You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
hey there,
firstly compliment to you for your great work so far. it was really intuitive getting along with the package, and it also works great loading a model from other programming language and using it. But working with it we discovered a problem. It appeared we got other results from this package in comparison to the results we got from the FT package in python while using the same trained model. We think the problem could be that the current version of this package isn't supporting the one vs all (ova) loss.
So here comes my issue, could you implement the ova loss while working with already trained models? We would really appreciate it!
Thanks in advance and stay healthy ✌️
The text was updated successfully, but these errors were encountered:
Can confirm. Loading an already trained model (supervised; ova-loss) works fine, and the inference does not throw an error message. However, the predictions do not align with the Python or CLI versions.
hey there,
firstly compliment to you for your great work so far. it was really intuitive getting along with the package, and it also works great loading a model from other programming language and using it. But working with it we discovered a problem. It appeared we got other results from this package in comparison to the results we got from the FT package in python while using the same trained model. We think the problem could be that the current version of this package isn't supporting the one vs all (ova) loss.
So here comes my issue, could you implement the ova loss while working with already trained models? We would really appreciate it!
Thanks in advance and stay healthy ✌️
The text was updated successfully, but these errors were encountered: