[Core] Remove tokenizer group in vLLM (#24078)
Signed-off-by: Zhuohan Li <zhuohan123@gmail.com>
This commit is contained in:
@@ -2,14 +2,13 @@
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
|
||||
from abc import ABC, abstractmethod
|
||||
from typing import Callable, List
|
||||
from typing import List
|
||||
|
||||
from vllm.config import SchedulerConfig
|
||||
from vllm.core.scheduler import Scheduler
|
||||
from vllm.engine.output_processor.stop_checker import StopChecker
|
||||
from vllm.sequence import Sequence, SequenceGroup, SequenceGroupOutput
|
||||
from vllm.sequence import SequenceGroup, SequenceGroupOutput
|
||||
from vllm.transformers_utils.detokenizer import Detokenizer
|
||||
from vllm.transformers_utils.tokenizer import AnyTokenizer
|
||||
from vllm.utils import Counter
|
||||
|
||||
|
||||
@@ -31,7 +30,6 @@ class SequenceGroupOutputProcessor(ABC):
|
||||
detokenizer: Detokenizer,
|
||||
scheduler: List[Scheduler],
|
||||
seq_counter: Counter,
|
||||
get_tokenizer_for_seq: Callable[[Sequence], AnyTokenizer],
|
||||
stop_checker: "StopChecker",
|
||||
):
|
||||
"""Create an output processor.
|
||||
|
||||
@@ -1,13 +1,12 @@
|
||||
# SPDX-License-Identifier: Apache-2.0
|
||||
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
||||
|
||||
from typing import Callable, List, Optional, Tuple
|
||||
from typing import List, Optional, Tuple
|
||||
|
||||
from vllm.lora.request import LoRARequest
|
||||
from vllm.reasoning import ReasoningParser
|
||||
from vllm.sampling_params import SamplingParams
|
||||
from vllm.sequence import Sequence, SequenceStatus
|
||||
from vllm.transformers_utils.tokenizer import AnyTokenizer
|
||||
|
||||
|
||||
class StopChecker:
|
||||
@@ -20,12 +19,10 @@ class StopChecker:
|
||||
def __init__(
|
||||
self,
|
||||
max_model_len: int,
|
||||
get_tokenizer_for_seq: Callable[[Sequence], AnyTokenizer],
|
||||
reasoner: Optional[ReasoningParser] = None,
|
||||
):
|
||||
# Do not use it directly, but use `self._get_max_model_len`.
|
||||
self._max_model_len = max_model_len
|
||||
self.get_tokenizer_for_seq = get_tokenizer_for_seq
|
||||
self.reasoner = reasoner
|
||||
|
||||
def _get_max_model_len(self, lora_req: Optional[LoRARequest]):
|
||||
|
||||
Reference in New Issue
Block a user