cache-dit 0.2.31__py3-none-any.whl → 0.2.32__py3-none-any.whl

cache_dit/_version.py

@@ -28,7 +28,7 @@ version_tuple: VERSION_TUPLE
 commit_id: COMMIT_ID
 __commit_id__: COMMIT_ID
 
-__version__ = version = '0.2.31'
-__version_tuple__ = version_tuple = (0, 2, 31)
+__version__ = version = '0.2.32'
+__version_tuple__ = version_tuple = (0, 2, 32)
 
 __commit_id__ = commit_id = None

cache_dit/cache_factory/block_adapters/__init__.py

@@ -254,7 +254,7 @@ def skyreelsv2_adapter(pipe, **kwargs) -> BlockAdapter:
     )
 
 
-@BlockAdapterRegistry.register("SD3")
+@BlockAdapterRegistry.register("StableDiffusion3")
 def sd3_adapter(pipe, **kwargs) -> BlockAdapter:
     from diffusers import SD3Transformer2DModel
 

cache_dit/cache_factory/cache_adapters.py

@@ -114,7 +114,7 @@ class CachedAdapter:
         **cache_context_kwargs,
     ):
         # Check cache_context_kwargs
-        if not cache_context_kwargs["enable_spearate_cfg"]:
+        if cache_context_kwargs["enable_spearate_cfg"] is None:
             # Check cfg for some specific case if users don't set it as True
             if BlockAdapterRegistry.has_separate_cfg(block_adapter):
                 cache_context_kwargs["enable_spearate_cfg"] = True
@@ -131,6 +131,12 @@ class CachedAdapter:
                     f"register: {cache_context_kwargs['enable_spearate_cfg']}, "
                     f"Pipeline: {block_adapter.pipe.__class__.__name__}."
                 )
+        else:
+            logger.info(
+                f"Use custom 'enable_spearate_cfg' from cache context "
+                f"kwargs: {cache_context_kwargs['enable_spearate_cfg']}. "
+                f"Pipeline: {block_adapter.pipe.__class__.__name__}."
+            )
 
         if (
             cache_type := cache_context_kwargs.pop("cache_type", None)

cache_dit/cache_factory/cache_blocks/pattern_base.py

@@ -345,12 +345,19 @@ class CachedBlocks_Pattern_Base(torch.nn.Module):
 
         # compute hidden_states residual
         hidden_states = hidden_states.contiguous()
-        encoder_hidden_states = encoder_hidden_states.contiguous()
 
         hidden_states_residual = hidden_states - original_hidden_states
-        encoder_hidden_states_residual = (
-            encoder_hidden_states - original_encoder_hidden_states
-        )
+
+        if (
+            encoder_hidden_states is not None
+            and original_encoder_hidden_states is not None
+        ):
+            encoder_hidden_states = encoder_hidden_states.contiguous()
+            encoder_hidden_states_residual = (
+                encoder_hidden_states - original_encoder_hidden_states
+            )
+        else:
+            encoder_hidden_states_residual = None
 
         return (
             hidden_states,
@@ -400,9 +407,16 @@ class CachedBlocks_Pattern_Base(torch.nn.Module):
                 Bn_i_hidden_states_residual = (
                     hidden_states - Bn_i_original_hidden_states
                 )
-                Bn_i_encoder_hidden_states_residual = (
-                    encoder_hidden_states - Bn_i_original_encoder_hidden_states
-                )
+                if (
+                    encoder_hidden_states is not None
+                    and Bn_i_original_encoder_hidden_states is not None
+                ):
+                    Bn_i_encoder_hidden_states_residual = (
+                        encoder_hidden_states
+                        - Bn_i_original_encoder_hidden_states
+                    )
+                else:
+                    Bn_i_encoder_hidden_states_residual = None
 
                 # Save original_hidden_states for diff calculation.
                 self.cache_manager.set_Bn_buffer(

cache_dit/cache_factory/cache_interface.py

@@ -24,7 +24,7 @@ def enable_cache(
     max_continuous_cached_steps: int = -1,
     residual_diff_threshold: float = 0.08,
     # Cache CFG or not
-    enable_spearate_cfg: bool = False,
+    enable_spearate_cfg: bool | None = None,
     cfg_compute_first: bool = False,
     cfg_diff_compute_separate: bool = True,
     # Hybird TaylorSeer
@@ -70,7 +70,7 @@ 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.
-        enable_spearate_cfg (`bool`, *required*,  defaults to False):
+        enable_spearate_cfg (`bool`, *required*,  defaults to None):
             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 enable_spearate_cfg as False, for example:
             CogVideoX, HunyuanVideo, Mochi, etc.

{cache_dit-0.2.31.dist-info → cache_dit-0.2.32.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: cache_dit
-Version: 0.2.31
+Version: 0.2.32
 Summary: 🤗 A Unified and Training-free Cache Acceleration Toolbox for Diffusion Transformers
 Author: DefTruth, vipshop.com, etc.
 Maintainer: DefTruth, vipshop.com, etc
@@ -59,32 +59,37 @@ 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>most</b> mainstream Diffusers' <b>DiT</b> Pipelines🎉<br>
-  🔥<a href="#supported">Qwen-Image</a> | <a href="#supported">FLUX.1</a> | <a href="#supported">Wan 2.1</a> | <a href="#supported"> Wan 2.2 </a> | <a href="#supported">HunyuanVideo</a>🔥<br>
-  🔥<a href="#supported">HunyuanDiT</a> | <a href="#supported">HiDream</a> | <a href="#supported">Mochi</a> | <a href="#supported"> CogVideoX </a> | <a href="#supported">CogVideoX1.5</a>🔥<br>
-  🔥<a href="#supported">CogView3Plus</a> | <a href="#supported">CogView4</a> | <a href="#supported">Chroma</a> | <a href="#supported"> LTXVideo </a> | <a href="#supported">PixArt</a>🔥<br>
-  🔥<a href="#supported">Cosmos</a> | <a href="#supported">SkyReelsV2</a> | <a href="#supported">VisualCloze</a> | <a href="#supported"> ... </a> | <a href="#supported">Lumina2</a>🔥
+    🎉Now, <b>cache-dit</b> covers <b>most</b> mainstream Diffusers' <b>DiT</b> Pipelines🎉<br>
+    🔥<a href="#supported">Qwen-Image</a> | <a href="#supported">FLUX.1</a> | <a href="#supported">Qwen-Image-Lightning</a> | <a href="#supported"> Wan 2.1/2.2 </a>🔥<br>
+    🔥<a href="#supported">HunyuanVideo</a> | <a href="#supported">HunyuanDiT</a> | <a href="#supported">HiDream</a> | <a href="#supported">Mochi</a> | <a href="#supported">CogVideoX 1/1.5</a>🔥<br>
+    🔥<a href="#supported">CogView3Plus</a> | <a href="#supported">CogView4</a> | <a href="#supported">Chroma</a> | <a href="#supported"> LTXVideo </a> | <a href="#supported">PixArt</a>🔥<br>
+    🔥<a href="#supported">Cosmos</a> | <a href="#supported">SkyReelsV2</a> | <a href="#supported">VisualCloze</a> | <a href="#supported"> OmniGen </a> | <a href="#supported">Lumina 1/2</a>🔥<br>
+    🔥<a href="#supported">Allegro</a> | <a href="#supported">EasyAnimate</a> | <a href="#supported">SD 3/3.5</a> | <a href="#supported"> ... </a> | <a href="#supported">DiT-XL</a>🔥
   </p>
 </div>
 <div align='center'>
   <img src=https://github.com/vipshop/cache-dit/raw/main/assets/gifs/wan2.2.C0_Q0_NONE.gif width=160px>
   <img src=https://github.com/vipshop/cache-dit/raw/main/assets/gifs/wan2.2.C1_Q0_DBCACHE_F1B0_W2M8MC2_T1O2_R0.08.gif width=160px>
   <img src=https://github.com/vipshop/cache-dit/raw/main/assets/gifs/wan2.2.C1_Q1_fp8_w8a8_dq_DBCACHE_F1B0_W2M8MC2_T1O2_R0.08.gif width=160px>
-  <p><b>🔥Wan2.2 MoE</b> Baseline | <b><a href="https://github.com/vipshop/cache-dit">+cache-dit</a>:~2.0x↑🎉</b> | +FP8 DQ:<b>~2.4x↑🎉</b></p>
+  <p><b>🔥Wan2.2 MoE</b> | <b><a href="https://github.com/vipshop/cache-dit">+cache-dit</a>:~2.0x↑🎉</b> | +FP8 DQ:<b>~2.4x↑🎉</b></p>
   <img src=https://github.com/vipshop/cache-dit/raw/main/assets/qwen-image.C0_Q0_NONE.png width=160px>
   <img src=https://github.com/vipshop/cache-dit/raw/main/assets/qwen-image.C1_Q0_DBCACHE_F8B0_W8M0MC0_T1O4_R0.12_S23.png width=160px>
   <img src=https://github.com/vipshop/cache-dit/raw/main/assets/qwen-image.C1_Q1_fp8_w8a8_dq_DBCACHE_F8B0_W8M0MC0_T1O4_R0.12_S18.png width=160px>
-  <p><b>🔥Qwen-Image</b> Baseline | <b><a href="https://github.com/vipshop/cache-dit">+cache-dit</a>:~1.8x↑🎉</b> | +FP8 DQ:<b>~2.2x↑🎉</b><br>♥️ Please consider to leave a <b>⭐️ Star</b> to support us ~ ♥️</p>
-  </p>
+  <p><b>🔥Qwen-Image</b> | <b><a href="https://github.com/vipshop/cache-dit">+cache-dit</a>:~1.8x↑🎉</b> | +FP8 DQ:<b>~2.2x↑🎉</b></p>
+  <img src=./assets/qwen-image-lightning.4steps.C0_L1_Q0_NONE.png width=200px>
+  <img src=./assets/qwen-image-lightning.4steps.C0_L1_Q0_DBCACHE_F16B16_W2M1MC1_T0O2_R0.9_S1.png width=200px>
+  <p><b>🔥Qwen-Image-Lightning</b> 4 steps | <b><a href="https://github.com/vipshop/cache-dit">+cache-dit</a></b> 3.5 steps:<b>~1.14x↑🎉</b>
+  <br>♥️ Please consider to leave a <b>⭐️ Star</b> to support us ~ ♥️</p>
 </div>
 
 ## 🔥News  
 
+- [2025-09-08] 🔥[**Qwen-Image-Lightning**](./examples/pipeline/run_qwen_image_lightning.py) **7.1/3.5 steps🎉** inference with **[DBCache: F16B16](https://github.com/vipshop/cache-dit)**.
 - [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 to [run_qwen_image.py](./examples/pipeline/run_qwen_image.py) as an example.
-- [2025-07-13] 🎉[**FLUX.1-Dev**](https://github.com/xlite-dev/flux-faster) **3.3x↑🎉** speedup! NVIDIA L20 with **[cache-dit](https://github.com/vipshop/cache-dit)** + `compile + FP8 DQ`.
+- [2025-07-13] 🎉[**FLUX.1-dev**](https://github.com/xlite-dev/flux-faster) **3.3x↑🎉** speedup! NVIDIA L20 with **[cache-dit](https://github.com/vipshop/cache-dit)** + **compile + FP8 DQ**.
 
 <details>
 <summary> Previous News </summary>  
@@ -134,6 +139,7 @@ pip3 install git+https://github.com/vipshop/cache-dit.git
 
 Currently, **cache-dit** library supports almost **Any** Diffusion Transformers (with **Transformer Blocks** that match the specific Input and Output **patterns**). Please check [🎉Unified Cache APIs](#unified) for more details. Here are just some of the tested models listed:
 
+- [🚀Qwen-Image-Lightning](https://github.com/vipshop/cache-dit/raw/main/examples)  
 - [🚀Qwen-Image-Edit](https://github.com/vipshop/cache-dit/raw/main/examples)  
 - [🚀Qwen-Image](https://github.com/vipshop/cache-dit/raw/main/examples)  
 - [🚀FLUX.1-dev](https://github.com/vipshop/cache-dit/raw/main/examples)  
@@ -147,7 +153,10 @@ Currently, **cache-dit** library supports almost **Any** Diffusion Transformers
 - [🚀mochi-1-preview](https://github.com/vipshop/cache-dit/raw/main/examples)
 - [🚀HunyuanVideo](https://github.com/vipshop/cache-dit/raw/main/examples)
 - [🚀HunyuanDiT](https://github.com/vipshop/cache-dit/raw/main/examples)
-- [🚀HiDream](https://github.com/vipshop/cache-dit/raw/main/examples)
+- [🚀HiDream-I1-Full](https://github.com/vipshop/cache-dit/raw/main/examples)
+- [🚀PixArt-Alpha](https://github.com/vipshop/cache-dit/raw/main/examples)
+- [🚀PixArt-Sigma](https://github.com/vipshop/cache-dit/raw/main/examples)
+- [🚀SD-3/3.5](https://github.com/vipshop/cache-dit/raw/main/examples)
 
 </details>
 

{cache_dit-0.2.31.dist-info → cache_dit-0.2.32.dist-info}/RECORD

@@ -1,21 +1,21 @@
 cache_dit/__init__.py,sha256=kX9V-FegZG4c8LMwI4PTmMqH794MEW0pzDArdhC0cJw,1241
-cache_dit/_version.py,sha256=cMx3p02rk8iaGjj6X7bw0aOcGW7d-iY_EBO9S_9o-b4,706
+cache_dit/_version.py,sha256=J0YTFDgdG9rY1Xk5pUbWWGgbT2rbSasvUHcntxayVtA,706
 cache_dit/logger.py,sha256=0zsu42hN-3-rgGC_C29ms1IvVpV4_b4_SwJCKSenxBE,4304
 cache_dit/utils.py,sha256=WK7eqgH6gCYNHXNLmWyxBDU0XSHTPg7CfOcyXlGXBqE,10510
 cache_dit/cache_factory/.gitignore,sha256=5Cb-qT9wsTUoMJ7vACDF7ZcLpAXhi5v-xdcWSRit988,23
 cache_dit/cache_factory/__init__.py,sha256=Iw6-iJLFbdzCsIDZXXOw371L-HPmoeZO_P9a3sDjP5s,1103
-cache_dit/cache_factory/cache_adapters.py,sha256=6YbBSfKEGdWi9oY1ceuxi-MpHcaDYoQ-t6NTaLZITR4,17938
-cache_dit/cache_factory/cache_interface.py,sha256=y1nY6R3MucRmAnG2UJRI_tIKrRk27FktGWLbfckf3zE,8543
+cache_dit/cache_factory/cache_adapters.py,sha256=dmNX68nBD52HtQvHnNAuSn1zjDWrQdycD0qXy-w-mwc,18212
+cache_dit/cache_factory/cache_interface.py,sha256=LpyCy-tQ_GcTRAYLpMMf9hFVIktABHI6CObn5Ll8bMw,8548
 cache_dit/cache_factory/cache_types.py,sha256=ooukxQRG55uTLmaZ0SKw6gIeY6SQHhMxkbv55uj2Sqk,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=x2ivShzOy2z3p1WUArzoChR4jaLHhNXkXMSk-RPzR3g,17534
+cache_dit/cache_factory/block_adapters/__init__.py,sha256=OZM5vJwmQIkoIwVmMxKXiHqKvs31NyAva1Z91C_ko3w,17547
 cache_dit/cache_factory/block_adapters/block_adapters.py,sha256=EQBiJYyoInKU1ND69wTm7M0n5Ja4I8QW01SgRpBjSn8,21671
 cache_dit/cache_factory/block_adapters/block_registers.py,sha256=ZeN2wGPmuf2u3puSsBx8x-rl3wRo8-cWcuWNcrssVfA,2553
 cache_dit/cache_factory/cache_blocks/__init__.py,sha256=08Ox7kD05lkRKCOsVTdEZeKAWBheqpxfrAT1Nz7eclI,2916
 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=nf2f5wdxp6tfq9AhFyMyBeKiZfxh63WG1g8q-c2BBSg,10182
-cache_dit/cache_factory/cache_blocks/pattern_base.py,sha256=_sajtb-Cz8yrCRBRSiJREzFG7h6265K9pXeAz5i1meY,20814
+cache_dit/cache_factory/cache_blocks/pattern_base.py,sha256=f1ojREQcDoBtDG3dzl8t1g_Vru8140LVDRPWlY-kAXw,21311
 cache_dit/cache_factory/cache_blocks/utils.py,sha256=dGOC1tMMOvcbvEgx44eTESKn_jsv-0RZ3tRHPa3wmQ4,1315
 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
@@ -41,9 +41,9 @@ cache_dit/quantize/__init__.py,sha256=kWYoMAyZgBXu9BJlZjTQ0dRffW9GqeeY9_iTkXrb70
 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-0.2.31.dist-info/licenses/LICENSE,sha256=Dqb07Ik2dV41s9nIdMUbiRWEfDqo7-dQeRiY7kPO8PE,3769
-cache_dit-0.2.31.dist-info/METADATA,sha256=MrRvt7HL8pNm0ZsBxKO25pBcCJhHPG7HddwjT_euy_I,23198
-cache_dit-0.2.31.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-cache_dit-0.2.31.dist-info/entry_points.txt,sha256=FX2gysXaZx6NeK1iCLMcIdP8Q4_qikkIHtEmi3oWn8o,65
-cache_dit-0.2.31.dist-info/top_level.txt,sha256=ZJDydonLEhujzz0FOkVbO-BqfzO9d_VqRHmZU-3MOZo,10
-cache_dit-0.2.31.dist-info/RECORD,,
+cache_dit-0.2.32.dist-info/licenses/LICENSE,sha256=Dqb07Ik2dV41s9nIdMUbiRWEfDqo7-dQeRiY7kPO8PE,3769
+cache_dit-0.2.32.dist-info/METADATA,sha256=WQ9GP-Om05j3NBvtifkmbz5t20XBU_-KJQptrK7jQBs,24222
+cache_dit-0.2.32.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+cache_dit-0.2.32.dist-info/entry_points.txt,sha256=FX2gysXaZx6NeK1iCLMcIdP8Q4_qikkIHtEmi3oWn8o,65
+cache_dit-0.2.32.dist-info/top_level.txt,sha256=ZJDydonLEhujzz0FOkVbO-BqfzO9d_VqRHmZU-3MOZo,10
+cache_dit-0.2.32.dist-info/RECORD,,