search

justanhduc / neuralnet-pytorch

A high level framework for general purpose neural networks in Pytorch.
https://neuralnet-pytorch.readthedocs.io
MIT License
27 stars 6 forks source link

How to modify original pytorch code into this? #1

Open xiaomingjie opened 4 years ago

xiaomingjie commented 4 years ago

Hi @justanhduc , I am very interested in your GraphX. It seems that your nnt mainly rewrite nn.sequential which need to add output_shape in each torch operation such as FC and Conv2d. So can original pytorch code be easily transform to nnt type?

justanhduc commented 4 years ago

thanks for your interest. it's very easy. you can use this to wrap a pure Pytorch module. i will release a new version of nnt soon (and GraphX as well) which supports better wrapping.