mesh icon indicating copy to clipboard operation
mesh copied to clipboard

Tensorflow Mesh needs documentation. Will this be provided anytime soon?

Open shyamalschandra opened this issue 3 years ago • 1 comments

I read the paper, Switch Transformers, as carefully as possible. However, none of these parameters were glossarized and well-defined in the code and paper. For example, you have the following uncommented lines with variables:

https://github.com/tensorflow/mesh/blob/5ce96838da567061515b71ca5b767e21bdca5768/mesh_tensorflow/transformer/moe.py#L41-L87

https://github.com/tensorflow/mesh/blob/5ce96838da567061515b71ca5b767e21bdca5768/mesh_tensorflow/transformer/moe.py#L133-L159

However, I see "some" documentation that is not inline with the code in the following section:

https://github.com/tensorflow/mesh/blob/5ce96838da567061515b71ca5b767e21bdca5768/mesh_tensorflow/transformer/moe.py#L504-L609

How can I understand the paper if the explanation for Switch Transformers in MoE is unclear and too abstract for most people unless they have access to the authors of the paper?

shyamalschandra avatar Jan 14 '21 21:01 shyamalschandra

Anyone out there with some credence? Any help could help?

shyamalschandra avatar Jan 20 '21 22:01 shyamalschandra