CacheType
 from cache_dit.cache_factory import BlockAdapter
 from cache_dit.cache_factory import BasicCacheConfig
 from cache_dit.cache_factory import CalibratorConfig
+from cache_dit.cache_factory import FakeDiffusionPipeline
 from cache_dit.parallelism import ParallelismConfig
 from cache_dit.logger import init_logger
 
@@ -64,6 +65,7 @@ def summary(
     adapter_or_others: Union[
         BlockAdapter,
         DiffusionPipeline,
+        FakeDiffusionPipeline,
         torch.nn.Module,
     ],
     details: bool = False,
@@ -73,31 +75,43 @@ def summary(
     if adapter_or_others is None:
         return [CacheStats()]
 
+    if isinstance(adapter_or_others, FakeDiffusionPipeline):
+        raise ValueError(
+            "Please pass DiffusionPipeline, BlockAdapter or transfomer, "
+            "not FakeDiffusionPipeline."
+        )
+
     if not isinstance(adapter_or_others, BlockAdapter):
         if not isinstance(adapter_or_others, DiffusionPipeline):
-            transformer = adapter_or_others
+            transformer = adapter_or_others  # transformer-only
             transformer_2 = None
         else:
             transformer = adapter_or_others.transformer
-            transformer_2 = None
+            transformer_2 = None  # Only for Wan2.2
             if hasattr(adapter_or_others, "transformer_2"):
                 transformer_2 = adapter_or_others.transformer_2
 
-        if not BlockAdapter.is_cached(transformer):
+        if all(
+            (
+                not BlockAdapter.is_cached(transformer),
+                not BlockAdapter.is_parallelized(transformer),
+            )
+        ):
             return [CacheStats()]
 
         blocks_stats: List[CacheStats] = []
-        for blocks in BlockAdapter.find_blocks(transformer):
-            blocks_stats.append(
-                _summary(
-                    blocks,
-                    details=details,
-                    logging=logging,
-                    **kwargs,
+        if BlockAdapter.is_cached(transformer):
+            for blocks in BlockAdapter.find_blocks(transformer):
+                blocks_stats.append(
+                    _summary(
+                        blocks,
+                        details=details,
+                        logging=logging,
+                        **kwargs,
+                    )
                 )
-            )
 
-        if transformer_2 is not None:
+        if transformer_2 is not None and BlockAdapter.is_cached(transformer_2):
             for blocks in BlockAdapter.find_blocks(transformer_2):
                 blocks_stats.append(
                     _summary(
@@ -126,7 +140,11 @@ def summary(
                 )
             )
 
-        blocks_stats = [stats for stats in blocks_stats if stats.cache_options]
+        blocks_stats = [
+            stats
+            for stats in blocks_stats
+            if (stats.cache_options or stats.parallelism_config)
+        ]
 
         return blocks_stats if len(blocks_stats) else [CacheStats()]
 
@@ -155,11 +173,20 @@ def strify(
     adapter_or_others: Union[
         BlockAdapter,
         DiffusionPipeline,
+        FakeDiffusionPipeline,
+        torch.nn.Module,
         CacheStats,
         List[CacheStats],
         Dict[str, Any],
     ],
 ) -> str:
+    if isinstance(adapter_or_others, FakeDiffusionPipeline):
+        raise ValueError(
+            "Please pass DiffusionPipeline, BlockAdapter or transfomer, "
+            "not FakeDiffusionPipeline."
+        )
+
+    parallelism_config: ParallelismConfig = None
     if isinstance(adapter_or_others, BlockAdapter):
         stats = summary(adapter_or_others, logging=False)[-1]
         cache_options = stats.cache_options
@@ -168,6 +195,10 @@ def strify(
         stats = summary(adapter_or_others, logging=False)[-1]
         cache_options = stats.cache_options
         cached_steps = len(stats.cached_steps)
+    elif isinstance(adapter_or_others, torch.nn.Module):
+        stats = summary(adapter_or_others, logging=False)[-1]
+        cache_options = stats.cache_options
+        cached_steps = len(stats.cached_steps)
     elif isinstance(adapter_or_others, CacheStats):
         stats = adapter_or_others
         cache_options = stats.cache_options
@@ -182,18 +213,22 @@ def strify(
         cache_options = adapter_or_others
         cached_steps = None
         cache_type = cache_options.get("cache_type", CacheType.NONE)
-
         stats = None
+        parallelism_config = cache_options.get("parallelism_config", None)
 
         if cache_type == CacheType.NONE:
             return "NONE"
     else:
         raise ValueError(
             "Please set pipe_or_stats param as one of: "
-            "DiffusionPipeline | CacheStats | Dict[str, Any]"
+            "DiffusionPipeline | CacheStats | Dict[str, Any] | List[CacheStats]"
+            " | BlockAdapter | Transformer"
         )
 
-    if not cache_options:
+    if stats is not None:
+        parallelism_config = stats.parallelism_config
+
+    if not cache_options and parallelism_config is None:
         return "NONE"
 
     def cache_str():
@@ -219,14 +254,14 @@ def strify(
         return "T0O0"
 
     def parallelism_str():
-        if stats is None:
-            return ""
-        parallelism_config: ParallelismConfig = stats.parallelism_config
         if parallelism_config is not None:
             return f"_{parallelism_config.strify()}"
         return ""
 
-    cache_type_str = f"{cache_str()}_{calibrator_str()}{parallelism_str()}"
+    cache_type_str = f"{cache_str()}"
+    if cache_type_str != "NONE":
+        cache_type_str += f"_{calibrator_str()}"
+    cache_type_str += f"{parallelism_str()}"
 
     if cached_steps:
         cache_type_str += f"_S{cached_steps}"
@@ -245,6 +280,7 @@ def _summary(
 ) -> CacheStats:
     cache_stats = CacheStats()
 
+    # Get stats from transformer
     if not isinstance(pipe_or_module, torch.nn.Module):
         assert hasattr(pipe_or_module, "transformer")
         module = pipe_or_module.transformer

{cache_dit-1.0.8.dist-info → cache_dit-1.0.10.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: cache_dit
-Version: 1.0.8
+Version: 1.0.10
 Summary: A Unified, Flexible and Training-free Cache Acceleration Framework for 🤗Diffusers.
 Author: DefTruth, vipshop.com, etc.
 Maintainer: DefTruth, vipshop.com, etc
@@ -14,6 +14,8 @@ Requires-Dist: pyyaml
 Requires-Dist: torch>=2.7.1
 Requires-Dist: diffusers>=0.35.1
 Requires-Dist: transformers>=4.55.2
+Provides-Extra: parallelism
+Requires-Dist: einops>=0.8.1; extra == "parallelism"
 Provides-Extra: quantization
 Requires-Dist: torchao>=0.12.0; extra == "quantization"
 Provides-Extra: metrics
@@ -36,6 +38,7 @@ Requires-Dist: opencv-python-headless; extra == "dev"
 Requires-Dist: ftfy; extra == "dev"
 Requires-Dist: scikit-image; extra == "dev"
 Provides-Extra: all
+Requires-Dist: cache-dit[parallelism]; extra == "all"
 Requires-Dist: cache-dit[quantization]; extra == "all"
 Requires-Dist: cache-dit[metrics]; extra == "all"
 Dynamic: license-file
@@ -48,16 +51,17 @@ Dynamic: requires-python
   <p align="center">
     A <b>Unified</b>, Flexible and Training-free <b>Cache Acceleration</b> Framework for <b>🤗Diffusers</b> <br>
     ♥️ Cache Acceleration with <b>One-line</b> Code ~ ♥️ <br>
-    🔥<b><a href="./docs/User_Guide.md">DBCache</a> | <a href="./docs/User_Guide.md">DBPrune</a> | <a href="./docs/User_Guide.md">Hybrid TaylorSeer</a> | <a href="./docs/User_Guide.md">Hybrid Cache CFG</a></b>🔥 <br> 
-    🔥<b><a href="./docs/User_Guide.md">Hybrid Context Paralleism</a> | <a href="./docs/User_Guide.md">Diffusers Native</a> | <a href="./docs/User_Guide.md">SOTA</a></b>🔥
+    🔥<a href="./docs/User_Guide.md">Forward Pattern Matching</a> | <a href="./docs/User_Guide.md">Automatic Block Adapter</a>🔥 <br>
+    🔥<a href="./docs/User_Guide.md"><b>DBCache</b></a> | <a href="./docs/User_Guide.md"><b>DBPrune</b></a> | <a href="./docs/User_Guide.md">Hybrid <b>TaylorSeer</b> Calibrator</a> | <a href="./docs/User_Guide.md">Hybrid <b>Cache CFG</b></a>🔥<br>
+    🔥<a href="./docs/User_Guide.md"><b>Context Parallelism</b></a> | <a href="./docs/User_Guide.md"><b>Tensor Parallelism</b></a> | <a href="./docs/User_Guide.md">Low-bits Quantization</a>🔥<br>
+    🔥<a href="./docs/User_Guide.md">Compile Compatible</a> | <a href="./docs/User_Guide.md"><b>🎉State-of-the-Art Performance</b></a>🎉 
   </p>
   <div align='center'>
       <img src=https://img.shields.io/badge/Language-Python-brightgreen.svg >
-      <img src=https://img.shields.io/badge/PRs-welcome-blue.svg >
       <img src=https://img.shields.io/badge/PyPI-pass-brightgreen.svg >
-      <img src=https://static.pepy.tech/badge/cache-dit >
+      <a href="https://pepy.tech/projects/cache-dit"><img src=https://static.pepy.tech/personalized-badge/cache-dit?period=total&units=INTERNATIONAL_SYSTEM&left_color=GRAY&right_color=GREEN&left_text=downloads></a>
+      <img src=https://img.shields.io/github/issues/vipshop/cache-dit.svg >
       <img src=https://img.shields.io/github/stars/vipshop/cache-dit.svg?style=dark >
-      <img src=https://img.shields.io/badge/Release-v1.0-brightgreen.svg >
   </div>
   <p align="center">
     🎉Now, <b>cache-dit</b> covers almost <b>All</b> Diffusers' <b>DiT</b> Pipelines🎉<br>
@@ -166,10 +170,11 @@ Dynamic: requires-python
 
 </details>
 
-## 🔥Hightlight <a href="https://huggingface.co/docs/diffusers/main/en/optimization/cache_dit"><img src=https://img.shields.io/badge/🤗Diffusers-ecosystem-yellow.svg ></a>
+## 🔥Hightlight <a href="https://huggingface.co/docs/diffusers/main/en/optimization/cache_dit"><img src=https://img.shields.io/badge/🤗Diffusers-ecosystem-yellow.svg ></a> <a href="https://hellogithub.com/repository/vipshop/cache-dit" target="_blank"><img src="https://api.hellogithub.com/v1/widgets/recommend.svg?rid=b8b03b3b32a449ea84cfc2b96cd384f3&claim_uid=ofSCbzTmdeQk3FD&theme=small" alt="Featured｜HelloGitHub" /></a> <a href="https://pypi.org/project/cache-dit/"><img src=https://img.shields.io/pypi/dm/cache-dit.svg ></a> <img src=https://img.shields.io/badge/Models-30+-hotpink.svg > <img src=https://img.shields.io/badge/Pipelines-~100+-hotpink.svg >
+
 
 We are excited to announce that the **first API-stable version (v1.0.0)** of cache-dit has finally been released!
-**[cache-dit](https://github.com/vipshop/cache-dit)** is a **Unified**, **Flexible**, and **Training-free** cache acceleration framework for 🤗 Diffusers, enabling cache acceleration with just **one line** of code. Key features: **Unified Cache APIs**, **Forward Pattern Matching**, **Automatic Block Adapter**, **Hybrid Forward Pattern**, **DBCache**, **DBPrune**, **TaylorSeer Calibrator**, and **Cache CFG**.   
+**[cache-dit](https://github.com/vipshop/cache-dit)** is a **Unified**, **Flexible**, and **Training-free** cache acceleration framework for 🤗 Diffusers, enabling cache acceleration with just **one line** of code. Key features: **Unified Cache APIs**, **Forward Pattern Matching**, **Automatic Block Adapter**, **DBCache**, **DBPrune**, **Hybrid TaylorSeer Calibrator**, **Hybrid Cache CFG**, **Context Parallelism**, **Tensor Parallelism**, **Torch Compile Compatible** and **🎉SOTA** performance.
 
 ```bash
 pip3 install -U cache-dit # pip3 install git+https://github.com/vipshop/cache-dit.git
@@ -192,7 +197,7 @@ You can install the stable release of cache-dit from PyPI, or the latest develop
 - **[🎉Easy New Model Integration](./docs/User_Guide.md#automatic-block-adapter)**: Features like **Unified Cache APIs**, **Forward Pattern Matching**, **Automatic Block Adapter**, **Hybrid Forward Pattern**, and **Patch Functor** make it highly functional and flexible. For example, we achieved 🎉 Day 1 support for [HunyuanImage-2.1](https://github.com/Tencent-Hunyuan/HunyuanImage-2.1) with 1.7x speedup w/o precision loss—even before it was available in the Diffusers library.  
 - **[🎉State-of-the-Art Performance](./bench/)**: Compared with algorithms including Δ-DiT, Chipmunk, FORA, DuCa, TaylorSeer and FoCa, cache-dit achieved the **SOTA** performance w/ **7.4x↑🎉** speedup on ClipScore!
 - **[🎉Support for 4/8-Steps Distilled Models](./bench/)**: Surprisingly, cache-dit's **DBCache** works for extremely few-step distilled models—something many other methods fail to do.  
-- **[🎉Compatibility with Other Optimizations](./docs/User_Guide.md#️torch-compile)**: Designed to work seamlessly with torch.compile, Quantization ([torchao](./examples/quantize/), [🔥nunchaku](./examples/quantize/)), CPU or Sequential Offloading, **[🔥Context Parallelism](./docs/User_Guide.md/#️hybrid-context-parallelism)**, Tensor Parallelism, etc.  
+- **[🎉Compatibility with Other Optimizations](./docs/User_Guide.md#️torch-compile)**: Designed to work seamlessly with torch.compile, Quantization ([torchao](./examples/quantize/), [🔥nunchaku](./examples/quantize/)), CPU or Sequential Offloading, **[🔥Context Parallelism](./docs/User_Guide.md/#️hybrid-context-parallelism)**, **[🔥Tensor Parallelism](./docs/User_Guide.md#️hybrid-tensor-parallelism)**, etc.  
 - **[🎉Hybrid Cache Acceleration](./docs/User_Guide.md#taylorseer-calibrator)**: Now supports hybrid **Block-wise Cache + Calibrator** schemes (e.g., DBCache or DBPrune + TaylorSeerCalibrator). DBCache or DBPrune acts as the **Indicator** to decide *when* to cache, while the Calibrator decides *how* to cache. More mainstream cache acceleration algorithms (e.g., FoCa) will be supported in the future, along with additional benchmarks—stay tuned for updates!  
 - **[🤗Diffusers Ecosystem Integration](https://huggingface.co/docs/diffusers/main/en/optimization/cache_dit)**: 🔥**cache-dit** has joined the Diffusers community ecosystem as the **first** DiT-specific cache acceleration framework! Check out the documentation here: <a href="https://huggingface.co/docs/diffusers/main/en/optimization/cache_dit"><img src=https://img.shields.io/badge/🤗Diffusers-ecosystem-yellow.svg ></a>
 
@@ -200,6 +205,9 @@ You can install the stable release of cache-dit from PyPI, or the latest develop
 
 ## 🔥Important News
 
+- 2025.10.28: 🔥Day 1 support [**LongCat-Video**](https://huggingface.co/meituan-longcat/LongCat-Video) with cache acceleration, ~**1.7x↑🎉**, 📚[Example](https://github.com/vipshop/cache-dit/blob/main/examples/pipeline/run_longcat_video.py).
+- 2025.10.28: 🎉Currently, **cache-dit** has been recommended by [<a href="https://hellogithub.com/repository/vipshop/cache-dit" target="_blank"><img src="https://api.hellogithub.com/v1/widgets/recommend.svg?rid=b8b03b3b32a449ea84cfc2b96cd384f3&claim_uid=ofSCbzTmdeQk3FD&theme=small" alt="Featured｜HelloGitHub" /></a>](https://hellogithub.com/repository/vipshop/cache-dit) ![](https://img.shields.io/github/stars/521xueweihan/HelloGitHub.svg)🔥
+- 2025.10.23: 🎉Now cache-dit supported the [Kandinsky5 T2V](https://github.com/ai-forever/Kandinsky-5) and [Photoroom/PRX](https://github.com/huggingface/diffusers/pull/12456) pipelines.
 - 2025.10.20: 🔥Now cache-dit supported the **[Hybrid Cache + Context Parallelism](./docs/User_Guide.md/#️hybrid-context-parallelism)** scheme!🔥
 - 2025.10.16: 🎉cache-dit + [**🔥nunchaku 4-bits**](https://github.com/nunchaku-tech/nunchaku) supported: [Qwen-Image-Lightning 4/8 steps](./examples/quantize/).
 - 2025.10.15: 🎉cache-dit now supported [**🔥nunchaku**](https://github.com/nunchaku-tech/nunchaku): Qwen-Image/FLUX.1 [4-bits examples](./examples/quantize/)
@@ -230,7 +238,7 @@ You can install the stable release of cache-dit from PyPI, or the latest develop
 
 <div id="user-guide"></div>
 
-For more advanced features such as **Unified Cache APIs**, **Forward Pattern Matching**, **Automatic Block Adapter**, **Hybrid Forward Pattern**, **Patch Functor**, **DBCache**, **TaylorSeer Calibrator**, and **Hybrid Cache CFG**, please refer to the [🎉User_Guide.md](./docs/User_Guide.md) for details.
+For more advanced features such as **Unified Cache APIs**, **Forward Pattern Matching**, **Automatic Block Adapter**, **Hybrid Forward Pattern**, **Patch Functor**, **DBCache**, **DBPrune**, **TaylorSeer Calibrator**, **Hybrid Cache CFG**, **Context Parallelism** and **Tensor Parallelism**, please refer to the [🎉User_Guide.md](./docs/User_Guide.md) for details.
 
 - [⚙️Installation](./docs/User_Guide.md#️installation)
 - [🔥Benchmarks](./docs/User_Guide.md#benchmarks)
@@ -241,12 +249,15 @@ For more advanced features such as **Unified Cache APIs**, **Forward Pattern Mat
   - [🔥Automatic Block Adapter](./docs/User_Guide.md#automatic-block-adapter)
   - [📚Hybrid Forward Pattern](./docs/User_Guide.md#hybrid-forward-pattern)
   - [📚Implement Patch Functor](./docs/User_Guide.md#implement-patch-functor)
+  - [📚Transformer-Only Interface](./docs/User_Guide.md#transformer-only-interface)
   - [🤖Cache Acceleration Stats](./docs/User_Guide.md#cache-acceleration-stats-summary)
 - [⚡️DBCache: Dual Block Cache](./docs/User_Guide.md#️dbcache-dual-block-cache)
 - [⚡️DBPrune: Dynamic Block Prune](./docs/User_Guide.md#️dbprune-dynamic-block-prune)
 - [⚡️Hybrid Cache CFG](./docs/User_Guide.md#️hybrid-cache-cfg)
 - [🔥Hybrid TaylorSeer Calibrator](./docs/User_Guide.md#taylorseer-calibrator)
-- [⚡️Hybrid Context Parallelism](./docs/User_Guide.md#context-paralleism)
+- [⚡️Hybrid Context Parallelism](./docs/User_Guide.md#context-parallelism)
+- [⚡️Hybrid Tensor Parallelism](./docs/User_Guide.md#tensor-parallelism)
+- [🤖Low-bits Quantization](./docs/User_Guide.md#quantization)
 - [🛠Metrics Command Line](./docs/User_Guide.md#metrics-cli)
 - [⚙️Torch Compile](./docs/User_Guide.md#️torch-compile)
 - [📚API Documents](./docs/User_Guide.md#api-documentation)
@@ -269,7 +280,7 @@ How to contribute? Star ⭐️ this repo to support us or check [CONTRIBUTE.md](
 
 ## 🎉Projects Using CacheDiT
 
-Here is a curated list of open-source projects integrating **CacheDiT**, including popular repositories like [jetson-containers](https://github.com/dusty-nv/jetson-containers/blob/master/packages/diffusion/cache_edit/build.sh) ![](https://img.shields.io/github/stars/dusty-nv/jetson-containers.svg), [flux-fast](https://github.com/huggingface/flux-fast) ![](https://img.shields.io/github/stars/huggingface/flux-fast.svg), and [sdnext](https://github.com/vladmandic/sdnext/discussions/4269) ![](https://img.shields.io/github/stars/vladmandic/sdnext.svg). **CacheDiT** has also been **recommended** by [Wan2.2](https://github.com/Wan-Video/Wan2.2) ![](https://img.shields.io/github/stars/Wan-Video/Wan2.2.svg), [Qwen-Image-Lightning](https://github.com/ModelTC/Qwen-Image-Lightning) ![](https://img.shields.io/github/stars/ModelTC/Qwen-Image-Lightning.svg), [Qwen-Image](https://github.com/QwenLM/Qwen-Image) ![](https://img.shields.io/github/stars/QwenLM/Qwen-Image.svg), and <a href="https://huggingface.co/docs/diffusers/main/en/optimization/cache_dit"><img src="https://img.shields.io/badge/🤗Diffusers-ecosystem-yellow.svg"></a> ![](https://img.shields.io/github/stars/huggingface/diffusers.svg), among others. We would be grateful if you could let us know if you have used CacheDiT.
+Here is a curated list of open-source projects integrating **CacheDiT**, including popular repositories like [jetson-containers](https://github.com/dusty-nv/jetson-containers/blob/master/packages/diffusion/cache_edit/build.sh) ![](https://img.shields.io/github/stars/dusty-nv/jetson-containers.svg), [flux-fast](https://github.com/huggingface/flux-fast) ![](https://img.shields.io/github/stars/huggingface/flux-fast.svg), and [sdnext](https://github.com/vladmandic/sdnext/discussions/4269) ![](https://img.shields.io/github/stars/vladmandic/sdnext.svg). 🎉**CacheDiT** has also been **recommended** by many famous projects: [Wan2.2](https://github.com/Wan-Video/Wan2.2) ![](https://img.shields.io/github/stars/Wan-Video/Wan2.2.svg), [Qwen-Image-Lightning](https://github.com/ModelTC/Qwen-Image-Lightning) ![](https://img.shields.io/github/stars/ModelTC/Qwen-Image-Lightning.svg), [Qwen-Image](https://github.com/QwenLM/Qwen-Image) ![](https://img.shields.io/github/stars/QwenLM/Qwen-Image.svg), [LongCat-Video](https://github.com/meituan-longcat/LongCat-Video) ![](https://img.shields.io/github/stars/meituan-longcat/LongCat-Video.svg), <a href="https://huggingface.co/docs/diffusers/main/en/optimization/cache_dit"><img src="https://img.shields.io/badge/🤗Diffusers-ecosystem-yellow.svg"></a> ![](https://img.shields.io/github/stars/huggingface/diffusers.svg), [<a href="https://hellogithub.com/repository/vipshop/cache-dit" target="_blank"><img src="https://api.hellogithub.com/v1/widgets/recommend.svg?rid=b8b03b3b32a449ea84cfc2b96cd384f3&claim_uid=ofSCbzTmdeQk3FD&theme=small" alt="Featured｜HelloGitHub" /></a>](https://hellogithub.com/repository/vipshop/cache-dit) ![](https://img.shields.io/github/stars/521xueweihan/HelloGitHub.svg), among others.
 
 ## ©️Acknowledgements
 
@@ -290,7 +301,7 @@ Special thanks to vipshop's Computer Vision AI Team for supporting document, tes
   title={cache-dit: A Unified, Flexible and Training-free Cache Acceleration Framework for Diffusers.},
   url={https://github.com/vipshop/cache-dit.git},
   note={Open-source software available at https://github.com/vipshop/cache-dit.git},
-  author={vipshop.com},
+  author={DefTruth, vipshop.com},
   year={2025}
 }
 ```

{cache_dit-1.0.8.dist-info → cache_dit-1.0.10.dist-info}/RECORD

@@ -1,19 +1,19 @@
 cache_dit/__init__.py,sha256=Azqj-3QMQK4HZDTGgyUtAfatUwuU-YQ4w8erJSyrsbE,2082
-cache_dit/_version.py,sha256=09KGe_qkXH8vhvX180khtkldtbrAX-u8refqVsC-Ky4,704
+cache_dit/_version.py,sha256=tHFvhjm1Ch2x5K0Hyz16rJbtaKyru5Di9vC_Puo-s9w,706
 cache_dit/logger.py,sha256=0zsu42hN-3-rgGC_C29ms1IvVpV4_b4_SwJCKSenxBE,4304
-cache_dit/utils.py,sha256=3NcEb324fNY0NYnrBTjsLURKQuckKeFe3V9Dfc_g4sc,17851
+cache_dit/utils.py,sha256=xlrjGHD-JepiAumGsxTKF7racsylEZL59_scf5y4e24,19295
 cache_dit/cache_factory/.gitignore,sha256=5Cb-qT9wsTUoMJ7vACDF7ZcLpAXhi5v-xdcWSRit988,23
-cache_dit/cache_factory/__init__.py,sha256=5UjrpxLVlmjHttTL0O14fD5oU5uKI3FKYevL613ibFQ,1848
-cache_dit/cache_factory/cache_interface.py,sha256=244uTVx83hpCpbCDgEOydi5HqG7hKHHzEoz1ApJW6lI,14627
+cache_dit/cache_factory/__init__.py,sha256=9b29gAb3Nd2Fpj-XvoMcbOMki4LwhvdBUk5GjWqXczQ,1921
+cache_dit/cache_factory/cache_interface.py,sha256=BQgN7Z4FiYKHZCjZMz8G11hYlD3WRBqPQ0ZQxyxK1xQ,18248
 cache_dit/cache_factory/cache_types.py,sha256=QnWfaS52UOXQtnoCUOwwz4ziY0dyBta6vQ6hvgtdV44,1404
-cache_dit/cache_factory/forward_pattern.py,sha256=FumlCuZ-TSmSYH0hGBHctSJ-oGLCftdZjLygqhsmdR4,2258
+cache_dit/cache_factory/forward_pattern.py,sha256=DAnldDC-B_FOMK8REtX2hx8mLZ9GLe1UWYfkxzyyMgo,2197
 cache_dit/cache_factory/params_modifier.py,sha256=2T98IbepAolWW6GwQsqUDsRzu0k65vo7BOrN3V8mKog,3606
 cache_dit/cache_factory/utils.py,sha256=S3SD6Zhexzhkqnmfo830v6oNLm8stZe32nF4VdxD_bA,2497
-cache_dit/cache_factory/block_adapters/__init__.py,sha256=eeBcWUMIvS-x3GcD1LNesW2SuB9V5mtwG9MoUBWHsL8,19765
-cache_dit/cache_factory/block_adapters/block_adapters.py,sha256=2TVK_KqiYXC7AKZ2s07fzdOzUoeUBc9P1SzQtLVzhf4,22249
-cache_dit/cache_factory/block_adapters/block_registers.py,sha256=KU0cqtLYRlij2WvuQ6erqZbxUWkb6DjvmY_sB3o_fQM,2594
+cache_dit/cache_factory/block_adapters/__init__.py,sha256=K3fQPgHW_Zoz8sjYwQlD_O8jHlo4ib4XBAh12yK7-n0,21122
+cache_dit/cache_factory/block_adapters/block_adapters.py,sha256=VJoGHUgQfWuwCY4YidvNiFMk7UQ5o3RmfgRCuePW948,25285
+cache_dit/cache_factory/block_adapters/block_registers.py,sha256=V3VJr6P0mOi49F_dWIXq_n6UkQLObZJYfLsbHVoLX28,3833
 cache_dit/cache_factory/cache_adapters/__init__.py,sha256=py71WGD3JztQ1uk6qdLVbzYcQ1rvqFidNNaQYo7tqTo,79
-cache_dit/cache_factory/cache_adapters/cache_adapter.py,sha256=WYrgV3DKxOxttl-wEKymyKIB1Po0eW73Q2_vOlGEKdQ,24080
+cache_dit/cache_factory/cache_adapters/cache_adapter.py,sha256=knMNq02aHCucGqub3XGqNYyTcmdtTCB6xHrh2eYmZ6w,25720
 cache_dit/cache_factory/cache_blocks/__init__.py,sha256=cpxzmDcUhbXcReHqaKSnWyEEbIg1H91Pz5hE3z9Xj3k,9984
 cache_dit/cache_factory/cache_blocks/offload_utils.py,sha256=wusgcqaCrwEjvv7Guy-6VXhNOgPPUrBV2sSVuRmGuvo,3513
 cache_dit/cache_factory/cache_blocks/pattern_0_1_2.py,sha256=j4bTafqU5DLQhzP_X5XwOk-QUVLWkGrX-Q6JZvBGHh0,666
@@ -21,13 +21,13 @@ cache_dit/cache_factory/cache_blocks/pattern_3_4_5.py,sha256=2qPnXVZwpQIm2oJ-Yrn
 cache_dit/cache_factory/cache_blocks/pattern_base.py,sha256=uNcPZU8b8i_-re_X1xBHkSDQSacQO7Fa69vjbfAYxOA,25275
 cache_dit/cache_factory/cache_blocks/pattern_utils.py,sha256=qOxoVTlYPQzPMrR06-7_Ce_lwNg6n5pt1KQrvxzAJhE,3124
 cache_dit/cache_factory/cache_contexts/__init__.py,sha256=7uY8fX9uhpC71VNm1HH4aDIicYn-dD3kRpPQhvc9-EI,853
-cache_dit/cache_factory/cache_contexts/cache_config.py,sha256=G0PVWgckDqeyARc72Ne_0lRtO_LftsOeMERRhbh2gCA,5739
+cache_dit/cache_factory/cache_contexts/cache_config.py,sha256=UAJpxGj5DxRAcAqSeNeKF4YCj6a2hzkQ-Hj0_UOggBE,5953
 cache_dit/cache_factory/cache_contexts/cache_context.py,sha256=fjZMEHaT1DZvUKnzY41GP0Ep8tmPEZTOsCSvG-5it5k,11269
-cache_dit/cache_factory/cache_contexts/cache_manager.py,sha256=tKtP35GDwZDoxGrQ_Okg_enlh3L-t-iqpytx8TFO_fw,30519
-cache_dit/cache_factory/cache_contexts/context_manager.py,sha256=j5zP_kwZAKla3EXbfr6JKI1vIxZuUEbZVhAPrtC4COw,853
+cache_dit/cache_factory/cache_contexts/cache_manager.py,sha256=6Orzs8prjNkCPRpsLiS_viJVbWCK8PQkT9OxC61X8zY,35398
+cache_dit/cache_factory/cache_contexts/context_manager.py,sha256=ajxAk9YvtmKZMtPcyVfx5DeGVDtCnHgNqPq2udpaazU,1077
 cache_dit/cache_factory/cache_contexts/prune_config.py,sha256=WMTh6zb480a0oJiYMlgI0cwCsDSVvs6UjyeJLiXbjP8,3216
 cache_dit/cache_factory/cache_contexts/prune_context.py,sha256=ywiT9P0w_GjIFLowzUDa6jhTohNsSGfTbanZcs9wMic,6359
-cache_dit/cache_factory/cache_contexts/prune_manager.py,sha256=rZG7HD9ATqgH4VZdMq1XtP_h2pokaotFOVx1svB3J7E,5478
+cache_dit/cache_factory/cache_contexts/prune_manager.py,sha256=u5z9SBLg-Ui1Um2Z80HlEsovG0b-Q_0MoPeyH4rtFXk,6123
 cache_dit/cache_factory/cache_contexts/calibrators/__init__.py,sha256=QTbyT8xcFEjfIp9xjbnsnlnVCNvMjUc20NjB0W-s95k,6269
 cache_dit/cache_factory/cache_contexts/calibrators/base.py,sha256=mn6ZBkChGpGwN5csrHTUGMoX6BBPvqHXSLbIExiW-EU,748
 cache_dit/cache_factory/cache_contexts/calibrators/foca.py,sha256=nhHGs_hxwW1M942BQDMJb9-9IuHdnOxp774Jrna1bJI,891
@@ -42,8 +42,8 @@ cache_dit/cache_factory/patch_functors/functor_hunyuan_dit.py,sha256=iSo5dD5uKnj
 cache_dit/cache_factory/patch_functors/functor_qwen_image_controlnet.py,sha256=D5i1Rrq1FQ49liupLcV2DW04moBqLnW9TICzfnMMzIU,10519
 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/kernels/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+cache_dit/kernels/triton_taylorseer.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 cache_dit/metrics/__init__.py,sha256=Y_JrBr9XE6NKXwyXc7d_-PaX9c_rk5FKms-IYgCyHmY,936
 cache_dit/metrics/clip_score.py,sha256=ERNCFQFJKzJdbIX9OAg-1LiSPuXUVHLOFxbf2gcENpc,3938
 cache_dit/metrics/config.py,sha256=ieOgD9ayz722RjVzk24bSIqS2D6o7TZjGk8KeXV-OLQ,551
@@ -53,18 +53,34 @@ cache_dit/metrics/inception.py,sha256=pBVe2X6ylLPIXTG4-GWDM9DWnCviMJbJ45R3ulhktR
 cache_dit/metrics/lpips.py,sha256=hrHrmdM-f2B4TKDs0xLqJO5JFaYcCjq2qNIR8oCrVkc,811
 cache_dit/metrics/metrics.py,sha256=AZbQyoavE-djvyRUZ_EfCIrWSQbiWQFo7n2dhn7XptE,40466
 cache_dit/parallelism/__init__.py,sha256=dheBG5_TZCuwctviMslpAEgB-B3N8F816bE51qsw_fU,210
-cache_dit/parallelism/parallel_backend.py,sha256=js1soTMenLeAyPMsBgdI3gWcdXoqjWgBD-PuFEywMr0,508
-cache_dit/parallelism/parallel_config.py,sha256=cOAXaniGf4CDPG5sbVktZy2rTZe49jSsnUIW2IBerGM,2106
-cache_dit/parallelism/parallel_interface.py,sha256=WPPYYaodo0PCYrn6-Haz8GcNZ2RK3EG7q6P_cH51Qj0,2202
-cache_dit/parallelism/backends/native_diffusers/__init__.py,sha256=T_6GeBA7TRiVbvtqGLLH2flkRiK0o7JBREt2xhS_-YE,242
-cache_dit/parallelism/backends/native_diffusers/parallel_difffusers.py,sha256=wHRjxRWK5E92cdSwDkZJpKQCQGZfxY53woW47rMFH2I,2844
-cache_dit/parallelism/backends/native_pytorch/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+cache_dit/parallelism/parallel_backend.py,sha256=GWpFrBjvRMRmAZYr1uek15-8aKDKKKQplGYENvbuSrg,576
+cache_dit/parallelism/parallel_config.py,sha256=rsPeyTvSjteKPWQanm_AgGKofAPS63zZzPZUeUPoxzk,2446
+cache_dit/parallelism/parallel_interface.py,sha256=wKZE-p2j89be7Y0MUD1IALBNBBL5QPtHGvHF1y1PgWg,2443
+cache_dit/parallelism/backends/native_diffusers/__init__.py,sha256=5woYGGhTi1XM-briRLSEIF5uNfYSMoY2W0wPQS-6iaY,114
+cache_dit/parallelism/backends/native_diffusers/parallel_difffusers.py,sha256=CY53v3lc36o8HXdc42xc5c0wsKdpDiswNDM3kLAuW-0,1607
+cache_dit/parallelism/backends/native_diffusers/utils.py,sha256=VAqtv9b8PTvcoYzD_CbvtRgdg9_VbtBug_5L38PADl0,266
+cache_dit/parallelism/backends/native_diffusers/context_parallelism/__init__.py,sha256=OT-CTUwBmWMErnVhiwZdwwYE67FOmWImtDXIrlxu8pU,3946
+cache_dit/parallelism/backends/native_diffusers/context_parallelism/cp_plan_registers.py,sha256=RMkr3fp7zTfYao_F-t8imDiTg9JIjiHPEsVfCLmy-mE,2392
+cache_dit/parallelism/backends/native_diffusers/context_parallelism/cp_planners.py,sha256=Z-wzi2VB8Gl8cqLSU9knGmrOJkGuXE2wzGWpO_j1Ubs,10186
+cache_dit/parallelism/backends/native_pytorch/__init__.py,sha256=qd2zxfi1zGoo-A_x5oBa7F6ZM0TvvZTRhKUCblOCckg,107
+cache_dit/parallelism/backends/native_pytorch/parallel_torch.py,sha256=1A-Gdv6JRueZNpwJdrVxRT7TBxwo2Bn-CwMbnlhJx0M,2047
+cache_dit/parallelism/backends/native_pytorch/tensor_parallelism/__init__.py,sha256=M-nEf6a22UeoIjZEhIajpUpGSQzWiNn_zmWiBNU70Fs,1662
+cache_dit/parallelism/backends/native_pytorch/tensor_parallelism/tp_plan_flux.py,sha256=1hSGbM84yBCfA2vv7smXLLfo69VmyAT_irgPi4Hs7Zo,6028
+cache_dit/parallelism/backends/native_pytorch/tensor_parallelism/tp_plan_qwen_image.py,sha256=AcRCfML0bwGeVY0gnVfMQtW9GAWc5wO4ZhxWGNDCkKc,2644
+cache_dit/parallelism/backends/native_pytorch/tensor_parallelism/tp_plan_registers.py,sha256=0mw3yJLjdFsnMscH-k-J23xdlJhlBtly3RQuVPAPOQA,1893
+cache_dit/parallelism/backends/native_pytorch/tensor_parallelism/tp_plan_wan.py,sha256=j8w5gZn3MhSsacxOVguss85GSI6bdLL61FMySDZKwJc,5191
+cache_dit/parallelism/backends/native_pytorch/tensor_parallelism/tp_planners.py,sha256=0rSroX_pVSFFWujdAqirLo3Gl6WZe-LFUtikhkOqprA,462
 cache_dit/quantize/__init__.py,sha256=rUu0V9VRjOgwXuIUHHAI-osivNjAdUsi-jpkDbFp6Gk,278
-cache_dit/quantize/quantize_ao.py,sha256=bbEUwsrMp3bMuRw8qJZREIvCHaJRQoZyfMjlu4ImRMI,6315
-cache_dit/quantize/quantize_interface.py,sha256=2s_R7xPSKuJeFpEGeLwRxnq_CqJcBG3a3lzyW5wh-UM,1241
-cache_dit-1.0.8.dist-info/licenses/LICENSE,sha256=Dqb07Ik2dV41s9nIdMUbiRWEfDqo7-dQeRiY7kPO8PE,3769
-cache_dit-1.0.8.dist-info/METADATA,sha256=cs6CuqsbXQ8w69_kg2e7yALeczPAZjd5hz8odNrsvZA,29462
-cache_dit-1.0.8.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-cache_dit-1.0.8.dist-info/entry_points.txt,sha256=FX2gysXaZx6NeK1iCLMcIdP8Q4_qikkIHtEmi3oWn8o,65
-cache_dit-1.0.8.dist-info/top_level.txt,sha256=ZJDydonLEhujzz0FOkVbO-BqfzO9d_VqRHmZU-3MOZo,10
-cache_dit-1.0.8.dist-info/RECORD,,
+cache_dit/quantize/quantize_backend.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+cache_dit/quantize/quantize_config.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+cache_dit/quantize/quantize_interface.py,sha256=dKLrLF-FDdRLQq-3CnaRzkAh70P4oObJE3-qWF7goM0,882
+cache_dit/quantize/backends/__init__.py,sha256=SL9EupOwBRzRcHZBI1ABqdHjCS9vEpFZXjA9R5ikTk8,33
+cache_dit/quantize/backends/bitsandbytes/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+cache_dit/quantize/backends/torchao/__init__.py,sha256=NK1widhJeKSx8ICCcyYZAWGUpGc2uKF5O7pLIcSCUDI,37
+cache_dit/quantize/backends/torchao/quantize_ao.py,sha256=VKxmr1c4n0yw7-1Rf4yKF2hXPdAkwQwNInDB43QcuiA,6917
+cache_dit-1.0.10.dist-info/licenses/LICENSE,sha256=Dqb07Ik2dV41s9nIdMUbiRWEfDqo7-dQeRiY7kPO8PE,3769
+cache_dit-1.0.10.dist-info/METADATA,sha256=YKa04RRkeRPHKtXuT8vfMyeJZDhz28vklpk6aMrmjUI,32329
+cache_dit-1.0.10.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+cache_dit-1.0.10.dist-info/entry_points.txt,sha256=FX2gysXaZx6NeK1iCLMcIdP8Q4_qikkIHtEmi3oWn8o,65
+cache_dit-1.0.10.dist-info/top_level.txt,sha256=ZJDydonLEhujzz0FOkVbO-BqfzO9d_VqRHmZU-3MOZo,10
+cache_dit-1.0.10.dist-info/RECORD,,