[CI/Build] mypy: Resolve some errors from checking vllm/engine (#9267)

Signed-off-by: Russell Bryant <rbryant@redhat.com>
This commit is contained in:
Russell Bryant
2024-10-16 22:55:59 +00:00
committed by GitHub
parent 8345045833
commit 776dbd74f1
20 changed files with 109 additions and 74 deletions
+12 -5
View File
@@ -3,7 +3,7 @@ import copy
import pickle
from contextlib import contextmanager, suppress
from typing import (Any, AsyncGenerator, Dict, Iterator, List, Mapping,
Optional, Union, overload)
Optional, Union, cast, overload)
import cloudpickle
import zmq
@@ -513,9 +513,14 @@ class MQLLMEngineClient(EngineClient):
assert (prompt is not None and pooling_params is not None
and request_id is not None)
return self._process_request(prompt, pooling_params, request_id,
lora_request, trace_headers, None,
priority)
return cast(
AsyncGenerator[EmbeddingRequestOutput, None],
self._process_request(prompt,
pooling_params,
request_id,
lora_request,
trace_headers,
priority=priority))
async def _process_request(
self,
@@ -543,7 +548,9 @@ class MQLLMEngineClient(EngineClient):
build_guided_decoding_logits_processor_async(
sampling_params=params,
tokenizer=await self.get_tokenizer(lora_request),
default_guided_backend=self.decoding_config.guided_decoding_backend
default_guided_backend=(self.decoding_config.guided_decoding_backend
if self.decoding_config
else DecodingConfig.guided_decoding_backend),
)
# 1) Create output queue for this requests.