e489ad7a21
- **Add SPDX license headers to python source files** - **Check for SPDX headers using pre-commit** commit 9d7ef44c3cfb72ca4c32e1c677d99259d10d4745 Author: Russell Bryant <rbryant@redhat.com> Date: Fri Jan 31 14:18:24 2025 -0500 Add SPDX license headers to python source files This commit adds SPDX license headers to python source files as recommended to the project by the Linux Foundation. These headers provide a concise way that is both human and machine readable for communicating license information for each source file. It helps avoid any ambiguity about the license of the code and can also be easily used by tools to help manage license compliance. The Linux Foundation runs license scans against the codebase to help ensure we are in compliance with the licenses of the code we use, including dependencies. Having these headers in place helps that tool do its job. More information can be found on the SPDX site: - https://spdx.dev/learn/handling-license-info/ Signed-off-by: Russell Bryant <rbryant@redhat.com> commit 5a1cf1cb3b80759131c73f6a9dddebccac039dea Author: Russell Bryant <rbryant@redhat.com> Date: Fri Jan 31 14:36:32 2025 -0500 Check for SPDX headers using pre-commit Signed-off-by: Russell Bryant <rbryant@redhat.com> --------- Signed-off-by: Russell Bryant <rbryant@redhat.com>
74 lines
2.1 KiB
Python
74 lines
2.1 KiB
Python
# SPDX-License-Identifier: Apache-2.0
|
|
|
|
from dataclasses import dataclass
|
|
from typing import TYPE_CHECKING, Any, Dict, List, Optional, Union
|
|
|
|
from vllm.sequence import Logprob
|
|
|
|
if TYPE_CHECKING:
|
|
from vllm.multimodal import MultiModalDataDict
|
|
|
|
|
|
@dataclass
|
|
class BeamSearchSequence:
|
|
"""A sequence for beam search.
|
|
It keeps track of the tokens and the log probability of the sequence.
|
|
The text field is optional and will only be filled when the sequence is
|
|
about to be returned to the user.
|
|
"""
|
|
# The tokens includes the prompt.
|
|
tokens: List[int]
|
|
logprobs: List[Dict[int, Logprob]]
|
|
cum_logprob: float = 0.0
|
|
text: Optional[str] = None
|
|
finish_reason: Optional[str] = None
|
|
stop_reason: Union[int, str, None] = None
|
|
multi_modal_data: Optional["MultiModalDataDict"] = None
|
|
mm_processor_kwargs: Optional[Dict[str, Any]] = None
|
|
|
|
|
|
@dataclass
|
|
class BeamSearchOutput:
|
|
"""The output of beam search.
|
|
It contains the list of the best beam search sequences.
|
|
The length of the list is equal to the beam width.
|
|
"""
|
|
sequences: List[BeamSearchSequence]
|
|
|
|
|
|
class BeamSearchInstance:
|
|
|
|
def __init__(self, prompt_tokens: List[int]):
|
|
self.beams: List[BeamSearchSequence] = [
|
|
BeamSearchSequence(tokens=prompt_tokens, logprobs=[])
|
|
]
|
|
self.completed: List[BeamSearchSequence] = []
|
|
|
|
|
|
def get_beam_search_score(
|
|
tokens: List[int],
|
|
cumulative_logprob: float,
|
|
eos_token_id: int,
|
|
length_penalty: float = 1.0,
|
|
) -> float:
|
|
"""Calculate the beam search score with length penalty.
|
|
|
|
Adapted from
|
|
|
|
https://github.com/huggingface/transformers/blob/ccb92be23def445f2afdea94c31286f84b89eb5b/src/transformers/generation/beam_search.py#L938
|
|
"""
|
|
seq_len = len(tokens)
|
|
if tokens[-1] == eos_token_id:
|
|
seq_len -= 1
|
|
|
|
return cumulative_logprob / (seq_len**length_penalty)
|
|
|
|
|
|
def create_sort_beams_key_function(eos_token_id: int, length_penalty: float):
|
|
|
|
def sort_beams_key(x: BeamSearchSequence) -> float:
|
|
return get_beam_search_score(x.tokens, x.cum_logprob, eos_token_id,
|
|
length_penalty)
|
|
|
|
return sort_beams_key
|