search

d2l-ai / d2l-en

Interactive deep learning book with multi-framework code, math, and discussions. Adopted at 500 universities from 70 countries including Stanford, MIT, Harvard, and Cambridge.
https://D2L.ai
Other
23.09k stars 4.26k forks source link

Need to tune performance for MXNet & TensorFlow for seq2seq #2099

Open astonzhang opened 2 years ago

astonzhang commented 2 years ago

http://preview.d2l.ai.s3-website-us-west-2.amazonaws.com/d2l-en/master/chapter_recurrent-modern/seq2seq.html http://preview.d2l.ai.s3-website-us-west-2.amazonaws.com/d2l-en/master/chapter_attention-mechanisms/bahdanau-attention.html http://preview.d2l.ai.s3-website-us-west-2.amazonaws.com/d2l-en/master/chapter_attention-mechanisms/transformer.html

We need to tune performance for MXNet & TensorFlow to obtain similar performance of PyTorch for each section, such as learning rate & max_epochs.

AnirudhDagar commented 2 years ago

http://preview.d2l.ai.s3-website-us-west-2.amazonaws.com/d2l-en/master/chapter_attention-mechanisms/bahdanau-attention.html was fixed in #2104. seq2seq and transformers for TF still need fixing.