Files
vllm/vllm/platforms/tpu.py
T

18 lines
378 B
Python

from typing import Tuple
import torch
from .interface import Platform, PlatformEnum
class TpuPlatform(Platform):
_enum = PlatformEnum.TPU
@staticmethod
def get_device_capability(device_id: int = 0) -> Tuple[int, int]:
raise RuntimeError("TPU does not have device capability.")
@staticmethod
def inference_mode():
return torch.no_grad()