[Chore][V0 Deprecation] Move LogProb to a separate file (#24055)

Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
This commit is contained in:
Woosuk Kwon
2025-09-01 12:07:53 -07:00
committed by GitHub
parent a0e0efd6bd
commit 5685370271
14 changed files with 49 additions and 42 deletions
+1 -24
View File
@@ -16,6 +16,7 @@ import msgspec
import torch
from vllm.inputs import SingletonInputs
from vllm.logprobs import Logprob, PromptLogprobs, SampleLogprobs
from vllm.multimodal import MultiModalKwargs, MultiModalPlaceholderDict
from vllm.pooling_params import PoolingParams
from vllm.sampling_params import RequestOutputKind, SamplingParams
@@ -38,30 +39,6 @@ def array_full(token_id: int, count: int):
return array(VLLM_TOKEN_ID_ARRAY_TYPE, [token_id]) * count
# We use dataclass for now because it is used for
# openai server output, and msgspec is not serializable.
# TODO(sang): Fix it.
@dataclass
class Logprob:
"""Infos for supporting OpenAI compatible logprobs and token ranks.
Attributes:
logprob: The logprob of chosen token
rank: The vocab rank of chosen token (>=1)
decoded_token: The decoded chosen token index
"""
logprob: float
rank: Optional[int] = None
decoded_token: Optional[str] = None
# {token_id -> logprob} per each sequence group. None if the corresponding
# sequence group doesn't require prompt logprob.
PromptLogprobs = list[Optional[dict[int, Logprob]]]
# {token_id -> logprob} for each sequence group.
SampleLogprobs = list[dict[int, Logprob]]
class SequenceStatus(enum.IntEnum):
"""Status of a sequence."""
WAITING = 0