cache-dit 0.2.37__py3-none-any.whl → 0.3.0__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 = '0.2.37'
-__version_tuple__ = version_tuple = (0, 2, 37)
+__version__ = version = '0.3.0'
+__version_tuple__ = version_tuple = (0, 3, 0)
 
 __commit_id__ = commit_id = None

{cache_dit-0.2.37.dist-info → cache_dit-0.3.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: cache_dit
-Version: 0.2.37
+Version: 0.3.0
 Summary: 🤗 A Unified and Training-free Cache Acceleration Toolbox for Diffusion Transformers
 Author: DefTruth, vipshop.com, etc.
 Maintainer: DefTruth, vipshop.com, etc
@@ -58,7 +58,7 @@ Dynamic: requires-python
       <img src=https://img.shields.io/badge/PyPI-pass-brightgreen.svg >
       <img src=https://static.pepy.tech/badge/cache-dit >
       <img src=https://img.shields.io/badge/Python-3.10|3.11|3.12-9cf.svg >
-      <img src=https://img.shields.io/badge/Release-v0.2-brightgreen.svg >
+      <img src=https://img.shields.io/badge/Release-v0.3-brightgreen.svg >
  </div>
   <p align="center">
     <b><a href="#unified">📚Unified Cache APIs</a></b> | <a href="#forward-pattern-matching">📚Forward Pattern Matching</a> | <a href="#automatic-block-adapter">📚Automatic Block Adapter</a><br>
@@ -286,24 +286,32 @@ Comparisons between different FnBn compute block configurations show that **more
 | Config | Clip Score(↑) | ImageReward(↑) | PSNR(↑) | TFLOPs(↓) | SpeedUp(↑) |
 | --- | --- | --- | --- | --- | --- |
 | [**FLUX.1**-dev]: 50 steps | 32.9217 | 1.0412 | INF | 3726.87 | 1.00x |
-| F8B0_W8MC0_R0.08 | 33.0070 | 1.0333 | 35.2008 | 2162.19 | 1.72x |
 | F8B0_W4MC0_R0.08 | 32.9871 | 1.0370 | 33.8317 | 2064.81 | 1.80x |
-| F4B0_W4MC2_R0.12 | 32.9718 | 1.0301 | 31.9394 | 1678.98 | 2.22x |
-| F8B0_W8MC3_R0.12 | 32.9613 | 1.0270 | 34.2834 | 1977.69 | 1.88x |
 | F8B0_W4MC2_R0.12 | 32.9535 | 1.0185 | 32.7346 | 1935.73 | 1.93x |
-| F8B0_W8MC2_R0.12 | 32.9302 | 1.0227 | 34.7449 | 2072.18 | 1.80x |
 | F8B0_W4MC3_R0.12 | 32.9234 | 1.0085 | 32.5385 | 1816.58 | 2.05x |
-| F8B0_W8MC4_R0.12 | 32.9041 | 1.0140 | 33.9466 | 1897.61 | 1.96x |
 | F4B0_W4MC3_R0.12 | 32.8981 | 1.0130 | 31.8031 | 1507.83 | 2.47x |
-| F4B0_W4MC0_R0.08 | 32.8544 | 1.0065 | 32.3555 | 1654.72 | 2.25x |
-| F8B0_W4MC4_R0.12 | 32.8443 | 1.0102 | 32.4231 | 1753.48 | 2.13x |
 | F4B0_W4MC4_R0.12 | 32.8384 | 1.0065 | 31.5292 | 1400.08 | 2.66x |
-| F1B0_W4MC4_R0.12 | 32.8291 | 1.0181 | 32.9462 | 1401.61 | 2.66x |
-| F1B0_W4MC3_R0.12 | 32.8236 | 1.0166 | 33.0037 | 1457.62 | 2.56x |
-| F1B0_W4MC10_R1.0 | 32.3183 | 0.8796 | 29.6757 | 651.90 | 5.72x |
 
 The comparison between **cache-dit: DBCache** and algorithms such as Δ-DiT, Chipmunk, FORA, DuCa, TaylorSeer and FoCa is as follows. Now, in the comparison with a speedup ratio less than **3x**, cache-dit achieved the best accuracy. Please check [📚How to Reproduce?](./bench/) for more details.
 
+| Method | TFLOPs(↓) | SpeedUp(↑) | ImageReward(↑) | Clip Score(↑) |
+| --- | --- | --- | --- | --- |
+| [**FLUX.1**-dev]: 50 steps | 3726.87 | 1.00× | 0.9898 | 32.404 |
+| [**FLUX.1**-dev]: 60% steps | 2231.70 | 1.67× | 0.9663 | 32.312 |
+| Δ-DiT(N=2) | 2480.01 | 1.50× | 0.9444 | 32.273 |
+| Δ-DiT(N=3) | 1686.76 | 2.21× | 0.8721 | 32.102 |
+| [**FLUX.1**-dev]: 34% steps | 1264.63 | 3.13× | 0.9453 | 32.114 |
+| Chipmunk | 1505.87 | 2.47× | 0.9936 | 32.776 |
+| FORA (N=3) | 1320.07 | 2.82× | 0.9776 | 32.266 |
+| **[DBCache(F=4,B=0,W=4,MC=4)](https://github.com/vipshop/cache-dit)** | **1400.08** | **2.66×** | **1.0065** | **32.838** |
+| DuCa(N=5) | 978.76 | 3.80× | 0.9955 | 32.241 |
+| TaylorSeer(N=4,O=2) | 1042.27 | 3.57× | 0.9857 | 32.413 |
+| **[DBCache+TaylorSeer(F=1,B=0,O=1)](https://github.com/vipshop/cache-dit)** | **1153.05** | **3.23×** | **1.0221** | **32.819** |
+| **[FoCa(N=5) arxiv.2508.16211](https://arxiv.org/pdf/2508.16211)** | **893.54** | **4.16×** | **1.0029** | **32.948** |
+
+<details>
+<summary> Show all comparison </summary>  
+
 | Method | TFLOPs(↓) | SpeedUp(↑) | ImageReward(↑) | Clip Score(↑) |
 | --- | --- | --- | --- | --- |
 | [**FLUX.1**-dev]: 50 steps | 3726.87 | 1.00× | 0.9898 | 32.404 |
@@ -336,22 +344,20 @@ The comparison between **cache-dit: DBCache** and algorithms such as Δ-DiT, Chi
 
 NOTE: Except for DBCache, other performance data are referenced from the paper [FoCa, arxiv.2508.16211](https://arxiv.org/pdf/2508.16211).
 
+</details>
+
 ### 📚Text2Image Distillation DrawBench: Qwen-Image-Lightning
 
 Surprisingly, cache-dit: DBCache still works in the extremely few-step distill model. For example,  **Qwen-Image-Lightning w/ 4 steps**, with the F16B16 configuration, the PSNR is 34.8163, the Clip Score is 35.6109, and the ImageReward is 1.2614. It maintained a relatively high precision.
 
 | Config                     |  PSNR(↑)      | Clip Score(↑) | ImageReward(↑) | TFLOPs(↓)   | SpeedUp(↑) |
 |----------------------------|-----------|------------|--------------|----------|------------|
-| [**Lightning**]: 4 steps | INF       | 35.5797    | 1.2630       | 274.33   | 1.00x       |
+| [**Lightning**]: 4 steps   | INF       | 35.5797    | 1.2630       | 274.33   | 1.00x       |
 | F24B24_W2MC1_R0.8          | 36.3242   | 35.6224    | 1.2630       | 264.74   | 1.04x       |
 | F16B16_W2MC1_R0.8          | 34.8163   | 35.6109    | 1.2614       | 244.25   | 1.12x       |
 | F12B12_W2MC1_R0.8          | 33.8953   | 35.6535    | 1.2549       | 234.63   | 1.17x       |
 | F8B8_W2MC1_R0.8            | 33.1374   | 35.7284    | 1.2517       | 224.29   | 1.22x       |
-| F48B0_W2MC1_R0.8           | 30.0533   | 35.8483    | 1.1979       | 265.56   | 1.03x       |
-| F32B0_W2MC1_R0.8           | 29.6490   | 35.7684    | 1.2302       | 261.05   | 1.05x       |
-| F24B0_W2MC1_R0.8           | 29.6081   | 35.8599    | 1.1874       | 245.54   | 1.12x       |
-| F16B0_W2MC1_R0.8           | 29.4844   | 36.0810    | 1.1586       | 227.06   | 1.21x       |
-
+| F1B0_W2MC1_R0.8            | 31.8317   | 35.6651    | 1.2397       | 206.90   | 1.33x       |
 
 ## 🎉Unified Cache APIs
 

{cache_dit-0.2.37.dist-info → cache_dit-0.3.0.dist-info}/RECORD

@@ -1,5 +1,5 @@
 cache_dit/__init__.py,sha256=hzaexC1VQ0TxiWY6TJ1lTm-04e65WOTNHOfYryu1vFA,1284
-cache_dit/_version.py,sha256=jVUPlUOcnlQRBFP8i5PUv2oJntFMrKgk1rs1guuDZ34,706
+cache_dit/_version.py,sha256=5zTqm8rgXsWYBpB2M3Zw_K1D-aV8wP7NsBLrmMKkrAQ,704
 cache_dit/logger.py,sha256=0zsu42hN-3-rgGC_C29ms1IvVpV4_b4_SwJCKSenxBE,4304
 cache_dit/utils.py,sha256=nuHHr6NB286qE9u6klLNfhAVRMOGipihOhM8LRqznmU,10775
 cache_dit/cache_factory/.gitignore,sha256=5Cb-qT9wsTUoMJ7vACDF7ZcLpAXhi5v-xdcWSRit988,23
@@ -43,9 +43,9 @@ cache_dit/metrics/metrics.py,sha256=7UV-H2NRbhfr6dvrXEzU97Zy-BSQ5zEfm9CKtaK4ldg,
 cache_dit/quantize/__init__.py,sha256=kWYoMAyZgBXu9BJlZjTQ0dRffW9GqeeY9_iTkXrb70A,59
 cache_dit/quantize/quantize_ao.py,sha256=Fx1KW4l3gdEkdrcAYtPoDW7WKBJWrs3glOHiEwW_TgE,6160
 cache_dit/quantize/quantize_interface.py,sha256=2s_R7xPSKuJeFpEGeLwRxnq_CqJcBG3a3lzyW5wh-UM,1241
-cache_dit-0.2.37.dist-info/licenses/LICENSE,sha256=Dqb07Ik2dV41s9nIdMUbiRWEfDqo7-dQeRiY7kPO8PE,3769
-cache_dit-0.2.37.dist-info/METADATA,sha256=dLOxpSzGT1izGxxPdbFc7jDKtiSgl-XAl-JZXRkX138,45826
-cache_dit-0.2.37.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-cache_dit-0.2.37.dist-info/entry_points.txt,sha256=FX2gysXaZx6NeK1iCLMcIdP8Q4_qikkIHtEmi3oWn8o,65
-cache_dit-0.2.37.dist-info/top_level.txt,sha256=ZJDydonLEhujzz0FOkVbO-BqfzO9d_VqRHmZU-3MOZo,10
-cache_dit-0.2.37.dist-info/RECORD,,
+cache_dit-0.3.0.dist-info/licenses/LICENSE,sha256=Dqb07Ik2dV41s9nIdMUbiRWEfDqo7-dQeRiY7kPO8PE,3769
+cache_dit-0.3.0.dist-info/METADATA,sha256=NW9YEZ1Dt3y0_o89jS3iO9o9-Y83Yo0qTz2iDOGF4j0,45943
+cache_dit-0.3.0.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+cache_dit-0.3.0.dist-info/entry_points.txt,sha256=FX2gysXaZx6NeK1iCLMcIdP8Q4_qikkIHtEmi3oWn8o,65
+cache_dit-0.3.0.dist-info/top_level.txt,sha256=ZJDydonLEhujzz0FOkVbO-BqfzO9d_VqRHmZU-3MOZo,10
+cache_dit-0.3.0.dist-info/RECORD,,