YAML Metadata Warning: The pipeline tag "text2text-generation" is not in the official list: text-classification, token-classification, table-question-answering, question-answering, zero-shot-classification, translation, summarization, feature-extraction, text-generation, fill-mask, sentence-similarity, text-to-speech, text-to-audio, automatic-speech-recognition, audio-to-audio, audio-classification, audio-text-to-text, voice-activity-detection, depth-estimation, image-classification, object-detection, image-segmentation, text-to-image, image-to-text, image-to-image, image-to-video, unconditional-image-generation, video-classification, reinforcement-learning, robotics, tabular-classification, tabular-regression, tabular-to-text, table-to-text, multiple-choice, text-ranking, text-retrieval, time-series-forecasting, text-to-video, image-text-to-text, visual-question-answering, document-question-answering, zero-shot-image-classification, graph-ml, mask-generation, zero-shot-object-detection, text-to-3d, image-to-3d, image-feature-extraction, video-text-to-text, keypoint-detection, visual-document-retrieval, any-to-any, video-to-video, other

Description

This is ai-forever/FRED-T5-1.7B model trained on Question-Answering, Question-Generation and Answer-Aware Question Generation tasks on russian dataset (hivaze/ru-AAQG-QA-QG)

Prompts

AAQG_PROMPT = "Сгенерируй вопрос по тексту, используя известный ответ. Текст: '{context}'. Ответ: '{answer}'."
QG_PROMPT = "Сгенерируй вопрос по тексту. Текст: '{context}'."
QA_PROMPT = "Сгенерируй ответ на вопрос по тексту. Текст: '{context}'. Вопрос: '{question}'."

Examples and code

from transformers import AutoTokenizer, T5ForConditionalGeneration
from functools import partial

saved_checkpoint = 'hivaze/AAQG-QA-QG-FRED-T5-1.7B'
tokenizer = AutoTokenizer.from_pretrained(saved_checkpoint)
model = T5ForConditionalGeneration.from_pretrained(saved_checkpoint).cuda()

def generate_text(prompt, tokenizer, model, n=1, temperature=0.8, num_beams=3):
  encoded_input = tokenizer.encode_plus(prompt, return_tensors='pt')
  encoded_input = {k: v.to(model.device) for k, v in encoded_input.items()}

  resulted_tokens = model.generate(**encoded_input,
                                   max_new_tokens=64,
                                   do_sample=True,
                                   num_beams=num_beams,
                                   num_return_sequences=n,
                                   temperature=temperature,
                                   top_p=0.9,
                                   top_k=50)
  resulted_texts = tokenizer.batch_decode(resulted_tokens, skip_special_tokens=True)

  return resulted_texts

generate_text = partial(generate_text, tokenizer=tokenizer, model=model)

test_context = "Путешественник Федор Конюхов и пилот Игорь Потапкин установили мировой рекорд высоты полета на паралёте, поднявшись на высоту 4728 метров — сайт Конюхова"

AAQG

generate_text(AAQG_PROMPT.format(
  context=test_context,
  answer='на паралёте'
), n=1)

"На чем путешественник Федор Конюхов и пилот Игорь Потапкин установили мировой рекорд высоты полета?"

generate_text(AAQG_PROMPT.format(
  context=test_context,
  answer='рекорд высоты полета'
), n=1)

"Что установили путешественник Федор Конюхов и пилот Игорь Потапкин?"

QA

generate_text(QA_PROMPT.format(
  context=test_context,
  question='Что установили путешественник Федор Конюхов и пилот Игорь Потапкин?'
), n=1)

"Мировой рекорд высоты полета на паралёте"

QG

generate_text(QG_PROMPT.format(context=test_context), n=1)

"Кто установил мировой рекорд высоты полета на паралёте?"

Metrics

Step	Training Loss	Validation Loss	Sbleu	Chr F	Rouge1	Rouge2	Rougel
500	1.020500	1.059296	41.556000	66.391100	0.104200	0.033700	0.104200
1000	1.050200	0.998357	43.035900	66.376800	0.105100	0.034100	0.105200
1500	0.994000	0.966051	43.692200	66.597600	0.106300	0.034400	0.106400
2000	0.947800	0.953637	44.012400	66.711100	0.106600	0.034900	0.106800
2500	0.978200	0.944621	44.027900	66.657400	0.106500	0.034600	0.106500

Authors

Sergei Bratchikov (https://t.me/nlpwanderer)

Downloads last month: 6

Safetensors

Model size

2B params

Tensor type

F32

hivaze
/

AAQG-QA-QG-FRED-T5-1.7B