mirror of
https://github.com/open-compass/opencompass.git
synced 2025-05-30 16:03:24 +08:00
Revert "[Dataset] Add Lifescience Sub-set Support for SciEval (#2059)"
This reverts commit c5048bfec7.
This commit is contained in:
Linchen Xiao
GitHub
parent
c5048bfec7
commit
b1b429b680
@ -695,12 +695,6 @@
|
|||||||
paper: https://arxiv.org/pdf/2009.03300
|
paper: https://arxiv.org/pdf/2009.03300
|
||||||
configpath: opencompass/configs/datasets/mmlu/mmlu_gen.py
|
configpath: opencompass/configs/datasets/mmlu/mmlu_gen.py
|
||||||
configpath_llmjudge: opencompass/configs/datasets/mmlu/mmlu_llm_judge_gen.py
|
configpath_llmjudge: opencompass/configs/datasets/mmlu/mmlu_llm_judge_gen.py
|
||||||
- SciEval:
|
|
||||||
name: SciEval
|
|
||||||
category: Understanding
|
|
||||||
paper: https://arxiv.org/pdf/2308.13149
|
|
||||||
configpath: opencompass/configs/datasets/SciEval_lifscience/SciEval_lifscience_gen.py
|
|
||||||
configpath_llmjudge: opencompass/configs/datasets/SciEval_lifscience/SciEval_lifscience_llm_judge_gen.py
|
|
||||||
- mmlu_cf:
|
- mmlu_cf:
|
||||||
name: MMLU-CF
|
name: MMLU-CF
|
||||||
category: Understanding
|
category: Understanding
|
||||||
|
|||||||
@ -1,61 +0,0 @@
|
|||||||
from opencompass.openicl.icl_prompt_template import PromptTemplate
|
|
||||||
from opencompass.openicl.icl_retriever import FixKRetriever
|
|
||||||
from opencompass.openicl.icl_inferencer import GenInferencer
|
|
||||||
from opencompass.openicl.icl_evaluator import AccwithDetailsEvaluator
|
|
||||||
from opencompass.utils.text_postprocessors import first_option_postprocess
|
|
||||||
from opencompass.datasets import SciEvalDataset # 你自己实现的类
|
|
||||||
|
|
||||||
# 只评测 biology + multiple-choice 的 test split
|
|
||||||
_hint = ('Given a question and four options, please select the right answer. '
|
|
||||||
"Your answer should be 'A', 'B', 'C' or 'D'.")
|
|
||||||
|
|
||||||
scieval_reader_cfg = dict(
|
|
||||||
input_columns=['input', 'A', 'B', 'C', 'D'],
|
|
||||||
output_column='target',
|
|
||||||
train_split='test',
|
|
||||||
)
|
|
||||||
|
|
||||||
scieval_infer_cfg = dict(
|
|
||||||
ice_template=dict(
|
|
||||||
type=PromptTemplate,
|
|
||||||
template=dict(round=[
|
|
||||||
dict(
|
|
||||||
role='HUMAN',
|
|
||||||
prompt=f'{_hint}\nQuestion: {{input}}\nA. {{A}}\nB. {{B}}\nC. {{C}}\nD. {{D}}\nAnswer: '
|
|
||||||
),
|
|
||||||
dict(role='BOT', prompt='{target}\n')
|
|
||||||
]),
|
|
||||||
),
|
|
||||||
prompt_template=dict(
|
|
||||||
type=PromptTemplate,
|
|
||||||
template=dict(
|
|
||||||
begin='</E>',
|
|
||||||
round=[
|
|
||||||
dict(
|
|
||||||
role='HUMAN',
|
|
||||||
prompt=f'{_hint}\nQuestion: {{input}}\nA. {{A}}\nB. {{B}}\nC. {{C}}\nD. {{D}}\nAnswer: '
|
|
||||||
),
|
|
||||||
],
|
|
||||||
),
|
|
||||||
ice_token='</E>',
|
|
||||||
),
|
|
||||||
retriever=dict(type=FixKRetriever, fix_id_list=[0, 1, 2, 3, 4]),
|
|
||||||
inferencer=dict(type=GenInferencer),
|
|
||||||
)
|
|
||||||
|
|
||||||
scieval_eval_cfg = dict(
|
|
||||||
evaluator=dict(type=AccwithDetailsEvaluator),
|
|
||||||
pred_postprocessor=dict(type=first_option_postprocess, options='ABCD'),
|
|
||||||
)
|
|
||||||
|
|
||||||
scieval_datasets = [
|
|
||||||
dict(
|
|
||||||
abbr='scieval_biology',
|
|
||||||
type=SciEvalDataset,
|
|
||||||
path='OpenDFM/SciEval',
|
|
||||||
name='default',
|
|
||||||
reader_cfg=scieval_reader_cfg,
|
|
||||||
infer_cfg=scieval_infer_cfg,
|
|
||||||
eval_cfg=scieval_eval_cfg,
|
|
||||||
)
|
|
||||||
]
|
|
||||||
@ -1,125 +0,0 @@
|
|||||||
# SciEval_lifescience_llmjudge_gen.py
|
|
||||||
|
|
||||||
from mmengine.config import read_base
|
|
||||||
from opencompass.openicl.icl_prompt_template import PromptTemplate
|
|
||||||
from opencompass.openicl.icl_retriever import ZeroRetriever
|
|
||||||
from opencompass.openicl.icl_inferencer import GenInferencer
|
|
||||||
from opencompass.utils.text_postprocessors import match_answer_pattern
|
|
||||||
from opencompass.evaluator import GenericLLMEvaluator
|
|
||||||
from opencompass.datasets import generic_llmjudge_postprocess
|
|
||||||
from opencompass.datasets import SciEvalDataset
|
|
||||||
|
|
||||||
with read_base():
|
|
||||||
from .SciEval_lifescience_sets import SciEval_lifescience_subsets
|
|
||||||
|
|
||||||
QUERY_TEMPLATE = """
|
|
||||||
Answer the following multiple choice question. The last line of your response should be of the following format: 'ANSWER: $LETTER' (without quotes) where LETTER is one of ABCD.
|
|
||||||
|
|
||||||
{input}
|
|
||||||
|
|
||||||
A) {A}
|
|
||||||
B) {B}
|
|
||||||
C) {C}
|
|
||||||
D) {D}
|
|
||||||
""".strip()
|
|
||||||
|
|
||||||
GRADER_TEMPLATE = """
|
|
||||||
Please as a grading expert, judge whether the final answers given by the candidates below are consistent with the standard answers, that is, whether the candidates answered correctly.
|
|
||||||
|
|
||||||
Here are some evaluation criteria:
|
|
||||||
1. Please refer to the given standard answer. You don't need to re-generate the answer to the question because the standard answer has been given. You only need to judge whether the candidate's answer is consistent with the standard answer according to the form of the question. Don't try to answer the original question. You can assume that the standard answer is definitely correct.
|
|
||||||
2. Because the candidate's answer may be different from the standard answer in the form of expression, before making a judgment, please understand the question and the standard answer first, and then judge whether the candidate's answer is correct, but be careful not to try to answer the original question.
|
|
||||||
3. Some answers may contain multiple items, such as multiple-choice questions, multiple-select questions, fill-in-the-blank questions, etc. As long as the answer is the same as the standard answer, it is enough. For multiple-select questions and multiple-blank fill-in-the-blank questions, the candidate needs to answer all the corresponding options or blanks correctly to be considered correct.
|
|
||||||
4. Some answers may be expressed in different ways, such as some answers may be a mathematical expression, some answers may be a textual description, as long as the meaning expressed is the same. And some formulas are expressed in different ways, but they are equivalent and correct.
|
|
||||||
|
|
||||||
Please judge whether the following answers are consistent with the standard answer based on the above criteria. Grade the predicted answer of this new question as one of:
|
|
||||||
A: CORRECT
|
|
||||||
B: INCORRECT
|
|
||||||
Just return the letters "A" or "B", with no text around it.
|
|
||||||
|
|
||||||
Here is your task. Simply reply with either CORRECT, INCORRECT. Don't apologize or correct yourself if there was a mistake; we are just trying to grade the answer.
|
|
||||||
|
|
||||||
<Original Question Begin>: {input}
|
|
||||||
A) {A}
|
|
||||||
B) {B}
|
|
||||||
C) {C}
|
|
||||||
D) {D}
|
|
||||||
<Original Question End>
|
|
||||||
|
|
||||||
<Gold Target Begin>:
|
|
||||||
{target}
|
|
||||||
<Gold Target End>
|
|
||||||
|
|
||||||
<Predicted Answer Begin>:
|
|
||||||
{prediction}
|
|
||||||
<Predicted End>
|
|
||||||
|
|
||||||
Judging the correctness of candidates' answers:
|
|
||||||
""".strip()
|
|
||||||
|
|
||||||
scieval_reader_cfg = dict(
|
|
||||||
input_columns=['input', 'A', 'B', 'C', 'D'],
|
|
||||||
output_column='target',
|
|
||||||
train_split='test',
|
|
||||||
)
|
|
||||||
|
|
||||||
scieval_datasets = []
|
|
||||||
for name in SciEval_lifescience_subsets:
|
|
||||||
scieval_infer_cfg = dict(
|
|
||||||
prompt_template=dict(
|
|
||||||
type=PromptTemplate,
|
|
||||||
template=dict(
|
|
||||||
round=[
|
|
||||||
dict(role='HUMAN', prompt=QUERY_TEMPLATE),
|
|
||||||
]
|
|
||||||
)
|
|
||||||
),
|
|
||||||
retriever=dict(type=ZeroRetriever),
|
|
||||||
inferencer=dict(type=GenInferencer),
|
|
||||||
)
|
|
||||||
|
|
||||||
scieval_eval_cfg = dict(
|
|
||||||
evaluator=dict(
|
|
||||||
type=GenericLLMEvaluator,
|
|
||||||
prompt_template=dict(
|
|
||||||
type=PromptTemplate,
|
|
||||||
template=dict(
|
|
||||||
begin=[
|
|
||||||
dict(
|
|
||||||
role='SYSTEM',
|
|
||||||
fallback_role='HUMAN',
|
|
||||||
prompt=(
|
|
||||||
'You are a helpful assistant who evaluates the correctness '
|
|
||||||
"and quality of models' outputs."
|
|
||||||
),
|
|
||||||
)
|
|
||||||
],
|
|
||||||
round=[
|
|
||||||
dict(role='HUMAN', prompt=GRADER_TEMPLATE),
|
|
||||||
],
|
|
||||||
),
|
|
||||||
),
|
|
||||||
dataset_cfg=dict(
|
|
||||||
type=SciEvalDataset,
|
|
||||||
path='OpenDFM/SciEval',
|
|
||||||
name='default',
|
|
||||||
reader_cfg=scieval_reader_cfg,
|
|
||||||
),
|
|
||||||
judge_cfg=dict(),
|
|
||||||
dict_postprocessor=dict(type=generic_llmjudge_postprocess),
|
|
||||||
),
|
|
||||||
pred_role='BOT',
|
|
||||||
)
|
|
||||||
|
|
||||||
scieval_datasets.append(
|
|
||||||
dict(
|
|
||||||
abbr=f'scieval_lifescience_{name}_llmjudge',
|
|
||||||
type=SciEvalDataset,
|
|
||||||
path='OpenDFM/SciEval',
|
|
||||||
name='default',
|
|
||||||
reader_cfg=scieval_reader_cfg,
|
|
||||||
infer_cfg=scieval_infer_cfg,
|
|
||||||
eval_cfg=scieval_eval_cfg,
|
|
||||||
mode='singlescore',
|
|
||||||
)
|
|
||||||
)
|
|
||||||
@ -1,3 +0,0 @@
|
|||||||
SciEval_lifescience_subsets = [
|
|
||||||
'biology', # 大学生物学
|
|
||||||
]
|
|
||||||
@ -1,62 +0,0 @@
|
|||||||
import re
|
|
||||||
from typing import List
|
|
||||||
|
|
||||||
from datasets import Dataset, DatasetDict, load_dataset
|
|
||||||
|
|
||||||
from opencompass.datasets.base import BaseDataset
|
|
||||||
from opencompass.registry import LOAD_DATASET
|
|
||||||
|
|
||||||
# 预编译的多选题正则，按 PEP-8 每行 < 79 字符
|
|
||||||
_PATTERN_MC = (
|
|
||||||
r'^(?P<stem>.*?)' # 题干
|
|
||||||
r'(?:A\.)\s*(?P<A>.*?)\s*' # 选项 A
|
|
||||||
r'B\.\s*(?P<B>.*?)\s*' # 选项 B
|
|
||||||
r'C\.\s*(?P<C>.*?)\s*' # 选项 C
|
|
||||||
r'D\.\s*(?P<D>.*?)' # 选项 D
|
|
||||||
r'Answer:' # 答案分隔符
|
|
||||||
)
|
|
||||||
|
|
||||||
|
|
||||||
@LOAD_DATASET.register_module()
|
|
||||||
class SciEvalDataset(BaseDataset):
|
|
||||||
"""Biology multiple-choice subset of SciEval."""
|
|
||||||
|
|
||||||
@staticmethod
|
|
||||||
def load(path: str, name: str, **kwargs) -> DatasetDict:
|
|
||||||
dataset = DatasetDict()
|
|
||||||
|
|
||||||
for split in ('test', ):
|
|
||||||
raw_iter = load_dataset(
|
|
||||||
path,
|
|
||||||
name=name,
|
|
||||||
split=split,
|
|
||||||
streaming=True,
|
|
||||||
)
|
|
||||||
|
|
||||||
examples: List[dict] = []
|
|
||||||
for ex in raw_iter:
|
|
||||||
if (ex.get('category') != 'biology'
|
|
||||||
or ex.get('type') != 'multiple-choice'):
|
|
||||||
continue
|
|
||||||
|
|
||||||
ans_list = ex.get('answer') or ex.get('answers') or []
|
|
||||||
if not ans_list:
|
|
||||||
continue
|
|
||||||
target = ans_list[0]
|
|
||||||
|
|
||||||
match = re.search(_PATTERN_MC, ex.get('question', ''), re.S)
|
|
||||||
if not match:
|
|
||||||
continue
|
|
||||||
|
|
||||||
examples.append({
|
|
||||||
'input': match.group('stem').strip(),
|
|
||||||
'A': match.group('A').strip(),
|
|
||||||
'B': match.group('B').strip(),
|
|
||||||
'C': match.group('C').strip(),
|
|
||||||
'D': match.group('D').strip(),
|
|
||||||
'target': target,
|
|
||||||
})
|
|
||||||
|
|
||||||
dataset[split] = Dataset.from_list(examples)
|
|
||||||
|
|
||||||
return dataset
|
|
||||||
@ -130,7 +130,6 @@ from .ruler import * # noqa: F401, F403
|
|||||||
from .safety import * # noqa: F401, F403
|
from .safety import * # noqa: F401, F403
|
||||||
from .scibench import ScibenchDataset, scibench_postprocess # noqa: F401, F403
|
from .scibench import ScibenchDataset, scibench_postprocess # noqa: F401, F403
|
||||||
from .scicode import * # noqa: F401, F403
|
from .scicode import * # noqa: F401, F403
|
||||||
from .SciEval_lifescience import SciEvalDataset # noqa: F401
|
|
||||||
from .simpleqa import * # noqa: F401, F403
|
from .simpleqa import * # noqa: F401, F403
|
||||||
from .siqa import * # noqa: F401, F403
|
from .siqa import * # noqa: F401, F403
|
||||||
from .smolinstruct import * # noqa: F401, F403
|
from .smolinstruct import * # noqa: F401, F403
|
||||||
|
|||||||
Loading…
Reference in New Issue
Block a user