Remove unused kwargs from model definitions (#13555)

vllm/model_executor/models/falcon.py

@@ -20,14 +20,14 @@
 """PyTorch Falcon model."""
 
 import math
-from typing import Iterable, List, Optional, Set, Tuple, Union
+from typing import Iterable, Optional, Set, Tuple, Union
 
 import torch
 from torch import nn
 from torch.nn import LayerNorm
 from transformers import FalconConfig as HF_FalconConfig
 
-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,
@@ -190,8 +190,6 @@ class FalconAttention(nn.Module):
         self,
         positions: torch.Tensor,
         hidden_states: torch.Tensor,
-        kv_cache: torch.Tensor,
-        attn_metadata: AttentionMetadata,
     ) -> torch.Tensor:
         qkv, bias = self.query_key_value(hidden_states)
         if bias is not None:
@@ -199,7 +197,7 @@ class FalconAttention(nn.Module):
         q, k, v = qkv.split([self.q_size, self.kv_size, self.kv_size], dim=-1)
         if self.use_rotary:
             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)
         attn_output, bias = self.dense(attn_output)
         return attn_output, bias
 
@@ -291,8 +289,6 @@ class FalconDecoderLayer(nn.Module):
         self,
         positions: torch.Tensor,
         hidden_states: torch.Tensor,
-        kv_cache: torch.Tensor,
-        attn_metadata: AttentionMetadata,
     ) -> torch.Tensor:
         residual = hidden_states
 
@@ -306,8 +302,6 @@ class FalconDecoderLayer(nn.Module):
         attention_output, attention_bias = self.self_attention(
             positions=positions,
             hidden_states=attention_layernorm_out,
-            kv_cache=kv_cache,
-            attn_metadata=attn_metadata,
         )
         if self.reduce_row_parallel_results and attention_bias is not None:
             attention_output += attention_bias
@@ -384,8 +378,6 @@ class FalconModel(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]:
@@ -396,14 +388,8 @@ class FalconModel(nn.Module):
                 hidden_states = self.get_input_embeddings(input_ids)
         else:
             hidden_states = intermediate_tensors["hidden_states"]
-        for i in range(self.start_layer, self.end_layer):
-            layer = self.h[i]
-            hidden_states = layer(
-                positions,
-                hidden_states,
-                kv_caches[i - self.start_layer],
-                attn_metadata,
-            )
+        for layer in self.h[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.ln_f(hidden_states)
@@ -450,14 +436,11 @@ class FalconForCausalLM(nn.Module, SupportsPP):
         self,
         input_ids: torch.LongTensor,
         positions: torch.Tensor,
-        kv_caches: List[torch.Tensor],
-        attn_metadata: AttentionMetadata,
         intermediate_tensors: Optional[IntermediateTensors] = None,
         inputs_embeds: Optional[torch.Tensor] = None,
     ) -> torch.Tensor:
-        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(