[mypy] Enable following imports for entrypoints (#7248)

Co-authored-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Co-authored-by: Fei <dfdfcai4@gmail.com>
This commit is contained in:
Cyrus Leung
2024-08-20 23:28:21 -07:00
committed by GitHub
co-authored by Woosuk Kwon Fei
parent 4506641212
commit baaedfdb2d
26 changed files with 480 additions and 320 deletions
+21 -10
View File
@@ -3,9 +3,9 @@ from contextlib import contextmanager
from typing import (TYPE_CHECKING, Any, ClassVar, Dict, Iterable, List,
Mapping, Optional)
from typing import Sequence as GenericSequence
from typing import Set, Tuple, Type, TypeVar, Union
from typing import Set, Tuple, Type, Union
from typing_extensions import assert_never
from typing_extensions import TypeVar, assert_never
import vllm.envs as envs
from vllm.config import (CacheConfig, DecodingConfig, DeviceConfig,
@@ -43,8 +43,9 @@ from vllm.tracing import (SpanAttributes, SpanKind, extract_trace_context,
init_tracer)
from vllm.transformers_utils.config import try_get_generation_config
from vllm.transformers_utils.detokenizer import Detokenizer
from vllm.transformers_utils.tokenizer import AnyTokenizer
from vllm.transformers_utils.tokenizer_group import (
AnyTokenizer, BaseTokenizerGroup, init_tokenizer_from_configs)
BaseTokenizerGroup, init_tokenizer_from_configs)
from vllm.usage.usage_lib import (UsageContext, is_usage_stats_enabled,
usage_message)
from vllm.utils import Counter, Device
@@ -67,6 +68,7 @@ def _load_generation_config_dict(model_config: ModelConfig) -> Dict[str, Any]:
return config.to_diff_dict()
_G = TypeVar("_G", bound=BaseTokenizerGroup, default=BaseTokenizerGroup)
_O = TypeVar("_O", RequestOutput, EmbeddingRequestOutput)
PromptComponents = Tuple[Optional[str], List[int],
@@ -493,12 +495,21 @@ class LLMEngine:
"skip_tokenizer_init is True")
def get_tokenizer_group(
self,
fail_msg: str = MISSING_TOKENIZER_GROUP_MSG) -> BaseTokenizerGroup:
if self.tokenizer is None:
raise ValueError(fail_msg)
self,
group_type: Type[_G] = BaseTokenizerGroup,
*,
missing_msg: str = MISSING_TOKENIZER_GROUP_MSG,
) -> _G:
tokenizer_group = self.tokenizer
return self.tokenizer
if tokenizer_group is None:
raise ValueError(missing_msg)
if not isinstance(tokenizer_group, group_type):
raise TypeError("Invalid type of tokenizer group. "
f"Expected type: {group_type}, but "
f"found type: {type(tokenizer_group)}")
return tokenizer_group
def get_tokenizer(
self,
@@ -693,8 +704,8 @@ class LLMEngine:
* prompt token ids
'''
tokenizer = self.get_tokenizer_group("prompts must be None if "
"skip_tokenizer_init is True")
tokenizer = self.get_tokenizer_group(
missing_msg="prompts must be None if skip_tokenizer_init is True")
return tokenizer.encode(request_id=request_id,
prompt=prompt,