Remove unused kwargs from model definitions (#13555)

vllm/model_executor/models/dbrx.py

@@ -1,11 +1,11 @@
 # SPDX-License-Identifier: Apache-2.0
 
-from typing import Iterable, List, Optional, Set, Tuple, Union
+from typing import Iterable, Optional, Set, Tuple, Union
 
 import torch
 import torch.nn as nn
 
-from vllm.attention import Attention, AttentionMetadata
+from vllm.attention import Attention
 from vllm.config import CacheConfig, VllmConfig
 from vllm.distributed import (get_pp_group, get_tensor_model_parallel_rank,
                               get_tensor_model_parallel_world_size)
@@ -230,15 +230,13 @@ class DbrxAttention(nn.Module):
         self,
         position_ids: torch.Tensor,
         hidden_states: torch.Tensor,
-        kv_cache: torch.Tensor,
-        attn_metadata: AttentionMetadata,
     ) -> torch.Tensor:
         qkv, _ = self.Wqkv(hidden_states)
         if self.clip_qkv is not None:
             qkv.clamp_(min=-self.clip_qkv, max=self.clip_qkv)
         q, k, v = qkv.split([self.q_size, self.kv_size, self.kv_size], dim=-1)
         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)
         hidden_states, _ = self.out_proj(attn_output)
         return hidden_states
 
@@ -265,16 +263,12 @@ class DbrxFusedNormAttention(nn.Module):
         self,
         position_ids: torch.Tensor,
         hidden_states: torch.Tensor,
-        kv_cache: torch.Tensor,
-        attn_metadata: AttentionMetadata,
     ) -> torch.Tensor:
         residual = hidden_states
         hidden_states = self.norm_1(hidden_states)
         x = self.attn(
             position_ids=position_ids,
             hidden_states=hidden_states,
-            kv_cache=kv_cache,
-            attn_metadata=attn_metadata,
         )
         hidden_states = residual + x
         residual = hidden_states
@@ -303,14 +297,10 @@ class DbrxBlock(nn.Module):
         self,
         position_ids: torch.Tensor,
         hidden_states: torch.Tensor,
-        kv_cache: torch.Tensor,
-        attn_metadata: AttentionMetadata,
     ) -> torch.Tensor:
         hidden_states, residual = self.norm_attn_norm(
             position_ids=position_ids,
             hidden_states=hidden_states,
-            kv_cache=kv_cache,
-            attn_metadata=attn_metadata,
         )
         hidden_states = self.ffn(hidden_states)
         hidden_states = hidden_states + residual
@@ -353,8 +343,6 @@ class DbrxModel(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]:
@@ -366,14 +354,8 @@ class DbrxModel(nn.Module):
         else:
             assert intermediate_tensors
             hidden_states = intermediate_tensors["hidden_states"]
-        for i in range(self.start_layer, self.end_layer):
-            block = self.blocks[i]
-            hidden_states = block(
-                position_ids,
-                hidden_states,
-                kv_caches[i - self.start_layer],
-                attn_metadata,
-            )
+        for block in self.blocks[self.start_layer:self.end_layer]:
+            hidden_states = block(position_ids, hidden_states)
         if not get_pp_group().is_last_rank:
             return IntermediateTensors({"hidden_states": hidden_states})
         hidden_states = self.norm_f(hidden_states)
@@ -415,14 +397,11 @@ class DbrxForCausalLM(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,
     ) -> 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(