Preparation for the next release about singa HOT 9 CLOSED

We may aim to include some internal features for developers. I wrote two discussions in the mail list, one for Code Quality and the other for Coverage. I am planning to write more discussions for other improvements soon.

The code quality improvements seems fine now. We have lgtm, linting and license checks. But the code coverage is not ready. There are many tasks to enable coverage as proposed in the discussion. I will try to work on some of them and everyone is welcome to participate.

About 2.1 vs 3.0, I vote for 2.1.

from singa.

Thanks for the suggestion, Moaz!
In terms of the version number, I think if we have major API changes that are not compatible with V2.0, then we need to release V3.0;

from singa.

@moazreyad pls help double check the release steps. This is our first release as TLP; hence it is a bit from past releases.

from singa.

1. At the end, we will need to publish the release.

2. The step of check license is not required. It will be done automatically when any pull request is merged with master.

from singa.

changelog of some of the mentioned points
https://gist.github.com/dcslin/1e8934ffc791fdea4b95810f437cf361

from singa.

changelog of some of the mentioned points
https://gist.github.com/joddiy/ff7a5f8e8dd35668a2f300c26ad01cdb

from singa.

@joddiy pls create the colab notebooks for the onnx examples

from singa.

@joddiy pls create the colab notebooks for the onnx examples

Got it, I am going to create it right now.

from singa.

@joddiy pls create the colab notebooks for the onnx examples

please check the colab examples. However, I just clean the code now, since the new version SINGA has not been released, it cannot run yet. After we release the new version, I will adjust it then.

Face Analysis

arcface, fer+ emotion

Object Detection

tiny yolo v2

Image Classification

mobilenet, vgg16, resnet18

Machine Comprehension

bert squad

Mnist demo for inference, re-training, and transfer learning

mnist

from singa.