Maziyar Panahi

I've noticed similar behavior between albert_base/1 and albert_base/3. While training NER with albert_base/1 is good (enough), albert_base/3 is terrible! The V1 starts from 86% and goes to 89% but the...

I believe due to its name `Generic Neural Elastic Search` GitHub has suggested `elasticsearch` as a tag and it was selected by mistake. I think it'll be removed soon :)

It would be great to have support for sbt as well. Currently, we do something like: ``` - name: Upload coverage data to Coveralls run: sbt ++${{ matrix.scala-version }} coverageAggregate...

If the linked PR bing merged which sets Dropout to 0.0 then the random embeddings will be resolved as the authors of the XLNet suggested.

Is this because of the way we compile the master or is there something in our notebooks that prevents the success? (Zeppelin 0.8.0) Update: I just tested the with master...

I had to create a dummy app so I can get it back! I don't think this UI change was supposed to be forced since it has a negative value...

I think you are on the right track by focusing on TF 2.0 where Keras is a first-class citizen. Also, TF 2.0 is much cleaner and much easier to use,...

I have observed a similar issue when it comes to context for word embeddings which can explain why it might behave the same on sentence level. In ELMO, BERT and...

Did anyone figure this out? I am still experiencing the same issue with no solution: https://github.com/zihangdai/xlnet/issues/264

Unfortunately, I couldn't find any solution. It seems for some reason (could be totally my mistake) the XLnet pre-trained models are not aware of their surrounding tokens. So no matter...