hf_text-generation-inference/server/text_generation_server/interceptor.py

38 lines
1.1 KiB
Python
Raw Normal View History

2023-07-12 09:05:50 -06:00
import torch
import grpc
from google.rpc import status_pb2, code_pb2
from grpc_status import rpc_status
from grpc_interceptor.server import AsyncServerInterceptor
from loguru import logger
from typing import Callable, Any
class ExceptionInterceptor(AsyncServerInterceptor):
async def intercept(
self,
method: Callable,
request_or_iterator: Any,
context: grpc.ServicerContext,
method_name: str,
) -> Any:
try:
response = method(request_or_iterator, context)
return await response
except Exception as err:
method_name = method_name.split("/")[-1]
logger.exception(f"Method {method_name} encountered an error.")
2024-04-12 10:38:34 -06:00
# Runtime Error cannot be recovered from
if isinstance(err, RuntimeError):
exit(1)
2023-07-12 09:05:50 -06:00
if torch.cuda.is_available():
torch.cuda.empty_cache()
await context.abort_with_status(
rpc_status.to_status(
status_pb2.Status(code=code_pb2.INTERNAL, message=str(err))
)
)