[torch.compile] Adding torch compile annotations to some models (#9614)

vllm/model_executor/models/baichuan.py

@@ -26,6 +26,7 @@ from torch import nn
 from transformers import PretrainedConfig
 
 from vllm.attention import Attention, AttentionMetadata
+from vllm.compilation.decorators import support_torch_compile
 from vllm.config import CacheConfig, LoRAConfig
 from vllm.distributed import (get_pp_group, get_tensor_model_parallel_rank,
                               get_tensor_model_parallel_world_size)
@@ -250,6 +251,7 @@ class BaiChuanDecoderLayer(nn.Module):
         return hidden_states, residual
 
 
+@support_torch_compile
 class BaiChuanModel(nn.Module):
 
     def __init__(self,

vllm/model_executor/models/bloom.py

@@ -24,6 +24,7 @@ from torch import nn
 from transformers import BloomConfig
 
 from vllm.attention import Attention, AttentionMetadata
+from vllm.compilation.decorators import support_torch_compile
 from vllm.config import CacheConfig
 from vllm.distributed import (get_pp_group, get_tensor_model_parallel_rank,
                               get_tensor_model_parallel_world_size)
@@ -218,6 +219,7 @@ class BloomBlock(nn.Module):
         return output
 
 
+@support_torch_compile
 class BloomModel(nn.Module):
 
     def __init__(

vllm/model_executor/models/commandr.py

@@ -28,6 +28,7 @@ from torch import nn
 from transformers import CohereConfig
 
 from vllm.attention import Attention, AttentionMetadata
+from vllm.compilation.decorators import support_torch_compile
 from vllm.config import CacheConfig, LoRAConfig
 from vllm.distributed import get_pp_group, get_tensor_model_parallel_world_size
 from vllm.model_executor.layers.activation import SiluAndMul
@@ -250,6 +251,7 @@ class CohereDecoderLayer(nn.Module):
         return hidden_states, residual
 
 
+@support_torch_compile
 class CohereModel(nn.Module):
 
     def __init__(

vllm/model_executor/models/exaone.py

@@ -29,6 +29,7 @@ import torch
 from torch import nn
 
 from vllm.attention import Attention, AttentionMetadata
+from vllm.compilation.decorators import support_torch_compile
 from vllm.config import CacheConfig, LoRAConfig
 from vllm.distributed import (get_pp_group, get_tensor_model_parallel_rank,
                               get_tensor_model_parallel_world_size)
@@ -311,6 +312,7 @@ class ExaoneDecoderLayer(nn.Module):
         return hidden_states, residual
 
 
+@support_torch_compile
 class ExaoneModel(nn.Module):
 
     def __init__(

vllm/model_executor/models/gemma.py

@@ -22,6 +22,7 @@ from torch import nn
 from transformers import GemmaConfig
 
 from vllm.attention import Attention, AttentionMetadata
+from vllm.compilation.decorators import support_torch_compile
 from vllm.config import CacheConfig, LoRAConfig
 from vllm.distributed import get_pp_group, get_tensor_model_parallel_world_size
 from vllm.logger import init_logger
@@ -239,6 +240,7 @@ class GemmaDecoderLayer(nn.Module):
         return hidden_states, residual
 
 
+@support_torch_compile
 class GemmaModel(nn.Module):
 
     def __init__(

vllm/model_executor/models/gpt2.py

@@ -24,6 +24,7 @@ from torch import nn
 from transformers import GPT2Config
 
 from vllm.attention import Attention, AttentionMetadata
+from vllm.compilation.decorators import support_torch_compile
 from vllm.config import CacheConfig
 from vllm.distributed.parallel_state import (
     get_pp_group, get_tensor_model_parallel_world_size)
@@ -182,6 +183,7 @@ class GPT2Block(nn.Module):
         return hidden_states
 
 
+@support_torch_compile
 class GPT2Model(nn.Module):
 
     def __init__(