Prepare for release

README.md

@@ -14,9 +14,10 @@ We have made several updates compared to the previous version:
 
 - The model is now implemented using MXNet.
 - A new graph generative model with molecule level recurrency is added to the repo. See the article for further detail.
-- The pre-trained models are now available in `ckpt.tar.gz`, along with the predictive model for GSK-3b and JNK3.
-- Samples generated by unconditional model are now available in `samples.tar.gz`
-- All datasets are now packed in `datasets.tar.gz`
+- The pre-trained models are now available in `ckpt.tar.gz` (download here), along with the predictive model for GSK-3b and JNK3.
+- Samples generated by unconditional model are now available in `samples.tar.gz` (download here)
+- All datasets are now packed in `datasets.tar.gz` (download here)
+- Large files (`ckpt.tar.gz`, `samples.tar.gz`  and `datasets.tar.gz` ) are now placed in the assets in the release.
 - We have provided a tutorial (`examples.ipynb`) to demonstrate the usage of our model
 
 ## Requirements:
@@ -51,7 +52,7 @@ To ease to installation process, a docker environment will be added to the repo
 
 ### Usage
 
-To train the model, first unpack`datasets.tar.gz` to the current directory, and call:
+To train the model, first unpack`datasets.tar.gz` (download here) to the current directory, and call:
 ```shell
 ./train.py {molmp|molrnn|scaffold|prop|kinase} path/to/output
 ```