Several Issues about Training on datasets

[Dean-UQ]

Hi,
I notice that the ReadMe.md file didn't make it clear in the following aspects.
For example,

1. How to train the MPT and its training file is missing.
2. The dataset you provided is incomplete and other actions in H3.6M/CMU-MoCap datasets are missing.
3. The test files are confusing. Such as, What is the difference between WalkDog.npy and WalkDog_1.npy and which one should be used in testing.
   Overall, I think the instruction in ReadMe is too sketchy to help its readers.