Merge branch 'v0.9.2-dev-update' into 'v0.9.2-dev'

修复w8a8 triton config 择优位运算可能引发torch compile 编译错误，修复smquant w8a8 权重后处理位置

See merge request dcutoolkit/deeplearing/vllm!320

vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors.py

@@ -39,7 +39,7 @@ from vllm.model_executor.layers.quantization.kv_cache import BaseKVCacheMethod
 from vllm.model_executor.layers.quantization.utils.nvfp4_emulation_utils import (  # noqa: E501
     cutlass_fp4_supported)
 from vllm.platforms import current_platform
-from vllm.utils import W8a8GetCacheJSON
+
 
 import os
 from vllm import _custom_ops as ops
@@ -616,33 +616,10 @@ class CompressedTensorsLinearMethod(LinearMethodBase):
 
     def __init__(self, quantization_config: CompressedTensorsConfig):
         self.quantization_config = quantization_config
-        self.tritonsingleton= W8a8GetCacheJSON()
-        self.w8a8_strategy=int(os.getenv('W8A8_SUPPORT_METHODS', '1'))
+
         
     def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
-        n=layer.weight.shape[0]
-        k=layer.weight.shape[1]
-        
-        if self.w8a8_strategy==1:
-            if [n,k] not in self.tritonsingleton.weight_shapes:
-                self.tritonsingleton.weight_shapes.append([n,k])
-                json_file=self.tritonsingleton.get_w8a8json_name(n,k)
-                configs_dict=self.tritonsingleton.get_triton_cache(json_file,n,k)
-                
-                if configs_dict:
-                    self.tritonsingleton.triton_json_dict.update(configs_dict)
-                    
-                    for key, value in configs_dict.items():
-                        m=int(key.split('_')[0])
-                        ops.triton_int8_gemm_helper(m=m,n=n,k=k,per_token_act_quant=True,per_out_channel_weight_quant=True,use_bias=False,device=layer.weight.device,best_config=value)
-        elif self.w8a8_strategy==3:
-            layer.weight.data = layer.weight.data.T
-        else: 
-            weight_data=layer.weight.data
-            _weight=weight_data.T.contiguous().reshape(n,-1)
-            layer.weight.data=_weight
-            
-        self.tritonsingleton.gen_model_json() 
+
         layer.scheme.process_weights_after_loading(layer)   
         
     def create_weights(self, layer: torch.nn.Module,

vllm/model_executor/layers/quantization/compressed_tensors/schemes/compressed_tensors_w8a8_int8.py

@@ -18,6 +18,8 @@ from vllm.model_executor.parameter import (BasevLLMParameter,
                                            ChannelQuantScaleParameter,
                                            ModelWeightParameter,
                                            PerTensorScaleParameter)
+from vllm.utils import W8a8GetCacheJSON
+from vllm import _custom_ops as ops
 
 logger = init_logger(__name__)
 
@@ -29,6 +31,7 @@ class CompressedTensorsW8A8Int8(CompressedTensorsScheme):
                  input_symmetric: bool):
         self.strategy = strategy
         self.is_static_input_scheme = is_static_input_scheme
+        self.tritonsingleton= W8a8GetCacheJSON()
         self.w8a8_strategy=int(os.getenv('W8A8_SUPPORT_METHODS', '1'))
         self.input_symmetric = input_symmetric
 
@@ -108,6 +111,30 @@ class CompressedTensorsW8A8Int8(CompressedTensorsScheme):
     # Checkpoints are serialized in compressed-tensors format, which is
     # different from the format the kernel may want. Handle repacking here.
     def process_weights_after_loading(self, layer: torch.nn.Module) -> None:
+        n=layer.weight.shape[0]
+        k=layer.weight.shape[1]
+        
+        if self.w8a8_strategy==1:
+            if [n,k] not in self.tritonsingleton.weight_shapes:
+                self.tritonsingleton.weight_shapes.append([n,k])
+                json_file=self.tritonsingleton.get_w8a8json_name(n,k)
+                configs_dict=self.tritonsingleton.get_triton_cache(json_file,n,k)
+                
+                if configs_dict:
+                    self.tritonsingleton.triton_json_dict.update(configs_dict)
+                    
+                    for key, value in configs_dict.items():
+                        m=int(key.split('_')[0])
+                        ops.triton_int8_gemm_helper(m=m,n=n,k=k,per_token_act_quant=True,per_out_channel_weight_quant=True,use_bias=False,device=layer.weight.device,best_config=value)
+        elif self.w8a8_strategy==3:
+            layer.weight.data = layer.weight.data.T
+        else: 
+            weight_data=layer.weight.data
+            _weight=weight_data.T.contiguous().reshape(n,-1)
+            layer.weight.data=_weight
+            
+        self.tritonsingleton.gen_model_json()         
+        
         self.kernel.process_weights_after_loading(layer)
 
     def apply_weights(self, layer: torch.nn.Module, x: torch.Tensor,

vllm/model_executor/layers/quantization/utils/w8a8_utils.py

@@ -455,11 +455,11 @@ def apply_int8_linear(
         elif f"1_{n}_{k}" in  W8A8_TRITONJSON.triton_json_dict:
             if m<=16:
                 m_=m
-                #best_config=W8A8_TRITONJSON.triton_json_dict[0][f"{m}_{n}_{k}"]
+                
             elif m<=64:
-                m_= (m + 3) & -4 #取值到最近的4的倍数
+                m_= (m //4) * 4 #取值到最近的4的倍数
             elif m<=160:
-                m_=(m + 7) & -8
+                m_=(m // 8) *8
                 
             elif m<200: #256
                 m_=160