RichardErkhov/p208p2002_-_gpt2-large-babi-4bits

Quantization made by Richard Erkhov.

Github

Discord

Request more models

gpt2-large-babi - bnb 4bits

• Model creator: https://huggingface.co/p208p2002/
• Original model: https://huggingface.co/p208p2002/gpt2-large-babi/

Original model description:

datasets: - facebook/babi_qa

Fine tune and evaluate transformer model on facebook's bAbi tasks.

Towards AI-Complete Question Answering: A Set of Prerequisite Toy Tasks

Training Code: p208p2002/bAbi-tasks-with-transformer-model

task_no	task_name	score
qa1	single-supporting-fact	100
qa2	two-supporting-facts	99.4
qa3	three-supporting-facts	62.0
qa4	two-arg-relations	100
qa5	three-arg-relations	96.5
qa6	yes-no-questions	100
qa7	counting	100
qa8	lists-sets	99.8
qa9	simple-negation	100
qa10	indefinite-knowledge	100
qa11	basic-coreference	100
qa12	conjunction	100
qa13	compound-coreference	100
qa14	time-reasoning	100
qa15	basic-deduction	100
qa16	basic-induction	100
qa17	positional-reasoning	100
qa18	size-reasoning	100
qa19	path-finding	100
qa20	agents-motivations	100

# Please use with the follow template

INPUT_TEMPLATE = """
Context:
{context}

Question:
{question}

Answer:
{answer}
"""

input_text = INPUT_TEMPLATE.format_map({
    "context":context,
    "question":question,
    "answer":answer
}).strip()