--- a
+++ b/lstm_kmean/readme.md
@@ -0,0 +1,4 @@
+1. First, you have to download the thought viz dataset pickle file and its equivalent images (images are not required for feature extraction learning, just class names are required).
+2. Then run train.py code and it will run for around 3000 epochs.
+3. After train, now run the find_bestckpt.py code, it will give you the path of best checkpoint with better test acc.
+4. Finally you can run inference.py for generating tsne plot.