Remove unused kwargs from model definitions (#13555)

vllm/model_executor/models/gpt_bigcode.py

@@ -19,13 +19,13 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 """Inference-only GPTBigCode model compatible with HuggingFace weights."""
-from typing import Iterable, List, Optional, Set, Tuple, Union
+from typing import Iterable, Optional, Set, Tuple, Union
 
 import torch
 from torch import nn
 from transformers import GPTBigCodeConfig
 
-from vllm.attention import Attention, AttentionMetadata
+from vllm.attention import Attention
 from vllm.compilation.decorators import support_torch_compile
 from vllm.config import CacheConfig, VllmConfig
 from vllm.distributed import get_pp_group, get_tensor_model_parallel_world_size
@@ -101,8 +101,6 @@ class GPTBigCodeAttention(nn.Module):
     def forward(
         self,
         hidden_states: torch.Tensor,
-        kv_cache: torch.Tensor,
-        attn_metadata: AttentionMetadata,
     ) -> torch.Tensor:
         qkv, _ = self.c_attn(hidden_states)
         q, k, v = qkv.split(
@@ -112,7 +110,7 @@ class GPTBigCodeAttention(nn.Module):
             ],
             dim=-1,
         )
-        attn_output = self.attn(q, k, v, kv_cache, attn_metadata)
+        attn_output = self.attn(q, k, v)
         attn_output, _ = self.c_proj(attn_output)
         return attn_output
 
@@ -173,16 +171,10 @@ class GPTBigCodeBlock(nn.Module):
     def forward(
         self,
         hidden_states: torch.Tensor,
-        kv_cache: torch.Tensor,
-        attn_metadata: AttentionMetadata,
     ) -> torch.Tensor:
         residual = hidden_states
         hidden_states = self.ln_1(hidden_states)
-        attn_output = self.attn(
-            hidden_states=hidden_states,
-            kv_cache=kv_cache,
-            attn_metadata=attn_metadata,
-        )
+        attn_output = self.attn(hidden_states=hidden_states, )
         # residual connection
         hidden_states = attn_output + residual
 
@@ -234,8 +226,6 @@ class GPTBigCodeModel(nn.Module):
         self,
         input_ids: torch.Tensor,
         position_ids: torch.Tensor,
-        kv_caches: List[torch.Tensor],
-        attn_metadata: AttentionMetadata,
         intermediate_tensors: Optional[IntermediateTensors],
         inputs_embeds: Optional[torch.Tensor] = None,
     ) -> Union[torch.Tensor, IntermediateTensors]:
@@ -246,11 +236,8 @@ class GPTBigCodeModel(nn.Module):
         else:
             hidden_states = intermediate_tensors["hidden_states"]
 
-        for i in range(self.start_layer, self.end_layer):
-            layer = self.h[i]
-            hidden_states = layer(hidden_states,
-                                  kv_caches[i - self.start_layer],
-                                  attn_metadata)
+        for layer in self.h[self.start_layer:self.end_layer]:
+            hidden_states = layer(hidden_states)
 
         if not get_pp_group().is_last_rank:
             return IntermediateTensors({"hidden_states": hidden_states})
@@ -302,14 +289,11 @@ class GPTBigCodeForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
         self,
         input_ids: torch.Tensor,
         positions: torch.Tensor,
-        kv_caches: List[torch.Tensor],
-        attn_metadata: AttentionMetadata,
         intermediate_tensors: Optional[IntermediateTensors] = None,
         inputs_embeds: Optional[torch.Tensor] = None,
     ) -> Union[torch.Tensor, IntermediateTensors]:
-        hidden_states = self.transformer(input_ids, positions, kv_caches,
-                                         attn_metadata, intermediate_tensors,
-                                         inputs_embeds)
+        hidden_states = self.transformer(input_ids, positions,
+                                         intermediate_tensors, inputs_embeds)
         return hidden_states
 
     def compute_logits(