mirror of
https://github.com/open-compass/opencompass.git
synced 2025-05-30 16:03:24 +08:00
[Update] Add dataset configurations of no max_out_len (#1967)
* [Update] Add dataset configurations of no max_out_len * update test torch version * update test torch version * update test torch version * update test torch version
This commit is contained in:
Linchen Xiao
GitHub
parent
64128916d0
commit
aa05993922
6
.github/workflows/pr-stage-check.yml
vendored
6
.github/workflows/pr-stage-check.yml
vendored
@ -20,7 +20,7 @@ jobs:
|
||||
matrix:
|
||||
python-version: ['3.10']
|
||||
include:
|
||||
- torch: 2.0.0
|
||||
- torch: 2.5.1
|
||||
steps:
|
||||
- uses: actions/checkout@v3
|
||||
- name: Set up Python ${{ matrix.python-version }}
|
||||
@ -30,7 +30,7 @@ jobs:
|
||||
- name: Upgrade pip
|
||||
run: python -m pip install --upgrade pip
|
||||
- name: Install PyTorch
|
||||
run: pip install torch==${{matrix.torch}}+cpu -f https://download.pytorch.org/whl/cpu/torch_stable.html
|
||||
run: pip install torch==${{matrix.torch}} -f https://download.pytorch.org/whl/cpu/torch_stable.html
|
||||
- name: Install system dependencies
|
||||
run: |
|
||||
sudo sed -i '$ a deb http://th.archive.ubuntu.com/ubuntu jammy main' /etc/apt/sources.list
|
||||
@ -106,7 +106,7 @@ jobs:
|
||||
- name: Upgrade pip
|
||||
run: python -m pip install pip --upgrade
|
||||
- name: Install PyTorch
|
||||
run: pip install torch==2.0.0+${{matrix.platform}} -f https://download.pytorch.org/whl/${{matrix.platform}}/torch_stable.html
|
||||
run: pip install torch==2.5.1 -f https://download.pytorch.org/whl/cpu/torch_stable.html
|
||||
- name: Install opencompass dependencies
|
||||
run: |
|
||||
pip install -r requirements.txt
|
||||
|
||||
@ -0,0 +1,56 @@
|
||||
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.arc_prize_public_evaluation import ARCPrizeDataset, ARCPrizeEvaluator
|
||||
|
||||
|
||||
# The system_prompt defines the initial instructions for the model,
|
||||
# setting the context for solving ARC tasks.
|
||||
system_prompt = '''You are a puzzle solving wizard. You are given a puzzle from the abstraction and reasoning corpus developed by Francois Chollet.'''
|
||||
|
||||
# User message template is a template for creating user prompts. It includes placeholders for training data and test input data,
|
||||
# guiding the model to learn the rule and apply it to solve the given puzzle.
|
||||
user_message_template = '''Here are the example input and output pairs from which you should learn the underlying rule to later predict the output for the given test input:
|
||||
----------------------------------------
|
||||
{training_data}
|
||||
----------------------------------------
|
||||
Now, solve the following puzzle based on its input grid by applying the rules you have learned from the training data.:
|
||||
----------------------------------------
|
||||
[{{'input': {input_test_data}, 'output': [[]]}}]
|
||||
----------------------------------------
|
||||
What is the output grid? Only provide the output grid in the form as in the example input and output pairs. Do not provide any additional information:'''
|
||||
|
||||
|
||||
arc_prize_public_evaluation_reader_cfg = dict(
|
||||
input_columns=['training_data', 'input_test_data'],
|
||||
output_column='output_test_data'
|
||||
)
|
||||
|
||||
arc_prize_public_evaluation_infer_cfg = dict(
|
||||
prompt_template=dict(
|
||||
type=PromptTemplate,
|
||||
template=dict(
|
||||
round=[
|
||||
dict(role='SYSTEM',fallback_role='HUMAN', prompt=system_prompt),
|
||||
dict(role='HUMAN', prompt=user_message_template),
|
||||
],
|
||||
)
|
||||
),
|
||||
retriever=dict(type=ZeroRetriever),
|
||||
inferencer=dict(type=GenInferencer)
|
||||
)
|
||||
|
||||
arc_prize_public_evaluation_eval_cfg = dict(
|
||||
evaluator=dict(type=ARCPrizeEvaluator)
|
||||
)
|
||||
|
||||
arc_prize_public_evaluation_datasets = [
|
||||
dict(
|
||||
abbr='ARC_Prize_Public_Evaluation',
|
||||
type=ARCPrizeDataset,
|
||||
path='opencompass/arc_prize_public_evaluation',
|
||||
reader_cfg=arc_prize_public_evaluation_reader_cfg,
|
||||
infer_cfg=arc_prize_public_evaluation_infer_cfg,
|
||||
eval_cfg=arc_prize_public_evaluation_eval_cfg
|
||||
)
|
||||
]
|
||||
@ -0,0 +1,45 @@
|
||||
import os
|
||||
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 GaokaoBenchDataset
|
||||
from mmengine.config import read_base
|
||||
|
||||
with read_base():
|
||||
from .GaokaoBench_prompts import MCQ_prompts, FBQ_prompts
|
||||
|
||||
GaokaoBench_datasets = []
|
||||
for folder, prompts in [
|
||||
('Multiple-choice_Questions', MCQ_prompts),
|
||||
('Fill-in-the-blank_Questions', FBQ_prompts),
|
||||
]:
|
||||
for p in prompts:
|
||||
reader_cfg = {
|
||||
'input_columns': ['question'],
|
||||
'output_column': 'answer',
|
||||
}
|
||||
infer_cfg = {
|
||||
'ice_template': {
|
||||
'type': PromptTemplate,
|
||||
'template': {'round': [{'role': 'HUMAN', 'prompt': p['prefix_prompt'] + '{question}'}]},
|
||||
'ice_token': '</E>',
|
||||
},
|
||||
'retriever': {'type': ZeroRetriever},
|
||||
'inferencer': {'type': GenInferencer},
|
||||
}
|
||||
eval_cfg = {
|
||||
'evaluator': {'type': 'GaokaoBenchEvaluator' + '_' + p['type']},
|
||||
'pred_role': 'BOT',
|
||||
}
|
||||
_base_path = 'opencompass/GAOKAO-BENCH'
|
||||
dataset = {
|
||||
'type': GaokaoBenchDataset,
|
||||
'abbr': 'GaokaoBench_' + p['keyword'],
|
||||
'path': _base_path,
|
||||
'filename': '/' + folder + '/' + p['keyword'] + '.json',
|
||||
'name': p['keyword'],
|
||||
'reader_cfg': reader_cfg,
|
||||
'infer_cfg': infer_cfg,
|
||||
'eval_cfg': eval_cfg,
|
||||
}
|
||||
GaokaoBench_datasets.append(dataset)
|
||||
@ -0,0 +1,81 @@
|
||||
from mmengine.config import read_base
|
||||
from copy import deepcopy
|
||||
from opencompass.openicl.icl_prompt_template import PromptTemplate
|
||||
from opencompass.openicl.icl_retriever import ZeroRetriever
|
||||
from opencompass.openicl.icl_inferencer import GenInferencer, PPLInferencer
|
||||
from opencompass.openicl.icl_evaluator import CircularEvaluator, AccEvaluator
|
||||
from opencompass.datasets import MathBenchDataset, math_postprocess_v2
|
||||
from opencompass.utils.text_postprocessors import first_option_postprocess
|
||||
|
||||
with read_base():
|
||||
from .mathbench_prompt import zero_shot_prompts, few_shot_prompts, mathbench_sets
|
||||
|
||||
# Max for this dataset is 4
|
||||
num_shot = 0
|
||||
# Generate reasoning path or not, only for single choice
|
||||
with_reasoning = True
|
||||
# Use circular evaluation or not
|
||||
with_circular_eval = True
|
||||
# Use PPL mode in single choice test or not
|
||||
use_ppl_single_choice = False
|
||||
|
||||
assert 0 <= num_shot <= 4
|
||||
if num_shot == 0:
|
||||
prompts = zero_shot_prompts
|
||||
else:
|
||||
prompts = {name: p[- 2 * num_shot - 2:] for name, p in few_shot_prompts.items()}
|
||||
|
||||
mathbench_datasets = []
|
||||
for _split in mathbench_sets:
|
||||
for _name in mathbench_sets[_split]:
|
||||
if 'single_choice' in _name:
|
||||
if with_reasoning:
|
||||
template_round = prompts[_name + '_with_reasoning']
|
||||
else:
|
||||
template_round = prompts[_name]
|
||||
else:
|
||||
template_round = prompts[_name]
|
||||
|
||||
if 'single_choice' in _name:
|
||||
pred_postprocessor = dict(type=first_option_postprocess, options='ABCD')
|
||||
else:
|
||||
pred_postprocessor = dict(type=math_postprocess_v2)
|
||||
|
||||
if 'single_choice' in _name and with_circular_eval:
|
||||
evaluator = dict(type=CircularEvaluator)
|
||||
else:
|
||||
evaluator = dict(type=AccEvaluator)
|
||||
|
||||
# assemble the final config
|
||||
mathbench_reader_cfg = dict(input_columns=['question'], output_column='answer')
|
||||
if use_ppl_single_choice and 'single_choice' in _name and not with_reasoning:
|
||||
template = {}
|
||||
for answer in ['A', 'B', 'C', 'D']:
|
||||
one_template_round = deepcopy(template_round)
|
||||
one_template_round['round'][-1]['prompt'] = one_template_round['round'][-1]['prompt'].format(answer=answer)
|
||||
template[answer] = dict(round=one_template_round)
|
||||
mathbench_infer_cfg = dict(
|
||||
prompt_template=dict(type=PromptTemplate, template=template),
|
||||
retriever=dict(type=ZeroRetriever),
|
||||
inferencer=dict(type=PPLInferencer),
|
||||
)
|
||||
else:
|
||||
mathbench_infer_cfg = dict(
|
||||
prompt_template=dict(type=PromptTemplate, template=dict(round=template_round)),
|
||||
retriever=dict(type=ZeroRetriever),
|
||||
inferencer=dict(type=GenInferencer),
|
||||
)
|
||||
mathbench_eval_cfg = dict(evaluator=evaluator, pred_postprocessor=pred_postprocessor)
|
||||
|
||||
mathbench_datasets.append(
|
||||
dict(
|
||||
abbr='mathbench-' + _split + '-' + _name,
|
||||
type=MathBenchDataset,
|
||||
path=f'data/mathbench_v1/{_split}',
|
||||
name=_name,
|
||||
with_circular=with_circular_eval,
|
||||
reader_cfg=mathbench_reader_cfg,
|
||||
infer_cfg=mathbench_infer_cfg,
|
||||
eval_cfg=mathbench_eval_cfg,
|
||||
)
|
||||
)
|
||||
189
opencompass/configs/datasets/bbh/bbh_llmjudge_gen_b5bdf1.py
Normal file
189
opencompass/configs/datasets/bbh/bbh_llmjudge_gen_b5bdf1.py
Normal file
@ -0,0 +1,189 @@
|
||||
# flake8: noqa
|
||||
|
||||
import os
|
||||
from opencompass.openicl.icl_prompt_template import PromptTemplate
|
||||
from opencompass.openicl.icl_retriever import ZeroRetriever
|
||||
from opencompass.openicl.icl_inferencer import GenInferencer
|
||||
from opencompass.evaluator import GenericLLMEvaluator
|
||||
from opencompass.datasets import BBHDataset
|
||||
from opencompass.datasets.generic import generic_llmjudge_academic_postprocess
|
||||
|
||||
|
||||
bbh_reader_cfg = dict(input_columns=['input'], output_column='target')
|
||||
|
||||
bbh_multiple_choice_sets = [
|
||||
'temporal_sequences',
|
||||
'disambiguation_qa',
|
||||
'date_understanding',
|
||||
'tracking_shuffled_objects_three_objects',
|
||||
'penguins_in_a_table',
|
||||
'geometric_shapes',
|
||||
'snarks',
|
||||
'ruin_names',
|
||||
'tracking_shuffled_objects_seven_objects',
|
||||
'tracking_shuffled_objects_five_objects',
|
||||
'logical_deduction_three_objects',
|
||||
'hyperbaton',
|
||||
'logical_deduction_five_objects',
|
||||
'logical_deduction_seven_objects',
|
||||
'movie_recommendation',
|
||||
'salient_translation_error_detection',
|
||||
'reasoning_about_colored_objects',
|
||||
]
|
||||
bbh_free_form_sets = [
|
||||
'multistep_arithmetic_two',
|
||||
'navigate',
|
||||
'dyck_languages',
|
||||
'word_sorting',
|
||||
'sports_understanding',
|
||||
'boolean_expressions',
|
||||
'object_counting',
|
||||
'formal_fallacies',
|
||||
'causal_judgement',
|
||||
'web_of_lies',
|
||||
]
|
||||
|
||||
|
||||
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.
|
||||
5. If the prediction is given with \\boxed{}, please ignore the \\boxed{} 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()
|
||||
|
||||
|
||||
bbh_sets = bbh_multiple_choice_sets + bbh_free_form_sets
|
||||
|
||||
# For zero shot inference in bbh
|
||||
bbh_datasets = []
|
||||
for _name in bbh_sets:
|
||||
bbh_infer_cfg = dict(
|
||||
prompt_template=dict(
|
||||
type=PromptTemplate,
|
||||
template=dict(round=[
|
||||
dict(
|
||||
role='HUMAN',
|
||||
prompt=f"Question: {{input}}\n You must give your final answer by starting with 'So the answer is' "
|
||||
)
|
||||
])),
|
||||
retriever=dict(type=ZeroRetriever),
|
||||
inferencer=dict(type=GenInferencer))
|
||||
|
||||
bbh_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=BBHDataset,
|
||||
name=_name,
|
||||
path='opencompass/bbh',
|
||||
reader_cfg=bbh_reader_cfg,
|
||||
),
|
||||
judge_cfg=dict(),
|
||||
dict_postprocessor=dict(type=generic_llmjudge_academic_postprocess, metric_name='score'),
|
||||
),
|
||||
pred_role='BOT',
|
||||
)
|
||||
|
||||
bbh_datasets.append(
|
||||
dict(
|
||||
type=BBHDataset,
|
||||
path='opencompass/bbh',
|
||||
name=_name,
|
||||
abbr='bbh-' + _name,
|
||||
reader_cfg=bbh_reader_cfg,
|
||||
infer_cfg=bbh_infer_cfg.copy(),
|
||||
eval_cfg=bbh_eval_cfg.copy())
|
||||
)
|
||||
|
||||
|
||||
# For original 3 shot inference in bbh
|
||||
bbh_3_shot_datasets = []
|
||||
for _name in bbh_sets:
|
||||
with open(os.path.join(os.path.dirname(__file__), 'lib_prompt', f'{_name}.txt'), 'r') as f:
|
||||
_hint = f.read()
|
||||
bbh_infer_cfg = dict(
|
||||
prompt_template=dict(
|
||||
type=PromptTemplate,
|
||||
template=dict(round=[
|
||||
dict(
|
||||
role='HUMAN',
|
||||
prompt=f"Follow the given examples and answer the question.\n{_hint}\n\nQ: {{input}}\nA: Let's think step by step."
|
||||
)
|
||||
])),
|
||||
retriever=dict(type=ZeroRetriever),
|
||||
inferencer=dict(type=GenInferencer))
|
||||
|
||||
bbh_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=BBHDataset,
|
||||
name=_name,
|
||||
path='opencompass/bbh',
|
||||
reader_cfg=bbh_reader_cfg,
|
||||
),
|
||||
judge_cfg=dict(),
|
||||
dict_postprocessor=dict(type=generic_llmjudge_academic_postprocess, metric_name='score'),
|
||||
),
|
||||
pred_role='BOT',
|
||||
)
|
||||
|
||||
bbh_3_shot_datasets.append(
|
||||
dict(
|
||||
type=BBHDataset,
|
||||
path='opencompass/bbh',
|
||||
name=_name,
|
||||
abbr='bbh-' + _name,
|
||||
reader_cfg=bbh_reader_cfg,
|
||||
infer_cfg=bbh_infer_cfg.copy(),
|
||||
eval_cfg=bbh_eval_cfg.copy()))
|
||||
@ -0,0 +1,45 @@
|
||||
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 (BigCodeBenchDataset, BigCodeBenchEvaluator)
|
||||
|
||||
bigcodebench_hard_reader_cfg = dict(
|
||||
input_columns=['complete_prompt'],
|
||||
output_column='test',
|
||||
)
|
||||
|
||||
bigcodebench_hard_infer_cfg = dict(prompt_template=dict(
|
||||
type=PromptTemplate,
|
||||
template=dict(
|
||||
begin=[dict(role='system', fallback_role='HUMAN', prompt='')],
|
||||
round=[
|
||||
dict(role='HUMAN', prompt='{complete_prompt}'),
|
||||
])),
|
||||
retriever=dict(type=ZeroRetriever),
|
||||
inferencer=dict(type=GenInferencer))
|
||||
|
||||
bigcodebench_hard_eval_cfg = dict(
|
||||
evaluator=dict(
|
||||
type=BigCodeBenchEvaluator,
|
||||
release_version='v0.1.2',
|
||||
eval_type='complete',
|
||||
# remote_execute_api='https://bigcode-bigcodebench-evaluator.hf.space/',
|
||||
remote_execute_api=
|
||||
'https://opencompass-opencompass-bigcodebench-evaluator.hf.space', # noqa: E501
|
||||
dataset_version='hard',
|
||||
),
|
||||
pred_role='BOT',
|
||||
)
|
||||
|
||||
bigcodebench_hard_complete_datasets = [
|
||||
dict(
|
||||
abbr='bigcodebench_hard_complete',
|
||||
type=BigCodeBenchDataset,
|
||||
path='opencompass/bigcodebench',
|
||||
reader_cfg=bigcodebench_hard_reader_cfg,
|
||||
infer_cfg=bigcodebench_hard_infer_cfg,
|
||||
eval_cfg=bigcodebench_hard_eval_cfg,
|
||||
release_version='v0.1.2',
|
||||
dataset_version='hard',
|
||||
)
|
||||
]
|
||||
@ -0,0 +1,45 @@
|
||||
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 (BigCodeBenchDataset, BigCodeBenchEvaluator)
|
||||
|
||||
bigcodebench_hard_reader_cfg = dict(
|
||||
input_columns=['instruct_prompt'],
|
||||
output_column='test',
|
||||
)
|
||||
|
||||
bigcodebench_hard_infer_cfg = dict(prompt_template=dict(
|
||||
type=PromptTemplate,
|
||||
template=dict(
|
||||
begin=[dict(role='system', fallback_role='HUMAN', prompt='')],
|
||||
round=[
|
||||
dict(role='HUMAN', prompt='{instruct_prompt}'),
|
||||
])),
|
||||
retriever=dict(type=ZeroRetriever),
|
||||
inferencer=dict(type=GenInferencer))
|
||||
|
||||
bigcodebench_hard_eval_cfg = dict(
|
||||
evaluator=dict(
|
||||
type=BigCodeBenchEvaluator,
|
||||
release_version='v0.1.2',
|
||||
eval_type='instruct',
|
||||
# remote_execute_api='https://bigcode-bigcodebench-evaluator.hf.space/',
|
||||
remote_execute_api=
|
||||
'https://opencompass-opencompass-bigcodebench-evaluator.hf.space', # noqa: E501
|
||||
dataset_version='hard',
|
||||
),
|
||||
pred_role='BOT',
|
||||
)
|
||||
|
||||
bigcodebench_hard_instruct_datasets = [
|
||||
dict(
|
||||
abbr='bigcodebench_hard_instruct',
|
||||
type=BigCodeBenchDataset,
|
||||
path='opencompass/bigcodebench',
|
||||
reader_cfg=bigcodebench_hard_reader_cfg,
|
||||
infer_cfg=bigcodebench_hard_infer_cfg,
|
||||
eval_cfg=bigcodebench_hard_eval_cfg,
|
||||
release_version='v0.1.2',
|
||||
dataset_version='hard',
|
||||
)
|
||||
]
|
||||
39
opencompass/configs/datasets/cmo_fib/cmo_fib_gen_2783e5.py
Normal file
39
opencompass/configs/datasets/cmo_fib/cmo_fib_gen_2783e5.py
Normal file
@ -0,0 +1,39 @@
|
||||
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 CMOFibDataset, MATHEvaluator, math_postprocess_v2
|
||||
|
||||
|
||||
cmo_fib_reader_cfg = dict(
|
||||
input_columns=['question'],
|
||||
output_column='answer'
|
||||
)
|
||||
|
||||
|
||||
cmo_fib_infer_cfg = dict(
|
||||
prompt_template=dict(
|
||||
type=PromptTemplate,
|
||||
template=dict(
|
||||
round=[
|
||||
dict(role='HUMAN', prompt='{question}\n请一步一步地推理，并将最终答案写入\\boxed{}.'),
|
||||
],
|
||||
)
|
||||
),
|
||||
retriever=dict(type=ZeroRetriever),
|
||||
inferencer=dict(type=GenInferencer)
|
||||
)
|
||||
|
||||
cmo_fib_eval_cfg = dict(
|
||||
evaluator=dict(type=MATHEvaluator, version='v2'), pred_postprocessor=dict(type=math_postprocess_v2)
|
||||
)
|
||||
|
||||
cmo_fib_datasets = [
|
||||
dict(
|
||||
abbr='cmo_fib',
|
||||
type=CMOFibDataset,
|
||||
path='opencompass/cmo_fib',
|
||||
reader_cfg=cmo_fib_reader_cfg,
|
||||
infer_cfg=cmo_fib_infer_cfg,
|
||||
eval_cfg=cmo_fib_eval_cfg
|
||||
)
|
||||
]
|
||||
@ -0,0 +1,37 @@
|
||||
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 GSM8KDataset, gsm8k_postprocess, gsm8k_dataset_postprocess, Gsm8kEvaluator
|
||||
from opencompass.datasets import MATHEvaluator, math_postprocess_v2
|
||||
|
||||
gsm8k_reader_cfg = dict(input_columns=['question'], output_column='answer')
|
||||
|
||||
gsm8k_infer_cfg = dict(
|
||||
prompt_template=dict(
|
||||
type=PromptTemplate,
|
||||
template=dict(
|
||||
round=[
|
||||
dict(role='HUMAN', prompt='{question}\nPlease reason step by step, and put your final answer within \\boxed{}.'),
|
||||
],
|
||||
),
|
||||
),
|
||||
retriever=dict(type=ZeroRetriever),
|
||||
inferencer=dict(type=GenInferencer),
|
||||
)
|
||||
|
||||
gsm8k_eval_cfg = dict(
|
||||
evaluator=dict(type=MATHEvaluator, version='v2'),
|
||||
pred_postprocessor=dict(type=math_postprocess_v2),
|
||||
dataset_postprocessor=dict(type=gsm8k_dataset_postprocess),
|
||||
)
|
||||
|
||||
gsm8k_datasets = [
|
||||
dict(
|
||||
abbr='gsm8k',
|
||||
type=GSM8KDataset,
|
||||
path='opencompass/gsm8k',
|
||||
reader_cfg=gsm8k_reader_cfg,
|
||||
infer_cfg=gsm8k_infer_cfg,
|
||||
eval_cfg=gsm8k_eval_cfg,
|
||||
)
|
||||
]
|
||||
@ -0,0 +1,117 @@
|
||||
from opencompass.datasets.korbench.korbench import korbenchDataset, korbenchEvaluator
|
||||
from opencompass.openicl.icl_inferencer import GenInferencer
|
||||
from opencompass.openicl.icl_prompt_template import PromptTemplate
|
||||
from opencompass.openicl.icl_retriever import ZeroRetriever
|
||||
from opencompass.evaluator import GenericLLMEvaluator
|
||||
from opencompass.datasets import generic_llmjudge_postprocess
|
||||
|
||||
categories = ['cipher', 'counterfactual', 'logic', 'operation', 'puzzle']
|
||||
|
||||
|
||||
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.
|
||||
5. If the prediction is given with \\boxed{}, please ignore the \\boxed{} 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{prompt}\n<Original Question End>\n\n
|
||||
<Gold Target Begin>: \n{answer}\n<Gold Target End>\n\n
|
||||
<Predicted Answer Begin>: \n{prediction}\n<Predicted End>\n\n
|
||||
|
||||
Judging the correctness of candidates' answers:
|
||||
""".strip()
|
||||
|
||||
korbench_0shot_single_datasets = []
|
||||
|
||||
for category in categories:
|
||||
# Prompt template
|
||||
prompt_template = dict(
|
||||
type=PromptTemplate,
|
||||
template=dict(
|
||||
begin=[
|
||||
dict(
|
||||
role='HUMAN',
|
||||
prompt=''
|
||||
)
|
||||
],
|
||||
round=[
|
||||
dict(
|
||||
role='HUMAN',
|
||||
prompt='{prompt}' # f-string
|
||||
)
|
||||
]
|
||||
)
|
||||
)
|
||||
|
||||
# Reader configuration
|
||||
reader_cfg = dict(
|
||||
input_columns=['prompt'],
|
||||
output_column='answer',
|
||||
)
|
||||
|
||||
# Inference configuration
|
||||
infer_cfg = dict(
|
||||
prompt_template=prompt_template,
|
||||
retriever=dict(type=ZeroRetriever),
|
||||
inferencer=dict(type=GenInferencer, max_out_len=1024),
|
||||
)
|
||||
|
||||
# Evaluation configuration
|
||||
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=korbenchDataset,
|
||||
path='opencompass/korbench',
|
||||
prompt_mode='0_shot',
|
||||
category=category,
|
||||
reader_cfg=reader_cfg,
|
||||
),
|
||||
judge_cfg=dict(),
|
||||
dict_postprocessor=dict(type=generic_llmjudge_postprocess),
|
||||
),
|
||||
pred_role='BOT',
|
||||
)
|
||||
|
||||
# Dataset
|
||||
korbench_dataset = dict(
|
||||
type=korbenchDataset,
|
||||
abbr=f'korbench_{category}',
|
||||
path='opencompass/korbench',
|
||||
prompt_mode='0_shot',
|
||||
category=category,
|
||||
reader_cfg=reader_cfg,
|
||||
infer_cfg=infer_cfg,
|
||||
eval_cfg=eval_cfg,
|
||||
mode='singlescore',
|
||||
)
|
||||
|
||||
korbench_0shot_single_datasets.append(korbench_dataset)
|
||||
@ -0,0 +1,96 @@
|
||||
from opencompass.openicl.icl_prompt_template import PromptTemplate
|
||||
from opencompass.openicl.icl_retriever import ZeroRetriever
|
||||
from opencompass.openicl.icl_inferencer import GenInferencer
|
||||
from opencompass.evaluator import GenericLLMEvaluator
|
||||
from opencompass.datasets import generic_llmjudge_postprocess
|
||||
from opencompass.datasets import MATHDataset
|
||||
|
||||
|
||||
# ----------------------------- Detailed Config -----------------------------
|
||||
|
||||
math_reader_cfg = dict(input_columns=['problem'], output_column='solution')
|
||||
|
||||
math_infer_cfg = dict(
|
||||
prompt_template=dict(
|
||||
type=PromptTemplate,
|
||||
template=dict(
|
||||
round=[
|
||||
dict(role='HUMAN', prompt='{problem}\nRemember to put your final answer within \\boxed{}.'),
|
||||
]
|
||||
),
|
||||
),
|
||||
retriever=dict(type=ZeroRetriever),
|
||||
inferencer=dict(type=GenInferencer),
|
||||
)
|
||||
|
||||
|
||||
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.
|
||||
5. If the prediction is given with \\boxed{}, please ignore the \\boxed{} 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{problem}\n<Original Question End>\n\n
|
||||
<Gold Target Begin>: \n{solution}\n<Gold Target End>\n\n
|
||||
<Predicted Answer Begin>: \n{prediction}\n<Predicted End>\n\n
|
||||
|
||||
Judging the correctness of candidates' answers:
|
||||
""".strip()
|
||||
|
||||
# Evaluation configuration
|
||||
math_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=MATHDataset,
|
||||
path='opencompass/math',
|
||||
file_name = 'test_prm800k_500.json',
|
||||
reader_cfg=math_reader_cfg,
|
||||
),
|
||||
judge_cfg=dict(),
|
||||
dict_postprocessor=dict(type=generic_llmjudge_postprocess),
|
||||
),
|
||||
pred_role='BOT',
|
||||
)
|
||||
|
||||
|
||||
math_datasets = [
|
||||
dict(
|
||||
type=MATHDataset,
|
||||
abbr='math_prm800k_500-llmjudge',
|
||||
path='opencompass/math',
|
||||
file_name = 'test_prm800k_500.json',
|
||||
reader_cfg=math_reader_cfg,
|
||||
infer_cfg=math_infer_cfg,
|
||||
eval_cfg=math_eval_cfg,
|
||||
mode='singlescore',
|
||||
)
|
||||
]
|
||||
@ -20,7 +20,7 @@ for k in [1]:
|
||||
)
|
||||
),
|
||||
retriever=dict(type=ZeroRetriever),
|
||||
inferencer=dict(type=GenInferencer, max_out_len=50)
|
||||
inferencer=dict(type=GenInferencer)
|
||||
)
|
||||
else:
|
||||
nq_infer_cfg = dict(
|
||||
|
||||
29
opencompass/configs/datasets/scicode/scicode_gen_62c139.py
Normal file
29
opencompass/configs/datasets/scicode/scicode_gen_62c139.py
Normal file
@ -0,0 +1,29 @@
|
||||
from opencompass.openicl.icl_prompt_template import PromptTemplate
|
||||
from opencompass.openicl.icl_retriever import ZeroRetriever
|
||||
from opencompass.openicl.icl_inferencer import ChatInferencer
|
||||
from opencompass.datasets import SciCodeDataset, SciCodeEvaluator
|
||||
|
||||
|
||||
SciCode_reader_cfg = dict(input_columns=['prompt'], output_column=None)
|
||||
|
||||
SciCode_infer_cfg = dict(
|
||||
ice_template=dict(
|
||||
type=PromptTemplate,
|
||||
template='',
|
||||
),
|
||||
|
||||
retriever=dict(type=ZeroRetriever),
|
||||
inferencer=dict(type=ChatInferencer, infer_mode='every'))
|
||||
|
||||
SciCode_eval_cfg = dict(evaluator=dict(type=SciCodeEvaluator, dataset_path='./data/scicode', with_bg=False))
|
||||
|
||||
SciCode_datasets = [
|
||||
dict(
|
||||
abbr='SciCode',
|
||||
type=SciCodeDataset,
|
||||
path='./data/scicode',
|
||||
with_bg=False,
|
||||
reader_cfg=SciCode_reader_cfg,
|
||||
infer_cfg=SciCode_infer_cfg,
|
||||
eval_cfg=SciCode_eval_cfg)
|
||||
]
|
||||
@ -0,0 +1,62 @@
|
||||
from opencompass.openicl.icl_prompt_template import PromptTemplate
|
||||
from opencompass.openicl.icl_retriever import ZeroRetriever, FixKRetriever
|
||||
from opencompass.openicl.icl_inferencer import GenInferencer
|
||||
from opencompass.datasets import TriviaQADatasetV2, TriviaQAEvaluator
|
||||
|
||||
|
||||
triviaqa_datasets = []
|
||||
for k in [1]:
|
||||
triviaqa_reader_cfg = dict(
|
||||
input_columns=['question'], output_column='answer', train_split='train', test_split='validation')
|
||||
|
||||
if k == 0:
|
||||
triviaqa_infer_cfg = dict(
|
||||
prompt_template=dict(
|
||||
type=PromptTemplate,
|
||||
template=dict(
|
||||
round=[
|
||||
dict(role='HUMAN', prompt='Q: {question}'),
|
||||
dict(role='BOT', prompt='A:'),
|
||||
]
|
||||
)
|
||||
),
|
||||
retriever=dict(type=ZeroRetriever),
|
||||
inferencer=dict(type=GenInferencer)
|
||||
)
|
||||
else:
|
||||
triviaqa_infer_cfg = dict(
|
||||
ice_template=dict(
|
||||
type=PromptTemplate,
|
||||
template=dict(
|
||||
round=[
|
||||
dict(role='HUMAN', prompt='Q: {question}'),
|
||||
dict(role='BOT', prompt='A: {answer}.\n'),
|
||||
]
|
||||
),
|
||||
),
|
||||
prompt_template=dict(
|
||||
type=PromptTemplate,
|
||||
template=dict(
|
||||
begin='</E>',
|
||||
round=[
|
||||
dict(role='HUMAN', prompt='Q: {question}'),
|
||||
dict(role='BOT', prompt='A:'),
|
||||
]
|
||||
),
|
||||
ice_token='</E>',
|
||||
),
|
||||
retriever=dict(type=FixKRetriever, fix_id_list=list(range(k))),
|
||||
inferencer=dict(type=GenInferencer),
|
||||
)
|
||||
|
||||
triviaqa_eval_cfg = dict(evaluator=dict(type=TriviaQAEvaluator), pred_role='BOT')
|
||||
|
||||
triviaqa_datasets.append(
|
||||
dict(
|
||||
type=TriviaQADatasetV2,
|
||||
abbr=f'triviaqa_wiki_{k}shot',
|
||||
path='opencompass/trivia_qa',
|
||||
reader_cfg=triviaqa_reader_cfg,
|
||||
infer_cfg=triviaqa_infer_cfg,
|
||||
eval_cfg=triviaqa_eval_cfg)
|
||||
)
|
||||
Loading…
Reference in New Issue
Block a user