[BugFix] Resolved Issues For LinearMethod --> QuantConfig (#4418)

vllm/model_executor/models/bloom.py

@@ -139,7 +139,6 @@ class BloomMLP(nn.Module):
             4 * hidden_size,
             quant_config=quant_config,
         )
-        quant_config = getattr(quant_config, "quant_config", None)
         self.gelu_impl = get_act_fn("gelu", quant_config, 4 * hidden_size)
         self.dense_4h_to_h = RowParallelLinear(
             4 * hidden_size,

vllm/model_executor/models/falcon.py

@@ -203,7 +203,6 @@ class FalconMLP(nn.Module):
                                                   bias=config.bias,
                                                   skip_bias_add=True,
                                                   quant_config=quant_config)
-        quant_config = getattr(quant_config, "quant_config", None)
         self.act = get_act_fn("gelu", quant_config, 4 * hidden_size)
         self.reduce_row_parallel_results = not (config.new_decoder_architecture
                                                 or config.parallel_attn)

vllm/model_executor/models/gpt2.py

@@ -107,7 +107,6 @@ class GPT2MLP(nn.Module):
             bias=True,
             quant_config=quant_config,
         )
-        quant_config = getattr(quant_config, "quant_config", None)
         self.act = get_act_fn(config.activation_function, quant_config,
                               intermediate_size)
 

vllm/model_executor/models/gpt_bigcode.py

@@ -128,7 +128,6 @@ class GPTBigMLP(nn.Module):
             bias=True,
             quant_config=quant_config,
         )
-        quant_config = getattr(quant_config, "quant_config", None)
         self.act = get_act_fn(config.activation_function, quant_config,
                               intermediate_size)
 

vllm/model_executor/models/gpt_j.py

@@ -120,7 +120,6 @@ class GPTJMLP(nn.Module):
             hidden_size,
             quant_config=quant_config,
         )
-        quant_config = getattr(quant_config, "quant_config", None)
         self.act = get_act_fn(config.activation_function, quant_config,
                               intermediate_size)
 

vllm/model_executor/models/gpt_neox.py

@@ -119,7 +119,6 @@ class GPTNeoXMLP(nn.Module):
             config.hidden_size,
             quant_config=quant_config,
         )
-        quant_config = getattr(quant_config, "quant_config", None)
         self.act = get_act_fn(config.hidden_act, quant_config,
                               config.intermediate_size)
 

vllm/model_executor/models/mpt.py

@@ -146,7 +146,6 @@ class MPTMLP(nn.Module):
             bias=not config.no_bias,
             quant_config=quant_config,
         )
-        quant_config = getattr(quant_config, "quant_config", None)
         self.act = get_act_fn("gelu", quant_config, intermediate_size)
         self.down_proj = RowParallelLinear(
             intermediate_size,

vllm/model_executor/models/opt.py

@@ -130,7 +130,6 @@ class OPTDecoderLayer(nn.Module):
             bias=config.enable_bias,
             quant_config=quant_config,
         )
-        quant_config = getattr(quant_config, "quant_config", None)
         self.activation_fn = get_act_fn(config.activation_function,
                                         quant_config, config.ffn_dim)
         self.fc2 = RowParallelLinear(

vllm/model_executor/models/phi.py

@@ -142,7 +142,6 @@ class PhiMLP(nn.Module):
             config.hidden_size,
             quant_config=quant_config,
         )
-        quant_config = getattr(quant_config, "quant_config", None)
         self.act = get_act_fn(config.hidden_act, quant_config, n_inner)
 
     def forward(self, hidden_states):

vllm/model_executor/models/starcoder2.py

@@ -136,7 +136,6 @@ class Starcoder2MLP(nn.Module):
             bias=config.use_bias,
             quant_config=quant_config,
         )
-        quant_config = getattr(quant_config, "quant_config", None)
         self.act = get_act_fn(config.hidden_act, quant_config,
                               config.intermediate_size)