|
datasets = [ |
|
[ |
|
dict( |
|
abbr='lukaemon_mmlu_college_biology', |
|
eval_cfg=dict( |
|
evaluator=dict( |
|
type='opencompass.openicl.icl_evaluator.AccEvaluator')), |
|
infer_cfg=dict( |
|
ice_template=dict( |
|
template=dict( |
|
A= |
|
'{input}\nA. {A}\nB. {B}\nC. {C}\nD. {D}\nAnswer: A\n', |
|
B= |
|
'{input}\nA. {A}\nB. {B}\nC. {C}\nD. {D}\nAnswer: B\n', |
|
C= |
|
'{input}\nA. {A}\nB. {B}\nC. {C}\nD. {D}\nAnswer: C\n', |
|
D='{input}\nA. {A}\nB. {B}\nC. {C}\nD. {D}\nAnswer: D\n' |
|
), |
|
type= |
|
'opencompass.openicl.icl_prompt_template.PromptTemplate'), |
|
inferencer=dict( |
|
type='opencompass.openicl.icl_inferencer.PPLInferencer'), |
|
prompt_template=dict( |
|
ice_token='</E>', |
|
template=dict( |
|
A= |
|
'The following are multiple choice questions (with answers) about college biology.\n\n</E>{input}\nA. {A}\nB. {B}\nC. {C}\nD. {D}\nAnswer: A', |
|
B= |
|
'The following are multiple choice questions (with answers) about college biology.\n\n</E>{input}\nA. {A}\nB. {B}\nC. {C}\nD. {D}\nAnswer: B', |
|
C= |
|
'The following are multiple choice questions (with answers) about college biology.\n\n</E>{input}\nA. {A}\nB. {B}\nC. {C}\nD. {D}\nAnswer: C', |
|
D='The following are multiple choice questions (with answers) about college biology.\n\n</E>{input}\nA. {A}\nB. {B}\nC. {C}\nD. {D}\nAnswer: D' |
|
), |
|
type= |
|
'opencompass.openicl.icl_prompt_template.PromptTemplate'), |
|
retriever=dict( |
|
fix_id_list=[ |
|
0, |
|
1, |
|
2, |
|
3, |
|
4, |
|
], |
|
type='opencompass.openicl.icl_retriever.FixKRetriever')), |
|
name='college_biology', |
|
path='./data/mmlu/', |
|
reader_cfg=dict( |
|
input_columns=[ |
|
'input', |
|
'A', |
|
'B', |
|
'C', |
|
'D', |
|
], |
|
output_column='target', |
|
train_split='dev'), |
|
type='opencompass.datasets.MMLUDataset'), |
|
], |
|
] |
|
eval = dict(runner=dict(task=dict())) |
|
models = [ |
|
dict( |
|
abbr='my_api', |
|
api_key='', |
|
batch_size=8, |
|
max_out_len=100, |
|
max_seq_len=2048, |
|
meta_template=dict(round=[ |
|
dict(api_role='HUMAN', role='HUMAN'), |
|
dict(api_role='BOT', generate=True, role='BOT'), |
|
]), |
|
path='my_api', |
|
run_cfg=dict(num_gpus=1, num_procs=1), |
|
type='opencompass.models.my_api.MyAPIModel', |
|
url='http://127.0.0.1:12345/testing'), |
|
] |
|
work_dir = './outputs/default/20240306_164404' |
|
|