diff --git a/configs/datasets/GLUE_CoLA/GULE_CoLA_ppl.py b/configs/datasets/GLUE_CoLA/GULE_CoLA_ppl.py new file mode 100644 index 000000000..1043c285f --- /dev/null +++ b/configs/datasets/GLUE_CoLA/GULE_CoLA_ppl.py @@ -0,0 +1,4 @@ +from mmengine.config import read_base + +with read_base(): + from .GULE_CoLA_ppl_77d0df import CoLA_datasets # noqa: F401, F403 diff --git a/configs/datasets/GLUE_CoLA/GULE_CoLA_ppl_77d0df.py b/configs/datasets/GLUE_CoLA/GULE_CoLA_ppl_77d0df.py new file mode 100644 index 000000000..48bed05e7 --- /dev/null +++ b/configs/datasets/GLUE_CoLA/GULE_CoLA_ppl_77d0df.py @@ -0,0 +1,50 @@ +from opencompass.openicl.icl_prompt_template import PromptTemplate +from opencompass.openicl.icl_retriever import FixKRetriever +from opencompass.openicl.icl_inferencer import PPLInferencer +from opencompass.openicl.icl_evaluator import AccEvaluator +from opencompass.datasets import HFDataset + + +_hint = "The following are text classification questions. \n" \ + "Please determine whether the following sentence is linguistically acceptable: " \ + "0 means unacceptable, 1 means acceptable.\n" + +CoLA_infer_cfg = dict( + ice_template=dict( + type=PromptTemplate, + template="Sentence: {sentence}\nResult: {label}", + ), + prompt_template=dict( + type=PromptTemplate, + template={ + answer: + f"{_hint}Sentence: {{sentence}}\nResult: {answer}" + for answer in [0, 1] + }, + ice_token='', + ), + retriever=dict(type=FixKRetriever), + inferencer=dict(type=PPLInferencer, fix_id_list=[17, 18, 19, 20, 21])) + +CoLA_eval_cfg = dict(evaluator=dict(type=AccEvaluator), ) + +CoLA_datasets = [] +for _split in ["validation"]: + + CoLA_reader_cfg = dict( + input_columns=['sentence'], + output_column='label', + test_split=_split + ) + + CoLA_datasets.append( + dict( + abbr=f'CoLA-{_split}', + type=HFDataset, + path='glue', + name='cola', + reader_cfg=CoLA_reader_cfg, + infer_cfg=CoLA_infer_cfg, + eval_cfg=CoLA_eval_cfg + ) + )