jiangzhs/OpenCompass
1
0
Fork 0
mirror of https://github.com/open-compass/opencompass.git synced 2025-05-30 16:03:24 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

[Feature] Add Datasets: ClimateQA,Physics (#2017)

Browse Source 
* feat ClimateQA

* feat PHYSICS

* fix

* fix

* fix

* fix
This commit is contained in:
Myhs_phz 2025-04-14 20:18:47 +08:00 committed by GitHub
parent 6a6a1a5c0b
commit 75e7834b59
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
12 changed files with 588 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
dataset-index.yml
View File

@ -997,3 +997,17 @@
paper: https://arxiv.org/pdf/2502.14739
configpath: opencompass/configs/datasets/supergpqa
configpath_llmjudge: ''
- climaqa:
name: ClimaQA
category: Science
paper: https://arxiv.org/pdf/2410.16701
configpath: ''
configpath_llmjudge:
- opencompass/configs/datasets/ClimaQA/ClimaQA_Gold_llm_judge.py
- opencompass/configs/datasets/ClimaQA/ClimaQA_Silver_llm_judge.py
- physics:
name: PHYSICS
category: Science
paper: https://arxiv.org/pdf/2503.21821
configpath: ''
configpath_llmjudge: opencompass/configs/datasets/PHYSICS/PHYSICS_llm_judge_gen_a133a2.py

4
opencompass/configs/datasets/ClimaQA/ClimaQA_Gold_llm_judge_gen.py Normal file
View File

@ -0,0 +1,4 @@
from mmengine.config import read_base
with read_base():
from .ClimaQA_Gold_llm_judge_gen_f15343 import climaqa_datasets # noqa: F401, F403

164
opencompass/configs/datasets/ClimaQA/ClimaQA_Gold_llm_judge_gen_f15343.py Normal file
View File

@ -0,0 +1,164 @@
from opencompass.openicl.icl_prompt_template import PromptTemplate
from opencompass.openicl.icl_retriever import ZeroRetriever
from opencompass.openicl.icl_inferencer import GenInferencer
from opencompass.datasets import ClimaQADataset, generic_llmjudge_postprocess
from opencompass.evaluator import GenericLLMEvaluator
climaqa_gold_sets = [
'mcq',
'cloze',
'ffq'
]
GRADER_TEMPLATE_mcq = """
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. The answer may be one of the four options: a, b, c, or d. Only when the options given by prediction are strictly consistent with the answer, the prediction can be considered correct.
3. If the prediction is given with 'The answer is:', please ignore the 'The answer is:', and only judge whether the candidate's answer is consistent with the standard answer.
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>: \n{input}\n<Original Question End>\n\n
<Gold Target Begin>: \n{target}\n<Gold Target End>\n\n
<Predicted Answer Begin>: \n{prediction}\n<Predicted End>\n\n
Judging the correctness of candidates' answers:
""".strip()
GRADER_TEMPLATE_cloze = """
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. The form of the answer is a word or a phrase. Please strictly compare the prediction and the answer. Only when the prediction and the answer are exactly the same, will the prediction be considered correct; otherwise, it will be considered incorrect.
3. If the prediction is given with 'The answer is:', please ignore the 'The answer is:' and only judge whether the candidate's answer is consistent with the standard answer.
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>: \n{input}\n<Original Question End>\n\n
<Gold Target Begin>: \n{target}\n<Gold Target End>\n\n
<Predicted Answer Begin>: \n{prediction}\n<Predicted End>\n\n
Judging the correctness of candidates' answers:
""".strip()
GRADER_TEMPLATE_ffq = """
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.
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. The type of question is open-ended Q&A. Please compare whether the prediction is close enough to the meaning of the answer and whether the prediction covers each key point in the answer. If the prediction meets the above requirements, it can be considered very close to the answer.
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.
5. If the prediction is given with 'The answer is:', please ignore the 'The answer is:' and only judge whether the candidate's answer is very close to the standard answer.
Please judge whether the following answers are close to the standard answer based on the above criteria. Grade the predicted answer of this new question as one of:
A: very close to the answer
B: not very close to the answer
Just return the letters "A" or "B", with no text around it.
Here is your task. Simply reply with either A or B. Don't apologize or correct yourself if there was a mistake; we are just trying to grade the answer.
<Original Question Begin>: \n{input}\n<Original Question End>\n\n
<Gold Target Begin>: \n{target}\n<Gold Target End>\n\n
<Predicted Answer Begin>: \n{prediction}\n<Predicted End>\n\n
Judging the correctness of candidates' answers:
""".strip()
climaqa_reader_cfg = dict(input_columns=['input'], output_column='target')
climaqa_datasets = []
for _task in climaqa_gold_sets:
if _task == 'mcq':
GRADER_TEMPLATE = GRADER_TEMPLATE_mcq
infer_prompt = f"Think step by step, and when you provide the final answer, please use the prefix \"The answer is:\"without any modification. The question is multiple choice with a single correct answer, the final answer must only be the letter corresponding to the correct answer. For example, \"The answer is: a\"\n\nQ: {{input}}\nA: "
if _task == 'ffq':
GRADER_TEMPLATE = GRADER_TEMPLATE_ffq
infer_prompt = f"Think step by step, and when you provide the final answer, please use the prefix \"The answer is:\".\n\nQ: {{input}}\nA: "
if _task == 'cloze':
GRADER_TEMPLATE = GRADER_TEMPLATE_cloze
infer_prompt = f"Fill the <Mask> in the sentence. Think step by step, and when you provide the final answer, please use the prefix \"The answer is:\"without any modification, and provide the answer directly, with no formatting, no bolding, and no markup. For instance: \"The answer is: 42\" or \"The answer is: yes\".\n\nQ: {{input}}\nA: "
climaqa_infer_cfg = dict(
prompt_template=dict(
type=PromptTemplate,
template=dict(
round=[
dict(
role='HUMAN',
prompt=infer_prompt,
)
]
),
),
retriever=dict(type=ZeroRetriever),
inferencer=dict(type=GenInferencer),
)
climaqa_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=ClimaQADataset,
path='opencompass/ClimaQA-Gold',
task=_task,
abbr='ClimaQA_Gold_' + _task,
reader_cfg=climaqa_reader_cfg,
),
judge_cfg=dict(),
dict_postprocessor=dict(type=generic_llmjudge_postprocess),
),
pred_role='BOT',
)
climaqa_datasets.append(
dict(
abbr='ClimaQA_Gold_' + _task,
type=ClimaQADataset,
path='opencompass/ClimaQA-Gold',
task=_task,
reader_cfg=climaqa_reader_cfg,
infer_cfg=climaqa_infer_cfg,
eval_cfg=climaqa_eval_cfg,
)
)

4
opencompass/configs/datasets/ClimaQA/ClimaQA_Silver_llm_judge_gen.py Normal file
View File

@ -0,0 +1,4 @@
from mmengine.config import read_base
with read_base():
from .ClimaQA_Silver_llm_judge_gen_f15343 import climaqa_datasets # noqa: F401, F403

160
opencompass/configs/datasets/ClimaQA/ClimaQA_Silver_llm_judge_gen_f15343.py Normal file
View File

@ -0,0 +1,160 @@
from opencompass.openicl.icl_prompt_template import PromptTemplate
from opencompass.openicl.icl_retriever import ZeroRetriever
from opencompass.openicl.icl_inferencer import GenInferencer
from opencompass.datasets import ClimaQADataset, generic_llmjudge_postprocess
from opencompass.evaluator import GenericLLMEvaluator
climaqa_silver_sets = [
'mcq',
'cloze',
'ffq'
]
GRADER_TEMPLATE_mcq = """
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. The answer may be one of the four options: a, b, c, or d. Only when the options given by prediction are strictly consistent with the answer, the prediction can be considered correct.
3. If the prediction is given with 'The answer is:', please ignore the 'The answer is:', and only judge whether the candidate's answer is consistent with the standard answer.
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>: \n{input}\n<Original Question End>\n\n
<Gold Target Begin>: \n{target}\n<Gold Target End>\n\n
<Predicted Answer Begin>: \n{prediction}\n<Predicted End>\n\n
Judging the correctness of candidates' answers:
""".strip()
GRADER_TEMPLATE_cloze = """
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. The form of the answer is a word or a phrase. Please strictly compare the prediction and the answer. Only when the prediction and the answer are exactly the same, will the prediction be considered correct; otherwise, it will be considered incorrect.
3. If the prediction is given with 'The answer is:', please ignore the 'The answer is:' and only judge whether the candidate's answer is consistent with the standard answer.
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>: \n{input}\n<Original Question End>\n\n
<Gold Target Begin>: \n{target}\n<Gold Target End>\n\n
<Predicted Answer Begin>: \n{prediction}\n<Predicted End>\n\n
Judging the correctness of candidates' answers:
""".strip()
GRADER_TEMPLATE_ffq = """
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.
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. The type of question is open-ended Q&A. Please compare whether the prediction is close enough to the meaning of the answer and whether the prediction covers each key point in the answer. If the prediction meets the above requirements, it can be considered very close to the answer.
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.
5. If the prediction is given with 'The answer is:', please ignore the 'The answer is:' and only judge whether the candidate's answer is very close to the standard answer.
Please judge whether the following answers are close to the standard answer based on the above criteria. Grade the predicted answer of this new question as one of:
A: very close to the answer
B: not very close to the answer
Just return the letters "A" or "B", with no text around it.
Here is your task. Simply reply with either A or B. Don't apologize or correct yourself if there was a mistake; we are just trying to grade the answer.
<Original Question Begin>: \n{input}\n<Original Question End>\n\n
<Gold Target Begin>: \n{target}\n<Gold Target End>\n\n
<Predicted Answer Begin>: \n{prediction}\n<Predicted End>\n\n
Judging the correctness of candidates' answers:
""".strip()
climaqa_reader_cfg = dict(input_columns=['input'], output_column='target')
climaqa_datasets = []
for _task in climaqa_silver_sets:
if _task == 'mcq':
GRADER_TEMPLATE = GRADER_TEMPLATE_mcq
infer_prompt = f"Think step by step, and when you provide the final answer, please use the prefix \"The answer is:\"without any modification. The question is multiple choice with a single correct answer, the final answer must only be the letter corresponding to the correct answer. For example, \"The answer is: a\"\n\nQ: {{input}}\nA: "
if _task == 'ffq':
GRADER_TEMPLATE = GRADER_TEMPLATE_ffq
infer_prompt = f"Think step by step, and when you provide the final answer, please use the prefix \"The answer is:\".\n\nQ: {{input}}\nA: "
if _task == 'cloze':
GRADER_TEMPLATE = GRADER_TEMPLATE_cloze
infer_prompt = f"Fill the <Mask> in the sentence. Think step by step, and when you provide the final answer, please use the prefix \"The answer is:\"without any modification, and provide the answer directly, with no formatting, no bolding, and no markup. For instance: \"The answer is: 42\" or \"The answer is: yes\".\n\nQ: {{input}}\nA: "
climaqa_infer_cfg = dict(
prompt_template=dict(
type=PromptTemplate,
template=dict(
round=[
dict(
role='HUMAN',
prompt=infer_prompt,
)
]
),
),
retriever=dict(type=ZeroRetriever),
inferencer=dict(type=GenInferencer),
)
climaqa_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=ClimaQADataset,
path='opencompass/ClimaQA-Silver',
task=_task,
abbr='ClimaQA_Silver_' + _task,
reader_cfg=climaqa_reader_cfg,
),
judge_cfg=dict(),
dict_postprocessor=dict(type=generic_llmjudge_postprocess),
),
pred_role='BOT',
)
climaqa_datasets.append(
dict(
abbr='ClimaQA_Silver_' + _task,
type=ClimaQADataset,
path='opencompass/ClimaQA-Silver',
task=_task,
reader_cfg=climaqa_reader_cfg,
infer_cfg=climaqa_infer_cfg,
eval_cfg=climaqa_eval_cfg,
)
)

4
opencompass/configs/datasets/PHYSICS/PHYSICS_llm_judge_gen.py Normal file
View File

@ -0,0 +1,4 @@
from mmengine.config import read_base
with read_base():
from .PHYSICS_llm_judge_gen_a133a2 import physics_datasets # noqa: F401, F403

131
opencompass/configs/datasets/PHYSICS/PHYSICS_llm_judge_gen_a133a2.py Normal file
View File

@ -0,0 +1,131 @@
from opencompass.openicl.icl_prompt_template import PromptTemplate
from opencompass.openicl.icl_retriever import ZeroRetriever
from opencompass.openicl.icl_inferencer import GenInferencer
from opencompass.datasets import (
PHYSICSDataset,
generic_llmjudge_postprocess,
)
from opencompass.evaluator import GenericLLMEvaluator
physics_sets = [
'atomic_dataset_textonly',
'electro_dataset_textonly',
'mechanics_dataset_textonly',
'optics_dataset_textonly',
'quantum_dataset_textonly',
'statistics_dataset_textonly',
]
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 questions may include multiple sub questions and sub answers. Each sub answer is given after a guide character in the form of <Answer 1:> or <Answer 2:>, etc. Please note that only when all sub predictions given in prediction correspond one-to-one with the answer and are all correct, will the prediction be considered correct; otherwise, it will be considered incorrect.
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.
5. The final answers in the prediction are generally given with \\boxed{}. If you cannot find sufficient \\boxed{} in the prediction, please try to find matching answers from other places within the prediction as much as possible.
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: All Sub Predictions Are Correct
B: Not Every Sub Predictions is Correct
Just return the letters "A" or "B", with no text around it.
Here is your task. Simply reply with either A, B. Don't apologize or correct yourself if there was a mistake; we are just trying to grade the answer.
<Original Question Begin>: \n{input}\n<Original Question End>\n\n
<Gold Target Begin>: \n{target}\n<Gold Target End>\n\n
<Predicted Answer Begin>: \n{prediction}\n<Predicted End>\n\n
Judging the correctness of candidates' answers:
""".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 questions may include multiple sub questions and sub answers. Each sub answer is given after a guide character in the form of <Answer 1:> or <Answer 2:>, etc. Please note that as long as at least one correct answer appears in the prediction, the prediction is 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.
# 5. The final answers in the prediction are generally given with \\boxed{}. If you cannot find sufficient \\boxed{} in the prediction, please try to find matching answers from other places within the prediction as much as possible.
#
# 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: At Least One Sub Prediction is Correct
# B: All Sub Predictions are Incorrect
# Just return the letters "A" or "B", with no text around it.
#
# Here is your task. Simply reply with either A, B. Don't apologize or correct yourself if there was a mistake; we are just trying to grade the answer.
#
# <Original Question Begin>: \n{input}\n<Original Question End>\n\n
# <Gold Target Begin>: \n{target}\n<Gold Target End>\n\n
# <Predicted Answer Begin>: \n{prediction}\n<Predicted End>\n\n
#
# Judging the correctness of candidates' answers:
# """.strip()
physics_reader_cfg = dict(input_columns=['input'], output_column='target')
physics_datasets = []
for _name in physics_sets:
physics_infer_cfg = dict(
prompt_template=dict(
type=PromptTemplate,
template=dict(
round=[
dict(
role='HUMAN',
prompt=f'Answer the given question step by step. Begin by explaining your reasoning process clearly. Conclude by providing the final answers at the end in LaTeX boxed format. Think step by step before answering. It should be noted that the question may include multiple sub questions, please ensure that each question is answered in order.\n\nQ: {{input}}\nA: ',
)
]
),
),
retriever=dict(type=ZeroRetriever),
inferencer=dict(type=GenInferencer),
)
physics_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=PHYSICSDataset,
path='opencompass/PHYSICS-textonly',
abbr='PHYSICS_' + _name,
name=_name,
reader_cfg=physics_reader_cfg,
),
judge_cfg=dict(),
dict_postprocessor=dict(type=generic_llmjudge_postprocess),
),
pred_role='BOT',
)
physics_datasets.append(
dict(
abbr='PHYSICS_' + _name,
type=PHYSICSDataset,
path='opencompass/PHYSICS-textonly',
name=_name,
reader_cfg=physics_reader_cfg,
infer_cfg=physics_infer_cfg,
eval_cfg=physics_eval_cfg,
)
)

14
opencompass/configs/summarizers/groups/PHYSICS.py Normal file
View File

@ -0,0 +1,14 @@
physics_summary_groups = []
# bbh
_physcis = [
'atomic_dataset_textonly',
'electro_dataset_textonly',
'mechanics_dataset_textonly',
'optics_dataset_textonly',
'quantum_dataset_textonly',
'statistics_dataset_textonly',
]
_physcis = ['PHYSICS_' + s for s in _physcis]
physics_summary_groups.append({'name': 'PHYSICS', 'subsets': _physcis})

2
opencompass/datasets/__init__.py
View File

@ -25,6 +25,7 @@ from .chinese_simpleqa import * # noqa: F401, F403
from .cibench import * # noqa: F401, F403
from .circular import * # noqa: F401, F403
from .civilcomments import * # noqa: F401, F403
from .climaqa import * # noqa: F401, F403
from .clozeTest_maxmin import * # noqa: F401, F403
from .cluewsc import * # noqa: F401, F403
from .cmb import * # noqa: F401, F403
@ -110,6 +111,7 @@ from .obqa import * # noqa: F401, F403
from .olymmath import * # noqa: F401, F403
from .OlympiadBench import * # noqa: F401, F403
from .OpenFinData import * # noqa: F401, F403
from .physics import * # noqa: F401, F403
from .piqa import * # noqa: F401, F403
from .py150 import * # noqa: F401, F403
from .qasper import * # noqa: F401, F403

30
opencompass/datasets/climaqa.py Normal file
View File

@ -0,0 +1,30 @@
import os
from datasets import load_dataset
from opencompass.datasets.base import BaseDataset
from opencompass.registry import LOAD_DATASET
from opencompass.utils import get_data_path
@LOAD_DATASET.register_module()
class ClimaQADataset(BaseDataset):
@staticmethod
def load(path: str, task: str, **kwargs):
path = get_data_path(path)
path = os.path.join(path, task)
climateqa = load_dataset(path)['train']
input_column = []
for i in range(len(climateqa)):
if 'Options' in climateqa[i].keys(
) and climateqa[i]['Options'] is not None:
input_column.append(climateqa[i]['Question'] + '\n' +
climateqa[i]['Options'])
else:
input_column.append(climateqa[i]['Question'])
climateqa = climateqa.add_column(name='input', column=input_column)
climateqa = climateqa.rename_column('Answer', 'target')
return climateqa

30
opencompass/datasets/physics.py Normal file
View File

@ -0,0 +1,30 @@
import os
from datasets import load_dataset
from opencompass.datasets.base import BaseDataset
from opencompass.registry import LOAD_DATASET
from opencompass.utils import get_data_path
@LOAD_DATASET.register_module()
class PHYSICSDataset(BaseDataset):
@staticmethod
def load(path: str, name: str, **kwargs):
path = get_data_path(path)
path = os.path.join(path, name)
physics = load_dataset(path)['train']
physics = physics.rename_column('questions', 'input')
target = []
for i in physics:
this_final_answer = ''
for j in range(len(i['final_answers'])):
this_final_answer += 'Answer ' + str(j + 1) + ': '
this_final_answer += i['final_answers'][j]
this_final_answer += '\n'
target.append(this_final_answer)
physics = physics.add_column(name='target', column=target)
return physics

31
opencompass/utils/datasets_info.py
View File

@ -420,9 +420,40 @@ DATASETS_MAPPING = {
"hf_id": "",
"local": "./data/OlympiadBench",
},
"opencompass/ClimaQA-Gold": {
"ms_id": "",
"hf_id": "",
"local": "./data/climaqa_gold",
},
"opencompass/ClimaQA-Silver": {
"ms_id": "",
"hf_id": "",
"local": "./data/climaqa_silver",
},
"opencompass/PHYSICS-textonly": {
"ms_id": "",
"hf_id": "",
"local": "./data/PHYSICS-textonly",
},
}
DATASETS_URL = {
"/climaqa_gold": {
"url":
"http://opencompass.oss-cn-shanghai.aliyuncs.com/datasets/data/climaqa_gold.zip",
"md5": "310cd0dc96db2bbbce798c40e2163ac2",
},
"/climaqa_silver": {
"url":
"http://opencompass.oss-cn-shanghai.aliyuncs.com/datasets/data/climaqa_silver.zip",
"md5": "acdd955f1c170539c5233c12f7227c58",
},
"/PHYSICS-textonly": {
"url":
"http://opencompass.oss-cn-shanghai.aliyuncs.com/datasets/data/PHYSICS-textonly.zip",
"md5": "92be6846a22dd4da942ca43f0638c709",
},
"/OlympiadBench": {
"url":
"http://opencompass.oss-cn-shanghai.aliyuncs.com/datasets/data/OlympiadBench.zip",