search

otori-bird / retrosynthesis

MIT License
55 stars 13 forks source link

Better results on Arxiv version #9

Closed mikolajsacha closed 1 year ago

mikolajsacha commented 1 year ago

Hi, I noticed that the version of your publication on Arxiv (https://arxiv.org/abs/2203.11444) has better Top K results than the one published in Chemical Science and documented in README in this repository. Where does this difference come from? Did you improve something in the model or the training procedure?

Mikołaj

otori-bird commented 1 year ago

I appreciate your attention, but considering the passage of time since this work was conducted, I have unfortunately forgotten the precise reasons for the improvements. I am sure that the specific model architecture is the same. In case you plan to reference the results of this article, I would advise relying on the officially published one.

mikolajsacha commented 1 year ago

I understand. Thank you for your response.