[GitHub] [incubator-tvm] antinucleon commented on issue #5377: [Blocksparse] Pipeline for lowering dense model to sparse-dense

[GitBox <gi...@apache.org>]

antinucleon commented on issue #5377:
URL: https://github.com/apache/incubator-tvm/pull/5377#issuecomment-616310534


   @vinx13 If directly convert a model from TensorFlow, like BERT, all dense layers are in the format of ```y = nn.desnse(x, transpose(w))```. To further convert it to a sparse dense model, we have to simplify it to ```y = nn.dense(x, wt)``` first, then transform to sparse model with sparse weight.


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org