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
{{ message }}
This repository has been archived by the owner on Sep 18, 2024. It is now read-only.
I work in an AI company, and we hope that NNI can not only save JSON representing the optimal model, but also the optimal model containing weights.
#2620
Open
xianglei3 opened this issue
Jun 30, 2020
· 2 comments
What would you like to be added:
I work in an AI company, and we hope that NNI can not only save JSON representing the optimal model, but also the optimal model containing weights. For example, when we use CIFAR10_pytorch. Py for classification, the optimal network structure can be obtained based on network_morphism.
But we also hope to save the best models and weights at the same time for other people who want to do classification tasks. In particular, I would like to ask if I can save the models and weights in PTH format for my friends who only install PyTorch but not nNI environment
Why is this needed:
Many businesses need not only models, but also weights
Without this feature, how does current nni work:
cannot do Components that may involve changes:
yes Brief description of your proposal if any:
yes
The text was updated successfully, but these errors were encountered:
Hello xianglei3, sorry for bothering. I'm an intern of nni group and I'm reaching out to understand better about your usage scenario. Would you like to leave an email so that we could have a discussion in more detail? Or you may want to join our wechat group (you could find the QR code at the main page).
What would you like to be added:
I work in an AI company, and we hope that NNI can not only save JSON representing the optimal model, but also the optimal model containing weights. For example, when we use CIFAR10_pytorch. Py for classification, the optimal network structure can be obtained based on network_morphism.
But we also hope to save the best models and weights at the same time for other people who want to do classification tasks. In particular, I would like to ask if I can save the models and weights in PTH format for my friends who only install PyTorch but not nNI environment
Why is this needed:
Many businesses need not only models, but also weights
Without this feature, how does current nni work:
cannot do
Components that may involve changes:
yes
Brief description of your proposal if any:
yes
The text was updated successfully, but these errors were encountered: