deep learning - Define custom model/architecture in TensorFlow -
from little have played around tensorflow see has already-implemented architectures rnn/lstm cells, convnets, etc. there way define one's "custom" architecture (e.g. "enhanced" lstm network few convolutional layers)?
yes, totally possible. output of lstm or network tensors cab used input of network.
see how combine them @ https://github.com/jazzsaxmafia/show_and_tell.tensorflow.
you can find more examples @ https://github.com/tensorflowkr/awesome_tensorflow_implementations.
Comments
Post a Comment