Multi-lingual Question Generating Model (mt5-small)

Give the model a passage and it will generate a question about the passage.

Trained on the following datasets:

Training details

I used flax summarization script and a TPU v3-8. Summarization expects a text column and a summary column. For question generation training, use the context column instead of text column and question instead of summary column.

Limitations and Intended Use

There is no guarantee that it will produce a question in the language of the passage, but it usually does. Lower resource languages will likely have lower quality questions.

Intended use is to make questions given a passage. With a larger model this might be able to generate training data for question-answering models, but this small one does not produce high-quality questions.

Using the model

PyTorch version

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
tokenizer = AutoTokenizer.from_pretrained("nbroad/mt5-small-qgen")
model = AutoModelForSeq2SeqLM.from_pretrained("nbroad/mt5-small-qgen", from_flax=True)

text = "Hugging Face has seen rapid growth in its \
popularity since the get-go. It is definitely doing\
 the right things to attract more and more people to \
 its platform, some of which are on the following lines:\
Community driven approach through large open source repositories \
along with paid services. Helps to build a network of like-minded\
 people passionate about open source. \
Attractive price point. The subscription-based features, e.g.: \
Inference based API, starts at a price of $9/month.\

inputs = tokenizer(text, return_tensors="pt")
output = model.generate(**inputs, max_length=40)

tokenizer.decode(output[0], skip_special_tokens=True)
# What is Hugging Face's price point?

Flax version

from transformers import AutoTokenizer, FlaxAutoModelForSeq2SeqLM
tokenizer = AutoTokenizer.from_pretrained("nbroad/mt5-small-qgen")
model = FlaxAutoModelForSeq2SeqLM.from_pretrained("nbroad/mt5-small-qgen")

text = "A un año y tres días de que el balón ruede \
en el Al Bayt Stadium inaugurando el Mundial 2022, \
ya se han dibujado los primeros bocetos de la próxima \
Copa del Mundo.13 selecciones están colocadas en el \
mapa con la etiqueta de clasificadas y tienen asegurado\
 pisar los verdes de Qatar en la primera fase final \
 otoñal. Serbia, Dinamarca, España, Países Bajos, \
 Suiza, Croacia, Francia, Inglaterra, Bélgica, Alemania,\
  Brasil, Argentina y Qatar, como anfitriona, entrarán en \
  el sorteo del 1 de abril de 2022 en Doha en el que 32 \
  países serán repartidos en sus respectivos grupos. \

inputs = tokenizer(text, return_tensors="pt")
output = model.generate(**inputs, max_length=40)

tokenizer.decode(output["sequences"][0], skip_special_tokens=True)
# ¿Cuántos países entrarán en el sorteo del Mundial 2022?

Model trained on Cloud TPUs from Google's TPU Research Cloud (TRC)

Downloads last month
Hosted inference API
Text2Text Generation
This model can be loaded on the Inference API on-demand.

Datasets used to train nbroad/mt5-small-qgen