cache-dit 0.2.7__py3-none-any.whl → 0.2.9__py3-none-any.whl

cache_dit/_version.py

@@ -17,5 +17,5 @@ __version__: str
 __version_tuple__: VERSION_TUPLE
 version_tuple: VERSION_TUPLE
 
-__version__ = version = '0.2.7'
-__version_tuple__ = version_tuple = (0, 2, 7)
+__version__ = version = '0.2.9'
+__version_tuple__ = version_tuple = (0, 2, 9)

cache_dit/cache_factory/dual_block_cache/cache_context.py

@@ -1162,7 +1162,6 @@ class DBCachedTransformerBlocks(torch.nn.Module):
 
         torch._dynamo.graph_break()
         if can_use_cache:
-            torch._dynamo.graph_break()
             add_cached_step()
             del Fn_hidden_states_residual
             hidden_states, encoder_hidden_states = apply_hidden_states_residual(
@@ -1189,7 +1188,6 @@ class DBCachedTransformerBlocks(torch.nn.Module):
                 )
             )
         else:
-            torch._dynamo.graph_break()
             set_Fn_buffer(Fn_hidden_states_residual, prefix="Fn_residual")
             if is_l1_diff_enabled():
                 # for hidden states L1 diff
@@ -1797,7 +1795,6 @@ class DBCachedTransformerBlocks(torch.nn.Module):
                 f"the number of single transformer blocks {len(self.single_transformer_blocks)}"
             )
 
-            torch._dynamo.graph_break()
             hidden_states = torch.cat(
                 [encoder_hidden_states, hidden_states], dim=1
             )
@@ -1829,13 +1826,11 @@ class DBCachedTransformerBlocks(torch.nn.Module):
                 ],
                 dim=1,
             )
-            torch._dynamo.graph_break()
         else:
             assert Bn_compute_blocks() <= len(self.transformer_blocks), (
                 f"Bn_compute_blocks {Bn_compute_blocks()} must be less than "
                 f"the number of transformer blocks {len(self.transformer_blocks)}"
             )
-            torch._dynamo.graph_break()
             if len(Bn_compute_blocks_ids()) > 0:
                 for i, block in enumerate(self._Bn_transformer_blocks()):
                     hidden_states, encoder_hidden_states = (
@@ -1864,7 +1859,6 @@ class DBCachedTransformerBlocks(torch.nn.Module):
                                 encoder_hidden_states,
                                 hidden_states,
                             )
-            torch._dynamo.graph_break()
 
         hidden_states = (
             hidden_states.reshape(-1)

cache_dit/compile/utils.py

@@ -1,6 +1,7 @@
 import os
 
 import torch
+import torch.distributed as dist
 from cache_dit.logger import init_logger, logging_rank_0
 
 logger = init_logger(__name__)
@@ -50,12 +51,13 @@ def set_custom_compile_configs(
         )
         return
 
-    # Enable compute comm overlap
-    torch._inductor.config.reorder_for_compute_comm_overlap = True
-    # L20 64 GB/s, PCIe; A100/A800 NVLink 300 GB/s.
-    torch._inductor.config.intra_node_bw = (
-        64 if "L20" in torch.cuda.get_device_name() else 300
-    )
+    if dist.is_initialized():
+        # Enable compute comm overlap
+        torch._inductor.config.reorder_for_compute_comm_overlap = True
+        # L20 64 GB/s, PCIe; A100/A800 NVLink 300 GB/s.
+        torch._inductor.config.intra_node_bw = (
+            64 if "L20" in torch.cuda.get_device_name() else 300
+        )
 
     # Below are default settings for torch.compile, you can change
     # them to your needs and test the performance

{cache_dit-0.2.7.dist-info → cache_dit-0.2.9.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: cache_dit
-Version: 0.2.7
+Version: 0.2.9
 Summary: 🤗 CacheDiT: A Training-free and Easy-to-use Cache Acceleration Toolbox for Diffusion Transformers
 Author: DefTruth, vipshop.com, etc.
 Maintainer: DefTruth, vipshop.com, etc
@@ -61,6 +61,11 @@ Dynamic: requires-python
   </p>
 </div> 
 
+## 🔥News🔥
+
+- [2025-07-13] An end2end speedup example for FLUX using cache-dit is released! **[🤗flux-faster](https://github.com/xlite-dev/flux-faster)**: A forked version of [huggingface/flux-fast](https://github.com/huggingface/flux-fast) that **makes flux-fast even faster** with **[cache-dit](https://github.com/vipshop/cache-dit)**, **3.3x** speedup on NVIDIA L20 while still maintaining **high precision**.
+  
+
 ## 🤗 Introduction 
 
 <div align="center">

{cache_dit-0.2.7.dist-info → cache_dit-0.2.9.dist-info}/RECORD

@@ -1,12 +1,12 @@
 cache_dit/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-cache_dit/_version.py,sha256=Xk20v7uvkFqkpy9aLJzVngs1eKQn0FYUP2oyA1MEQUU,511
+cache_dit/_version.py,sha256=Iq6CyehddPOWDVsW9Hnb65BEkCEkAnt4bl0MAuqXKLA,511
 cache_dit/logger.py,sha256=0zsu42hN-3-rgGC_C29ms1IvVpV4_b4_SwJCKSenxBE,4304
 cache_dit/primitives.py,sha256=A2iG9YLot3gOsZSPp-_gyjqjLgJvWQRx8aitD4JQ23Y,3877
 cache_dit/cache_factory/__init__.py,sha256=5RNuhWakvvqrOV4vkqrEBA7d-V1LwcNSsjtW14mkqK8,5255
 cache_dit/cache_factory/taylorseer.py,sha256=LKSNo2ode69EVo9xrxjxAMEjz0yDGiGADeDYnEqddA8,3987
 cache_dit/cache_factory/utils.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 cache_dit/cache_factory/dual_block_cache/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-cache_dit/cache_factory/dual_block_cache/cache_context.py,sha256=7kMk6hvMDi-m2HP1qlj4p6qJhzZPjJol6IBneVGDs3E,71396
+cache_dit/cache_factory/dual_block_cache/cache_context.py,sha256=itVEb6gT2eZuncAHUmP51ZS0r6v6cGtRvnPjyeXqKH8,71156
 cache_dit/cache_factory/dual_block_cache/diffusers_adapters/__init__.py,sha256=krNAICf-aS3JLmSG8vOB9tpLa04uYRcABsC8PMbVUKY,1870
 cache_dit/cache_factory/dual_block_cache/diffusers_adapters/cogvideox.py,sha256=3xUjvDzor9AkBkDUc0N7kZqM86MIdajuigesnicNzXE,2260
 cache_dit/cache_factory/dual_block_cache/diffusers_adapters/flux.py,sha256=cIsov6Pf0dRyddqkzTA2CU-jSDotof8LQr-HIoY9T9M,2615
@@ -30,7 +30,7 @@ cache_dit/cache_factory/first_block_cache/diffusers_adapters/hunyuan_video.py,sh
 cache_dit/cache_factory/first_block_cache/diffusers_adapters/mochi.py,sha256=lQTClo52OwPbNEE4jiBZQhfC7hbtYqnYIABp_vbm_dk,2363
 cache_dit/cache_factory/first_block_cache/diffusers_adapters/wan.py,sha256=dBNzHBECAuTTA1a7kLdvZL20YzaKTAS3iciVLzKKEWA,2638
 cache_dit/compile/__init__.py,sha256=DfMdPleFFGADXLsr7zXui8BTz_y9futY6rNmNdh9y7k,63
-cache_dit/compile/utils.py,sha256=KU60xc474Anbj7Y_FLRFmNxEjVYLLXkhbtCLXO7o_Tc,3699
+cache_dit/compile/utils.py,sha256=OTvkwcezSrApZ2M1IMkYtkEmFbkfpTknhHMgoBApd6U,3786
 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
@@ -38,9 +38,9 @@ cache_dit/metrics/config.py,sha256=ieOgD9ayz722RjVzk24bSIqS2D6o7TZjGk8KeXV-OLQ,5
 cache_dit/metrics/fid.py,sha256=9Ivtazl6mW0Bon2VXa-Ia5Xj2ewxRD3V1Qkd69zYM3Y,17066
 cache_dit/metrics/inception.py,sha256=pBVe2X6ylLPIXTG4-GWDM9DWnCviMJbJ45R3ulhktR0,12759
 cache_dit/metrics/metrics.py,sha256=tzAtG_-fM1xPIBfRVFIBupvOWYzIO3xDq29Vy5rOBWc,14730
-cache_dit-0.2.7.dist-info/licenses/LICENSE,sha256=Dqb07Ik2dV41s9nIdMUbiRWEfDqo7-dQeRiY7kPO8PE,3769
-cache_dit-0.2.7.dist-info/METADATA,sha256=S0C1VGcXoWVjxmfX_755xttdoE9J0toSuHCZW9xaUBM,27608
-cache_dit-0.2.7.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-cache_dit-0.2.7.dist-info/entry_points.txt,sha256=FX2gysXaZx6NeK1iCLMcIdP8Q4_qikkIHtEmi3oWn8o,65
-cache_dit-0.2.7.dist-info/top_level.txt,sha256=ZJDydonLEhujzz0FOkVbO-BqfzO9d_VqRHmZU-3MOZo,10
-cache_dit-0.2.7.dist-info/RECORD,,
+cache_dit-0.2.9.dist-info/licenses/LICENSE,sha256=Dqb07Ik2dV41s9nIdMUbiRWEfDqo7-dQeRiY7kPO8PE,3769
+cache_dit-0.2.9.dist-info/METADATA,sha256=TdvKAftNWwijdCW8K-8iO7fITEcfllWX3FJdZ-qcRqA,28032
+cache_dit-0.2.9.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+cache_dit-0.2.9.dist-info/entry_points.txt,sha256=FX2gysXaZx6NeK1iCLMcIdP8Q4_qikkIHtEmi3oWn8o,65
+cache_dit-0.2.9.dist-info/top_level.txt,sha256=ZJDydonLEhujzz0FOkVbO-BqfzO9d_VqRHmZU-3MOZo,10
+cache_dit-0.2.9.dist-info/RECORD,,