How to specify the number of layers for Transformer in OpenNMT-tf with Tensorflow 2?

Support
1

Hi all,

I’m trying to train 2 models with different number of layers, I can do this easilly when I was using OpenNMT-py with Tensorflow 1.14, now I have upgraded to Tensorflow 2 and OpenNMT-tf, the only command I found is something like this:
onmt-main --model_type Transformer --config data.yml --auto_config train
I found there’s no way to specify those parameters, not even in the data.yml file

Can anyone help me with this?
Thanks

2

Hi,

Non standard models should be defined in custom model definition files, see the documentation at https://opennmt.net/OpenNMT-tf/model.html#custom-models