Short model versions

This model was targeted MLM pre-trained for a few epochs after short former for 50 epochs. No long form training.

config.json

@@ -1,5 +1,5 @@
 {
-  "_name_or_path": "deberta-small/long-third-stage/checkpoint-22200/",
+  "_name_or_path": "deberta-small/shortformer-second-stage/",
   "architectures": [
     "DebertaV2ForMaskedLM"
   ],

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5c84141d71020be9ed515f2fe98230075c803c0cb863889599d13cfcdf845325
+oid sha256:575090b01c7a88217f7affe5f4de486504264ddb218cc25c47c9bc84a4644c26
 size 225288954