tensorflow / mesh

Mesh TensorFlow: Model Parallelism Made Easier
Apache License 2.0
1.59k stars 256 forks source link

Layers and Session Support #28

Open Quentin-Anthony opened 5 years ago

Quentin-Anthony commented 5 years ago
  1. I have an image classification model defined in Keras that I'm attempting to parallelize with MTF. However, it's not clear to me whether MTF support exists for keras.layers/tf.layers or if I'll need to recreate my model in MTF. Does MTF support keras.layers or tf.layers?

  2. Does MTF exclusively use sessions for training or is there support for TF 2.0 eager execution?

If the answer is "no" to either of the above questions, is there any plan to add support in the future?

kaczmarj commented 4 years ago

I have the same questions. Is it possible to construct tensorflow mesh models with the tf.keras API? I took a look at mesh_tensorflow/layers.py and the layers are not tf.keras.layers.Layer subclasses, suggesting the answer is no.