diffsynth-engine 0.6.1.dev26__py3-none-any.whl → 0.6.1.dev27__py3-none-any.whl

diffsynth_engine/pipelines/wan_s2v.py

@@ -656,6 +656,8 @@ class WanSpeech2VideoPipeline(WanVideoPipeline):
         )
 
         with LoRAContext():
+            cls._auto_enable_vsa(state_dicts.model, config)
+
             dit = WanS2VDiT.from_state_dict(
                 state_dicts.model,
                 config=model_config,

diffsynth_engine/pipelines/wan_video.py

@@ -4,7 +4,7 @@ from typing import Callable, List, Dict, Tuple, Optional
 from tqdm import tqdm
 from PIL import Image
 
-from diffsynth_engine.configs import WanPipelineConfig, WanStateDicts
+from diffsynth_engine.configs import WanPipelineConfig, WanStateDicts, AttnImpl, VideoSparseAttentionParams
 from diffsynth_engine.algorithm.noise_scheduler.flow_match import RecifitedFlowScheduler
 from diffsynth_engine.algorithm.sampler import FlowMatchEulerSampler
 from diffsynth_engine.models.wan.wan_dit import WanDiT
@@ -584,6 +584,8 @@ class WanVideoPipeline(BasePipeline):
             dit_state_dict = state_dicts.model
 
         with LoRAContext():
+            cls._auto_enable_vsa(dit_state_dict, config)
+
             dit = WanDiT.from_state_dict(
                 dit_state_dict,
                 config=dit_config,
@@ -668,6 +670,16 @@ class WanVideoPipeline(BasePipeline):
             vae_type = "wan2.2-vae"
         return vae_type
 
+    @staticmethod
+    def _auto_enable_vsa(state_dict: Dict[str, torch.Tensor], config: WanPipelineConfig):
+        def has_any_key(*xs):
+            return any(x in state_dict for x in xs)
+
+        if has_any_key("blocks.0.to_gate_compress.weight", "blocks.0.self_attn.gate_compress.weight"):
+            config.dit_attn_impl = AttnImpl.VSA
+            if config.attn_params is None:
+                config.attn_params = VideoSparseAttentionParams(sparsity=0.9)
+
     def compile(self):
         self.dit.compile_repeated_blocks()
         if self.dit2 is not None:

{diffsynth_engine-0.6.1.dev26.dist-info → diffsynth_engine-0.6.1.dev27.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: diffsynth_engine
-Version: 0.6.1.dev26
+Version: 0.6.1.dev27
 Author: MuseAI x ModelScope
 Classifier: Programming Language :: Python :: 3
 Classifier: Operating System :: OS Independent

{diffsynth_engine-0.6.1.dev26.dist-info → diffsynth_engine-0.6.1.dev27.dist-info}/RECORD

@@ -148,8 +148,8 @@ diffsynth_engine/pipelines/qwen_image.py,sha256=n6Nnin8OyC9Mfp8O-3N4GNq12Mws8_hH
 diffsynth_engine/pipelines/sd_image.py,sha256=nr-Nhsnomq8CsUqhTM3i2l2zG01YjwXdfRXgr_bC3F0,17891
 diffsynth_engine/pipelines/sdxl_image.py,sha256=v7ZACGPb6EcBunL6e5E9jynSQjE7GQx8etEV-ZLP91g,21704
 diffsynth_engine/pipelines/utils.py,sha256=HZbJHErNJS1DhlwJKvZ9dY7Kh8Zdlsw3zE2e88TYGRY,2277
-diffsynth_engine/pipelines/wan_s2v.py,sha256=AUVLhLP5F0gnOV7nqWQUSZbye5ov-m44151B3zWBrAk,29323
-diffsynth_engine/pipelines/wan_video.py,sha256=Hs1iVacfrwi_0X4VNgflVUlJP5vHp0x7CF6wegidP2c,29108
+diffsynth_engine/pipelines/wan_s2v.py,sha256=QHlCLMqlmnp55iYm2mzg4qCq4jceRAP3Zt5Mubz3mAM,29384
+diffsynth_engine/pipelines/wan_video.py,sha256=rJq60LiaCoLq1PkqUzzrdvFkp6h73fc-ZUu0MiMQC-c,29668
 diffsynth_engine/processor/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 diffsynth_engine/processor/canny_processor.py,sha256=hV30NlblTkEFUAmF_O-LJrNlGVM2SFrqq6okfF8VpOo,602
 diffsynth_engine/processor/depth_processor.py,sha256=dQvs3JsnyMbz4dyI9QoR8oO-mMFBFAgNvgqeCoaU5jk,1532
@@ -187,8 +187,8 @@ diffsynth_engine/utils/video.py,sha256=8FCaeqIdUsWMgWI_6SO9SPynsToGcLCQAVYFTc4CD
 diffsynth_engine/utils/memory/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 diffsynth_engine/utils/memory/linear_regression.py,sha256=oW_EQEw13oPoyUrxiL8A7Ksa5AuJ2ynI2qhCbfAuZbg,3930
 diffsynth_engine/utils/memory/memory_predcit_model.py,sha256=EXprSl_zlVjgfMWNXP-iw83Ot3hyMcgYaRPv-dvyL84,3943
-diffsynth_engine-0.6.1.dev26.dist-info/licenses/LICENSE,sha256=x7aBqQuVI0IYnftgoTPI_A0I_rjdjPPQkjnU6N2nikM,11346
-diffsynth_engine-0.6.1.dev26.dist-info/METADATA,sha256=z6sjXpooZoFJJGqqdE_DFtsi2f3aqhjLBbyXPX0RdgE,1164
-diffsynth_engine-0.6.1.dev26.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-diffsynth_engine-0.6.1.dev26.dist-info/top_level.txt,sha256=6zgbiIzEHLbhgDKRyX0uBJOV3F6VnGGBRIQvSiYYn6w,17
-diffsynth_engine-0.6.1.dev26.dist-info/RECORD,,
+diffsynth_engine-0.6.1.dev27.dist-info/licenses/LICENSE,sha256=x7aBqQuVI0IYnftgoTPI_A0I_rjdjPPQkjnU6N2nikM,11346
+diffsynth_engine-0.6.1.dev27.dist-info/METADATA,sha256=w8FRm_Fr7AZp3TPFh1TUHk93eWxm9CFAZcU8S4qwKj0,1164
+diffsynth_engine-0.6.1.dev27.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+diffsynth_engine-0.6.1.dev27.dist-info/top_level.txt,sha256=6zgbiIzEHLbhgDKRyX0uBJOV3F6VnGGBRIQvSiYYn6w,17
+diffsynth_engine-0.6.1.dev27.dist-info/RECORD,,