Remove unused kwargs from model definitions (#13555)

vllm/model_executor/models/qwen2.py

@@ -23,13 +23,13 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 """Inference-only Qwen2 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 Qwen2Config
 
-from vllm.attention import Attention, AttentionMetadata, AttentionType
+from vllm.attention import Attention, AttentionType
 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
@@ -170,13 +170,11 @@ class Qwen2Attention(nn.Module):
         self,
         positions: torch.Tensor,
         hidden_states: torch.Tensor,
-        kv_cache: torch.Tensor,
-        attn_metadata: AttentionMetadata,
     ) -> torch.Tensor:
         qkv, _ = self.qkv_proj(hidden_states)
         q, k, v = qkv.split([self.q_size, self.kv_size, self.kv_size], dim=-1)
         q, k = self.rotary_emb(positions, q, k)
-        attn_output = self.attn(q, k, v, kv_cache, attn_metadata)
+        attn_output = self.attn(q, k, v)
         output, _ = self.o_proj(attn_output)
         return output
 
@@ -233,8 +231,6 @@ class Qwen2DecoderLayer(nn.Module):
         self,
         positions: torch.Tensor,
         hidden_states: torch.Tensor,
-        kv_cache: torch.Tensor,
-        attn_metadata: AttentionMetadata,
         residual: Optional[torch.Tensor],
     ) -> Tuple[torch.Tensor, torch.Tensor]:
         # Self Attention
@@ -247,8 +243,6 @@ class Qwen2DecoderLayer(nn.Module):
         hidden_states = self.self_attn(
             positions=positions,
             hidden_states=hidden_states,
-            kv_cache=kv_cache,
-            attn_metadata=attn_metadata,
         )
 
         # Fully Connected
@@ -328,8 +322,6 @@ class Qwen2Model(nn.Module):
         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]:
@@ -343,13 +335,10 @@ class Qwen2Model(nn.Module):
             assert intermediate_tensors is not None
             hidden_states = intermediate_tensors["hidden_states"]
             residual = intermediate_tensors["residual"]
-        for i in range(self.start_layer, self.end_layer):
-            layer = self.layers[i]
+        for layer in self.layers[self.start_layer:self.end_layer]:
             hidden_states, residual = layer(
                 positions,
                 hidden_states,
-                kv_caches[i - self.start_layer],
-                attn_metadata,
                 residual,
             )
         if not get_pp_group().is_last_rank:
@@ -468,13 +457,10 @@ class Qwen2ForCausalLM(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.model(input_ids, positions, kv_caches,
-                                   attn_metadata, intermediate_tensors,
+        hidden_states = self.model(input_ids, positions, intermediate_tensors,
                                    inputs_embeds)
         return hidden_states
 
@@ -553,12 +539,9 @@ class Qwen2EmbeddingModel(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,
     ) -> torch.Tensor:
-        return self.model(input_ids, positions, kv_caches, attn_metadata,
-                          intermediate_tensors)
+        return self.model(input_ids, positions, intermediate_tensors)
 
     def pooler(
         self,