- train.js contains the training logic
- train.js uses data.js to load and format the training data
- sequence_utils.js is created to house the truncation/padding logic
for both the tfjs-node training and in-browser inference
- README.md is updated
* WIP
* Add mnist-transfer-cnn and update build files
* Update build scripts
* bugfixes
* yarn.lock updates
* Remove http-server in favor of yarn watch; fix licenses
* Serve local pretrained model with http-server
* Let user choose local vs. remote pretrained model.
Shanqing Cai