A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 13:30 UTC
[MISC] Use string annotation types for class definitions (#17244)
Signed-off-by: Jade Zheng <zheng.shoujian@outlook.com>
Jade Zheng committed
Apr 27, 2025 at 16:39 UTC
30215ca61f81cb313d184e74c2f616f1a1a10804
2 files changed
+4
-10
vllm/platforms/cuda.py
+2
-5
@@ -21,9 +21,6 @@ from .interface import DeviceCapability, Platform, PlatformEnum, _Backend
21
22
if TYPE_CHECKING:
23
from vllm.config import ModelConfig, VllmConfig
24
-else:
25
- ModelConfig = None
26
- VllmConfig = None
24
25
logger = init_logger(__name__)
26
@@ -109,7 +106,7 @@ class CudaPlatformBase(Platform):
106
pass
107
108
@classmethod
112
- def check_and_update_config(cls, vllm_config: VllmConfig) -> None:
109
+ def check_and_update_config(cls, vllm_config: "VllmConfig") -> None:
110
parallel_config = vllm_config.parallel_config
111
scheduler_config = vllm_config.scheduler_config
112
compilation_config = vllm_config.compilation_config
@@ -308,7 +305,7 @@ class CudaPlatformBase(Platform):
305
return cls.has_device_capability(89)
306
307
@classmethod
311
- def supports_v1(cls, model_config: ModelConfig) -> bool:
308
+ def supports_v1(cls, model_config: "ModelConfig") -> bool:
309
return True
310
311
@classmethod
vllm/platforms/rocm.py
+2
-5
@@ -13,9 +13,6 @@ from .interface import DeviceCapability, Platform, PlatformEnum, _Backend
13
14
if TYPE_CHECKING:
15
from vllm.config import ModelConfig, VllmConfig
16
-else:
17
- ModelConfig = None
18
- VllmConfig = None
16
17
logger = init_logger(__name__)
18
@@ -243,7 +240,7 @@ class RocmPlatform(Platform):
240
return True
241
242
@classmethod
246
- def check_and_update_config(cls, vllm_config: VllmConfig) -> None:
243
+ def check_and_update_config(cls, vllm_config: "VllmConfig") -> None:
244
cache_config = vllm_config.cache_config
245
if cache_config and cache_config.block_size is None:
246
cache_config.block_size = 16
@@ -332,7 +329,7 @@ class RocmPlatform(Platform):
329
return torch.float8_e4m3fn
330
331
@classmethod
335
- def supports_v1(cls, model_config: ModelConfig) -> bool:
332
+ def supports_v1(cls, model_config: "ModelConfig") -> bool:
333
# V1 support on AMD gpus is experimental
334
return True
335