llmdataset 1.0.0

llm dataset

LLM dataset

How to use

• install
  • https://pypi.org/project/llmdataset/

pip install llmdataset

• setting

dataloader = llmdataset.dataloader(
    data_type='train',
    batch_size=4,
    seed=3655,
    max_data=5
    )

• execution

for i in dataloader:
    print('------')
    print(i)

Useful dataset

• GSM8K
  • https://huggingface.co/datasets/gsm8k

llmdataset = main.LLMdataset(dataset_name="gsm8k", subset='main')

• MultiArith
  • https://huggingface.co/datasets/ChilleD/MultiArith

llmdataset = main.LLMdataset(dataset_name="ChilleD/MultiArith")

• Big Bench
  • https://huggingface.co/datasets/bigbench

llmdataset = main.LLMdataset(dataset_name="tasksource/bigbench", subset = "abstract_narrative_understanding")

• Big Bench Hard
  • https://huggingface.co/datasets/lukaemon/bbh

llmdataset = main.LLMdataset(dataset_name="reasoning-machines/gsm-hard", subset="boolean_expressions")

• GSM Hard
  • https://huggingface.co/datasets/reasoning-machines/gsm-hard

llmdataset = main.LLMdataset(dataset_name="reasoning-machines/gsm-hard")

• wikitablequestions
  • https://huggingface.co/datasets/wikitablequestions?row=0

llmdataset = main.LLMdataset(dataset_name="wikitablequestions")

• StrategyQA
  • https://huggingface.co/datasets/ChilleD/StrategyQA?row=0

llmdataset = main.LLMdataset(dataset_name="ChilleD/StrategyQA")

• ARC
  • https://huggingface.co/datasets/allenai/ai2_arc

llmdataset = main.LLMdataset(dataset_name="allenai/ai2_arc", subset = "ARC-Challenge")

• AQuA-RAT
  • https://huggingface.co/datasets/aqua_rat/viewer/raw

llmdataset = main.LLMdataset(dataset_name="aqua_rat", subset = "raw")

• GPQA
  • https://huggingface.co/datasets/Idavidrein/gpqa

llmdataset = main.LLMdataset(dataset_name="Idavidrein/gpqa", subset = "gpqa_diamond")

• SVAP
  • https://huggingface.co/datasets/ChilleD/SVAMP?row=0

llmdataset = main.LLMdataset(dataset_name="ChilleD/SVAMP")

• CommonsenseQA
  • https://huggingface.co/datasets/tau/commonsense_qa

llmdataset = main.LLMdataset(dataset_name="tau/commonsense_qa")