Remove ScaledActivation for AWQ (#10057)

Signed-off-by: mgoin <michael@neuralmagic.com>

vllm/model_executor/layers/activation.py

@@ -9,7 +9,6 @@ import torch.nn.functional as F
 from vllm.distributed import (divide, get_tensor_model_parallel_rank,
                               get_tensor_model_parallel_world_size)
 from vllm.model_executor.custom_op import CustomOp
-from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.utils import set_weight_attrs
 from vllm.utils import LazyDict
 
@@ -277,28 +276,14 @@ _ACTIVATION_REGISTRY = LazyDict({
 })
 
 
-def get_act_fn(
-    act_fn_name: str,
-    quant_config: Optional[QuantizationConfig] = None,
-    intermediate_size: Optional[int] = None,
-    input_is_parallel: bool = True,
-    params_dtype: Optional[torch.dtype] = None,
-) -> nn.Module:
+def get_act_fn(act_fn_name: str) -> nn.Module:
     """Get an activation function by name."""
     act_fn_name = act_fn_name.lower()
     if act_fn_name not in _ACTIVATION_REGISTRY:
         raise ValueError(
             f"Activation function {act_fn_name!r} is not supported.")
 
-    act_fn = _ACTIVATION_REGISTRY[act_fn_name]
-    if (quant_config is not None
-            and act_fn_name in quant_config.get_scaled_act_names()):
-        if intermediate_size is None:
-            raise ValueError("intermediate_size must be specified for scaled "
-                             "activation functions.")
-        return ScaledActivation(act_fn, intermediate_size, input_is_parallel,
-                                params_dtype)
-    return act_fn
+    return _ACTIVATION_REGISTRY[act_fn_name]
 
 
 _ACTIVATION_AND_MUL_REGISTRY = LazyDict({
@@ -307,25 +292,11 @@ _ACTIVATION_AND_MUL_REGISTRY = LazyDict({
 })
 
 
-def get_act_and_mul_fn(
-    act_fn_name: str,
-    quant_config: Optional[QuantizationConfig] = None,
-    intermediate_size: Optional[int] = None,
-    input_is_parallel: bool = True,
-    params_dtype: Optional[torch.dtype] = None,
-) -> nn.Module:
+def get_act_and_mul_fn(act_fn_name: str) -> nn.Module:
     """Get an activation-and-mul (i.e. SiluAndMul) function by name."""
     act_fn_name = act_fn_name.lower()
     if act_fn_name not in _ACTIVATION_AND_MUL_REGISTRY:
         raise ValueError(
             f"Activation function {act_fn_name!r} is not supported.")
 
-    act_fn = _ACTIVATION_AND_MUL_REGISTRY[act_fn_name]
-    if (quant_config is not None
-            and act_fn_name in quant_config.get_scaled_act_names()):
-        if intermediate_size is None:
-            raise ValueError("intermediate_size must be specified for scaled "
-                             "activation functions.")
-        return ScaledActivation(act_fn, intermediate_size, input_is_parallel,
-                                params_dtype)
-    return act_fn
+    return _ACTIVATION_AND_MUL_REGISTRY[act_fn_name]

vllm/model_executor/layers/quantization/aqlm.py

@@ -213,9 +213,6 @@ class AQLMConfig(QuantizationConfig):
             return AQLMLinearMethod(self)
         return None
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
 
 class AQLMLinearMethod(LinearMethodBase):
     """Linear method for AQLM.

vllm/model_executor/layers/quantization/awq.py

@@ -77,9 +77,6 @@ class AWQConfig(QuantizationConfig):
             return AWQLinearMethod(self)
         return None
 
-    def get_scaled_act_names(self) -> List[str]:
-        return ["gelu", "gelu_fast", "gelu_new", "gelu_pytorch_tanh"]
-
 
 def is_layer_skipped_awq(prefix: str, modules_to_not_convert: List[str]):
     return any(module_name in prefix for module_name in modules_to_not_convert)

vllm/model_executor/layers/quantization/awq_marlin.py

@@ -127,9 +127,6 @@ class AWQMarlinConfig(QuantizationConfig):
             return AWQMoEMethod(self)
         return None
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
     @classmethod
     def is_awq_marlin_compatible(cls, quant_config: Dict[str, Any]):
         # Extract data from quant config.

vllm/model_executor/layers/quantization/base_config.py

@@ -133,11 +133,3 @@ class QuantizationConfig(ABC):
             method.
         """
         raise NotImplementedError
-
-    @abstractmethod
-    def get_scaled_act_names(self) -> List[str]:
-        """Returns the activation function names that should be post-scaled.
-
-        For now, this is only used by AWQ.
-        """
-        raise NotImplementedError

vllm/model_executor/layers/quantization/bitsandbytes.py

@@ -114,9 +114,6 @@ class BitsAndBytesConfig(QuantizationConfig):
             return BitsAndBytesLinearMethod(self)
         return None
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
 
 def is_layer_skipped_bnb(prefix: str, llm_int8_skip_modules: List[str]):
     # Split the prefix into its dot-separated components

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

@@ -45,9 +45,6 @@ class CompressedTensorsConfig(QuantizationConfig):
     def get_linear_method(self) -> "CompressedTensorsLinearMethod":
         return CompressedTensorsLinearMethod(self)
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
     def get_supported_act_dtypes(cls) -> List[torch.dtype]:
         return [torch.float16, torch.bfloat16]
 

vllm/model_executor/layers/quantization/deepspeedfp.py

@@ -50,9 +50,6 @@ class DeepSpeedFPConfig(QuantizationConfig):
     def get_linear_method(self) -> "DeepSpeedFPLinearMethod":
         return DeepSpeedFPLinearMethod(self)
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
     @classmethod
     def get_supported_act_dtypes(cls) -> List[torch.dtype]:
         return [torch.half, torch.bfloat16]

vllm/model_executor/layers/quantization/experts_int8.py

@@ -45,9 +45,6 @@ class ExpertsInt8Config(QuantizationConfig):
             return ExpertsInt8MoEMethod(self)
         return None
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
 
 class ExpertsInt8MoEMethod(FusedMoEMethodBase):
 

vllm/model_executor/layers/quantization/fbgemm_fp8.py

@@ -64,9 +64,6 @@ class FBGEMMFp8Config(QuantizationConfig):
             return FBGEMMFp8LinearMethod(self)
         return None
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
 
 class FBGEMMFp8LinearMethod(LinearMethodBase):
 

vllm/model_executor/layers/quantization/fp8.py

@@ -92,9 +92,6 @@ class Fp8Config(QuantizationConfig):
             return Fp8KVCacheMethod(self)
         return None
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
 
 class Fp8LinearMethod(LinearMethodBase):
     """Linear method for FP8.

vllm/model_executor/layers/quantization/gguf.py

@@ -48,9 +48,6 @@ class GGUFConfig(QuantizationConfig):
             return GGUFEmbeddingMethod(self)
         return None
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
 
 def _fuse_mul_mat(x: torch.Tensor, qweight: torch.Tensor,
                   qweight_type: int) -> torch.Tensor:

vllm/model_executor/layers/quantization/gptq.py

@@ -80,9 +80,6 @@ class GPTQConfig(QuantizationConfig):
             return GPTQLinearMethod(self)
         return None
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
 
 class ExllamaState(Enum):
 

vllm/model_executor/layers/quantization/gptq_marlin.py

@@ -125,9 +125,6 @@ class GPTQMarlinConfig(QuantizationConfig):
             return GPTQMarlinMoEMethod(self)
         return None
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
     @classmethod
     def is_gptq_marlin_compatible(cls, quant_config: Dict[str, Any]):
         # Extract data from quant config.

vllm/model_executor/layers/quantization/gptq_marlin_24.py

@@ -127,9 +127,6 @@ class GPTQMarlin24Config(QuantizationConfig):
             return GPTQMarlin24LinearMethod(self)
         return None
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
 
 class GPTQMarlin24LinearMethod(LinearMethodBase):
     """Linear method for Marlin24.

vllm/model_executor/layers/quantization/ipex_quant.py

@@ -93,12 +93,6 @@ class IPEXConfig(QuantizationConfig):
             return self.quant_method(self)
         return None
 
-    def get_scaled_act_names(self) -> List[str]:
-        if self.method == "awq":
-            return ["gelu", "gelu_fast", "gelu_new", "gelu_pytorch_tanh"]
-        else:
-            return []
-
 
 class IPEXAWQLinearMethod(AWQLinearMethod):
     """AWQ linear method using IPEX for the CPU backend.

vllm/model_executor/layers/quantization/marlin.py

@@ -110,9 +110,6 @@ class MarlinConfig(QuantizationConfig):
             return MarlinLinearMethod(self)
         return None
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
 
 class MarlinLinearMethod(LinearMethodBase):
     """Linear method for Marlin.

vllm/model_executor/layers/quantization/modelopt.py

@@ -68,9 +68,6 @@ class ModelOptFp8Config(QuantizationConfig):
             return ModelOptFp8KVCacheMethod(self)
         return None
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
 
 class ModelOptFp8KVCacheMethod(BaseKVCacheMethod):
     """

vllm/model_executor/layers/quantization/neuron_quant.py

@@ -57,9 +57,6 @@ class NeuronQuantConfig(QuantizationConfig):
                 "Neuron Quantization is only supported through"
                 " transformers_neuronx.")
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
     def get_quantization_config(self):
         from transformers_neuronx.config import QuantizationConfig
         return QuantizationConfig(quant_dtype=self.quant_dtype,

vllm/model_executor/layers/quantization/qqq.py

@@ -112,9 +112,6 @@ class QQQConfig(QuantizationConfig):
             return QQQLinearMethod(self)
         return None
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
 
 class QQQLinearMethod(LinearMethodBase):
     """Linear method for QQQ.

vllm/model_executor/layers/quantization/tpu_int8.py

@@ -50,9 +50,6 @@ class Int8TpuConfig(QuantizationConfig):
             return TPUInt8LinearMethod(self)
         return None
 
-    def get_scaled_act_names(self) -> List[str]:
-        return []
-
 
 class TPUInt8LinearMethod(LinearMethodBase):
     """Int8 Linear method for TPU Quant. """

vllm/model_executor/models/bart.py

@@ -393,8 +393,7 @@ class BartEncoderLayer(nn.Module):
             cache_config=cache_config,
             quant_config=quant_config)
         self.self_attn_layer_norm = nn.LayerNorm(self.embed_dim)
-        self.activation_fn = get_act_fn(config.activation_function,
-                                        quant_config)
+        self.activation_fn = get_act_fn(config.activation_function)
 
         ffn_hidden_size = self.embed_dim
         ffn_intermediate_size = config.encoder_ffn_dim
@@ -405,7 +404,7 @@ class BartEncoderLayer(nn.Module):
             bias=ffn_has_bias,
             quant_config=quant_config,
         )
-        self.act = get_act_fn("gelu", quant_config, ffn_intermediate_size)
+        self.act = get_act_fn("gelu")
         self.fc2 = RowParallelLinear(
             ffn_intermediate_size,
             ffn_hidden_size,
@@ -473,8 +472,7 @@ class BartDecoderLayer(nn.Module):
             config=config,
             cache_config=cache_config,
             quant_config=quant_config)
-        self.activation_fn = get_act_fn(config.activation_function,
-                                        quant_config)
+        self.activation_fn = get_act_fn(config.activation_function)
 
         self.self_attn_layer_norm = nn.LayerNorm(self.embed_dim)
         '''

vllm/model_executor/models/bloom.py

@@ -146,7 +146,7 @@ class BloomMLP(nn.Module):
             4 * hidden_size,
             quant_config=quant_config,
         )
-        self.gelu_impl = get_act_fn("gelu", quant_config, 4 * hidden_size)
+        self.gelu_impl = get_act_fn("gelu")
         self.dense_4h_to_h = RowParallelLinear(
             4 * hidden_size,
             hidden_size,

vllm/model_executor/models/falcon.py

@@ -212,7 +212,7 @@ class FalconMLP(nn.Module):
                                                   bias=config.bias,
                                                   skip_bias_add=True,
                                                   quant_config=quant_config)
-        self.act = get_act_fn("gelu", quant_config, 4 * hidden_size)
+        self.act = get_act_fn("gelu")
         self.reduce_row_parallel_results = not (config.new_decoder_architecture
                                                 or config.parallel_attn)
         self.dense_4h_to_h = RowParallelLinear(

vllm/model_executor/models/gpt2.py

@@ -123,8 +123,7 @@ class GPT2MLP(nn.Module):
             quant_config=quant_config,
             prefix=f"{prefix}.c_proj",
         )
-        self.act = get_act_fn(config.activation_function, quant_config,
-                              intermediate_size)
+        self.act = get_act_fn(config.activation_function)
 
     def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
         hidden_states, _ = self.c_fc(hidden_states)

vllm/model_executor/models/gpt_bigcode.py

@@ -135,8 +135,7 @@ class GPTBigMLP(nn.Module):
             bias=True,
             quant_config=quant_config,
         )
-        self.act = get_act_fn(config.activation_function, quant_config,
-                              intermediate_size)
+        self.act = get_act_fn(config.activation_function)
 
     def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
         hidden_states, _ = self.c_fc(hidden_states)

vllm/model_executor/models/gpt_j.py

@@ -130,8 +130,7 @@ class GPTJMLP(nn.Module):
             hidden_size,
             quant_config=quant_config,
         )
-        self.act = get_act_fn(config.activation_function, quant_config,
-                              intermediate_size)
+        self.act = get_act_fn(config.activation_function)
 
     def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
         hidden_states, _ = self.fc_in(hidden_states)

vllm/model_executor/models/gpt_neox.py

@@ -128,8 +128,7 @@ class GPTNeoXMLP(nn.Module):
             config.hidden_size,
             quant_config=quant_config,
         )
-        self.act = get_act_fn(config.hidden_act, quant_config,
-                              config.intermediate_size)
+        self.act = get_act_fn(config.hidden_act)
 
     def forward(self, hidden_states):
         hidden_states, _ = self.dense_h_to_4h(hidden_states)

vllm/model_executor/models/mpt.py

@@ -153,7 +153,7 @@ class MPTMLP(nn.Module):
             bias=not config.no_bias,
             quant_config=quant_config,
         )
-        self.act = get_act_fn("gelu", quant_config, intermediate_size)
+        self.act = get_act_fn("gelu")
         self.down_proj = RowParallelLinear(
             intermediate_size,
             hidden_size,

vllm/model_executor/models/opt.py

@@ -147,8 +147,7 @@ class OPTDecoderLayer(nn.Module):
             quant_config=quant_config,
             prefix=f"{prefix}.fc1",
         )
-        self.activation_fn = get_act_fn(config.activation_function,
-                                        quant_config, config.ffn_dim)
+        self.activation_fn = get_act_fn(config.activation_function)
         self.fc2 = RowParallelLinear(
             config.ffn_dim,
             self.embed_dim,

vllm/model_executor/models/persimmon.py

@@ -60,7 +60,7 @@ class PersimmonMLP(nn.Module):
         self.dense_4h_to_h = RowParallelLinear(config.intermediate_size,
                                                config.hidden_size,
                                                quant_config=quant_config)
-        self.act = get_act_fn(config.hidden_act, quant_config)
+        self.act = get_act_fn(config.hidden_act)
 
     def forward(self, hidden_states) -> torch.Tensor:
         hidden_states, _ = self.dense_h_to_4h(hidden_states)

vllm/model_executor/models/phi.py

@@ -152,7 +152,7 @@ class PhiMLP(nn.Module):
             config.hidden_size,
             quant_config=quant_config,
         )
-        self.act = get_act_fn(config.hidden_act, quant_config, n_inner)
+        self.act = get_act_fn(config.hidden_act)
 
     def forward(self, hidden_states):
         hidden_states, _ = self.fc1(hidden_states)

vllm/model_executor/models/qwen.py

@@ -203,7 +203,7 @@ class QwenVMLP(nn.Module):
                                          intermediate_size,
                                          bias=True,
                                          quant_config=quant_config)
-        self.act_fn = get_act_fn("gelu", quant_config, intermediate_size)
+        self.act_fn = get_act_fn("gelu")
         self.c_proj = RowParallelLinear(
             intermediate_size,
             hidden_size,

vllm/model_executor/models/starcoder2.py

@@ -139,8 +139,7 @@ class Starcoder2MLP(nn.Module):
             bias=config.use_bias,
             quant_config=quant_config,
         )
-        self.act = get_act_fn(config.hidden_act, quant_config,
-                              config.intermediate_size)
+        self.act = get_act_fn(config.hidden_act)
 
     def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
         hidden_states, _ = self.c_fc(hidden_states)