Remove unused kwargs from model definitions (#13555)

vllm/model_executor/models/persimmon.py

@@ -21,13 +21,13 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 """Inference-only persimmon 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 PersimmonConfig
 
-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
@@ -142,8 +142,6 @@ class PersimmonAttention(nn.Module):
         self,
         position_ids: torch.Tensor,
         hidden_states: torch.Tensor,
-        kv_cache: torch.Tensor,
-        attn_metadata: AttentionMetadata,
     ) -> torch.Tensor:
         # [seq_length, 3 x hidden_size]
         qkv, _ = self.query_key_value(hidden_states)
@@ -161,7 +159,7 @@ class PersimmonAttention(nn.Module):
             k = self._merge_heads(k)
 
         q, k = self.rotary_emb(position_ids, q, k)
-        attn_output = self.attn(q, k, v, kv_cache, attn_metadata)
+        attn_output = self.attn(q, k, v)
         output, _ = self.dense(attn_output)
         return output
 
@@ -189,8 +187,6 @@ class PersimmonDecoderLayer(nn.Module):
         self,
         position_ids: torch.Tensor,
         hidden_states: torch.Tensor,
-        kv_cache: torch.Tensor,
-        attn_metadata: AttentionMetadata,
     ) -> torch.Tensor:
         residual = hidden_states
 
@@ -200,8 +196,6 @@ class PersimmonDecoderLayer(nn.Module):
         hidden_states = self.self_attn(
             position_ids=position_ids,
             hidden_states=hidden_states,
-            kv_cache=kv_cache,
-            attn_metadata=attn_metadata,
         )
         hidden_states = residual + hidden_states
 
@@ -248,8 +242,6 @@ class PersimmonModel(nn.Module):
         self,
         input_ids: torch.Tensor,
         positions: torch.Tensor,
-        kv_caches: List[torch.Tensor],
-        attn_metadata: AttentionMetadata,
         intermediate_tensors: Optional[IntermediateTensors],
         inputs_embeds: Optional[torch.Tensor] = None,
     ) -> Union[torch.Tensor, IntermediateTensors]:
@@ -261,13 +253,8 @@ class PersimmonModel(nn.Module):
         else:
             assert intermediate_tensors is not None
             hidden_states = intermediate_tensors["hidden_states"]
-        for i in range(self.start_layer, self.end_layer):
-            hidden_states = self.layers[i](
-                positions,
-                hidden_states,
-                kv_caches[i - self.start_layer],
-                attn_metadata,
-            )
+        for layer in self.layers[self.start_layer:self.end_layer]:
+            hidden_states = layer(positions, hidden_states)
         if not get_pp_group().is_last_rank:
             return IntermediateTensors({"hidden_states": hidden_states})
         hidden_states = self.final_layernorm(hidden_states)
@@ -298,16 +285,12 @@ class PersimmonForCausalLM(nn.Module, 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,
     ):
         hidden_states = self.model(
             input_ids=input_ids,
             positions=positions,
-            kv_caches=kv_caches,
-            attn_metadata=attn_metadata,
             intermediate_tensors=intermediate_tensors,
             inputs_embeds=inputs_embeds,
         )