GT4SD / gt4sd-core

GT4SD, an open-source library to accelerate hypothesis generation in the scientific discovery process.
https://gt4sd.github.io/gt4sd-core/
MIT License
333 stars 69 forks source link

Is there a demo of "Multitask Text and Chemistry T5"? #223

Closed hkqiu closed 1 year ago

hkqiu commented 1 year ago

Is there a demo of "Multitask Text and Chemistry T5"? Including the demo of training data for fine-tuning, the demo of inference.

Best regards.

jannisborn commented 1 year ago

Hi @HKQiu, Thanks for your interest!

Hope this helps, let me know?

hkqiu commented 1 year ago

Thanks for help! Have a nice day!

zw-SIMM commented 1 year ago

I see, t5-small(finetuned) achieved 0.95 BLEU score in the paragraph2action task, can you provide more details?

jannisborn commented 1 year ago

The paragraph2actions task has been originally described in this paper: https://www.nature.com/articles/s41467-020-17266-6

zw-SIMM commented 1 year ago

The paragraph2actions task has been originally described in this paper: https://www.nature.com/articles/s41467-020-17266-6

Thanks, I knew the paper. I mean the training details of your work to achieve 0.95 BLEU. It's so fascinating!

jannisborn commented 1 year ago

Thanks :) The training process is detailed here: https://github.com/GT4SD/multitask_text_and_chemistry_t5/tree/main#model-training

Note that it only includes a small subset of the data we used, however, the formatting of the sample data is exactly as you need it