add an evaluation function

1, add an evaluation function pick_the_best_out_of_multi_trained_model

1, add modelchecker for save the best model, 2, turn the baseline to .85

1, add early stop with torlerence 2, add regularizer 3, add optimizer adam, nadam and adamax

add tensorboard and early stop callback, while early stop doesnt work well, need more work here

add graph plot

1, move local config from the notebook to a local config python file 2, minor errors fixed

implement mutli-input with classification output

Merge https://github.com/DeviceChipsetAI/OneShot_SpeechCommand

multi-input for one shot

fix reference bug out_mfcc=base_mfcc

Merge https://github.com/DeviceChipsetAI/OneShot_SpeechCommand

add background noise as the basic mfcc

Add files via upload

add evaluate function on group. it can be a criteria for choosing right model