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
This comment indicates that considerable compute resources were needed to train NeuralGCM: #101 (comment)
This puts the value of neuralgcm out of reach for many groups who would like to try, fine-tune, or use the pretrained weights. It seems like their only route is to duplicate effort and train a new model from scratch rather than fine-tune or use the original pretrained weights.
It would be great if companies (big and small), nonprofits and partnerships between the two could use the pretrained model weights without worrying about the "noncommercial" part of the license.
The text was updated successfully, but these errors were encountered:
We are still figuring out our full licensing strategy around NeuralGCM, but certainly we are very eager to support non-profit and non-profit adjacent use-cases. Please do reach out directly to me at [email protected] with details and I can put you in touch with our partnerships team.
This comment indicates that considerable compute resources were needed to train NeuralGCM: #101 (comment)
This puts the value of neuralgcm out of reach for many groups who would like to try, fine-tune, or use the pretrained weights. It seems like their only route is to duplicate effort and train a new model from scratch rather than fine-tune or use the original pretrained weights.
It would be great if companies (big and small), nonprofits and partnerships between the two could use the pretrained model weights without worrying about the "noncommercial" part of the license.
The text was updated successfully, but these errors were encountered: