search

asyml / texar-pytorch

Integrating the Best of TF into PyTorch, for Machine Learning, Natural Language Processing, and Text Generation. This is part of the CASL project: http://casl-project.ai/
https://asyml.io
Apache License 2.0
745 stars 117 forks source link

Introduce NNI with distributed Adaptive API #331

Closed ZeyaWang closed 3 years ago

ZeyaWang commented 3 years ago

This PR adds an example to Texar-Pytorch about how to run NNI experiments with distributed Adaptive API with the help of AdaptDL.

examples/bert/bert_classifier_adaptive_nni.py is th the nni version of examples/bert/bert_classifier_adaptive.py. config_bert_classifier.yml is the configuration file to run the experiment.

codecov[bot] commented 3 years ago

Codecov Report

Merging #331 (26d93c4) into master (a061b45) will increase coverage by 0.02%. The diff coverage is n/a.

Impacted file tree graph

@@            Coverage Diff             @@
##           master     #331      +/-   ##
==========================================
+ Coverage   80.14%   80.17%   +0.02%     
==========================================
  Files         135      136       +1     
  Lines       11220    11242      +22     
==========================================
+ Hits         8992     9013      +21     
- Misses       2228     2229       +1
Impacted Files Coverage Δ
texar/torch/losses/info_loss.py 95.45% <0.00%> (ø)

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update a061b45...26d93c4. Read the comment docs.

ZeyaWang commented 3 years ago

We can wait and merge this only when all dependencies are open-sourced.

The docker build command here requires a petuum registry, any way for a non-petuum user to reproduce this?

In the updated ReadMe file, we further provide guidance about how to build and push a docker image to their own private registry and a link about how to create Secrete and use Secrete to pull image from the private registry in K8S. With this guide, the user is able to reproduce all the steps about the docker images.