[torch.compile] Adding torch compile annotations to some models (#9876)

Signed-off-by: youkaichao <youkaichao@gmail.com>
Co-authored-by: youkaichao <youkaichao@gmail.com>
This commit is contained in:
Yongzao
2024-11-01 00:25:47 -07:00
committed by GitHub
co-authored by youkaichao
parent 93a76dd21d
commit 2b5bf20988
7 changed files with 12 additions and 2 deletions
+2
View File
@@ -27,6 +27,7 @@ from torch.nn import LayerNorm
from transformers import FalconConfig as HF_FalconConfig
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,
@@ -329,6 +330,7 @@ class FalconDecoderLayer(nn.Module):
return output
@support_torch_compile
class FalconModel(nn.Module):
def __init__(
+2
View File
@@ -42,6 +42,7 @@ from torch import nn
from transformers import PhiConfig
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 get_act_fn
@@ -193,6 +194,7 @@ class PhiLayer(nn.Module):
return hidden_states
@support_torch_compile
class PhiModel(nn.Module):
def __init__(self,
+2
View File
@@ -20,6 +20,7 @@ from torchvision.transforms import InterpolationMode
from transformers import PretrainedConfig
from vllm.attention import Attention, AttentionMetadata
from vllm.compilation.decorators import support_torch_compile
from vllm.config import CacheConfig, LoRAConfig, MultiModalConfig
from vllm.distributed import get_pp_group, get_tensor_model_parallel_world_size
from vllm.inputs import (INPUT_REGISTRY, DecoderOnlyInputs, InputContext,
@@ -549,6 +550,7 @@ class QWenBlock(nn.Module):
return hidden_states, residual
@support_torch_compile
class QWenModel(nn.Module):
def __init__(
+2
View File
@@ -29,6 +29,7 @@ from torch import nn
from transformers import Qwen2Config
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
@@ -237,6 +238,7 @@ class Qwen2DecoderLayer(nn.Module):
return hidden_states, residual
@support_torch_compile
class Qwen2Model(nn.Module):
def __init__(
+2
View File
@@ -30,6 +30,7 @@ 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
from vllm.distributed import (get_pp_group,
get_tensor_model_parallel_world_size,
@@ -312,6 +313,7 @@ class Qwen2MoeDecoderLayer(nn.Module):
return hidden_states, residual
@support_torch_compile
class Qwen2MoeModel(nn.Module):
def __init__(