Adding "torch compile" annotations to moe models (#9758)

vllm/model_executor/models/phimoe.py

@@ -28,6 +28,7 @@ from torch import nn
 from transformers.configuration_utils 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_world_size
 from vllm.model_executor.layers.fused_moe import FusedMoE
@@ -429,6 +430,7 @@ class PhiMoEDecoderLayer(nn.Module):
         return hidden_states, residual
 
 
+@support_torch_compile
 class PhiMoEModel(nn.Module):
 
     def __init__(