update refined model

Using data augmentation, made the dataset more diverse. Although the training was on a much smaller dataset (Because of time limit), it resulted in significantly better classification outcome

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:b92d924aab8fca21a66040bfc0e562e1d75ce1b4cdd7adde1a3ed7c4832dbfbd
+oid sha256:f7c5665fe05193a23276b6effa6b1adb6366fbf0920ce678014484f1f093d669
 size 46757253