Remove unused kwargs from model definitions (#13555)

vllm/model_executor/models/mpt.py

@@ -2,12 +2,12 @@
 
 # Adapted from https://huggingface.co/mosaicml/mpt-7b/tree/main
 import math
-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.compilation.decorators import support_torch_compile
 from vllm.config import CacheConfig, VllmConfig
 from vllm.distributed import (get_pp_group, get_tensor_model_parallel_rank,
@@ -125,8 +125,6 @@ class MPTAttention(nn.Module):
         self,
         position_ids: torch.Tensor,
         hidden_states: torch.Tensor,
-        kv_cache: torch.Tensor,
-        attn_metadata: AttentionMetadata,
     ) -> torch.Tensor:
         del position_ids  # unused.
         qkv, _ = self.Wqkv(hidden_states)
@@ -136,7 +134,7 @@ class MPTAttention(nn.Module):
         if self.qk_ln:
             q = self.q_ln(q)
             k = self.k_ln(k)
-        attn_output = self.attn(q, k, v, kv_cache, attn_metadata)
+        attn_output = self.attn(q, k, v)
         output, _ = self.out_proj(attn_output)
         return output
 
@@ -196,15 +194,11 @@ class MPTBlock(nn.Module):
         self,
         position_ids: torch.Tensor,
         hidden_states: torch.Tensor,
-        kv_cache: torch.Tensor,
-        attn_metadata: AttentionMetadata,
     ) -> torch.Tensor:
         x = self.norm_1(hidden_states)
         x = self.attn(
             position_ids=position_ids,
             hidden_states=x,
-            kv_cache=kv_cache,
-            attn_metadata=attn_metadata,
         )
         hidden_states = hidden_states + x
         x = self.norm_2(hidden_states)
@@ -253,8 +247,6 @@ class MPTModel(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]:
@@ -267,14 +259,8 @@ class MPTModel(nn.Module):
             assert intermediate_tensors is not None
             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)
@@ -306,14 +292,11 @@ class MPTForCausalLM(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(