revert try 2

config.json

@@ -17,7 +17,7 @@
   "decoder_layerdrop": 0.0,
   "decoder_layers": 24,
   "decoder_start_token_id": 50258,
-  "dropout": 0.1,
+  "dropout": 0.03,
   "encoder_attention_heads": 16,
   "encoder_ffn_dim": 4096,
   "encoder_layerdrop": 0.0,

runs/Dec14_13-12-20_129-146-102-49/events.out.tfevents.1671023549.129-146-102-49.1608432.0

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:fde9aa65dd8258230f64cd1c0412fbd6e4555a173245c38182a88c17f2048f19
-size 17527
+oid sha256:57f9f639b674a5142f21cc272b0cbc03eaed6e6c81ce10dd31f0e8692d1354f2
+size 20196

runs/Dec14_17-13-31_129-146-102-49/1671038020.244651/events.out.tfevents.1671038020.129-146-102-49.1608432.3

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f4fb44bb37b9410e913d843fa3c79f7151010590810dbe5ccfc0312d744c8516
+size 5916

runs/Dec14_17-13-31_129-146-102-49/events.out.tfevents.1671038020.129-146-102-49.1608432.2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:864c3d4a93b53695bac1ca639630f5ea08fb61f76111ef76c92334e2f0d1194d
+size 4136

runs/Dec14_17-34-13_129-146-102-49/1671039262.5354552/events.out.tfevents.1671039262.129-146-102-49.1708481.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:763f6defe00c5912e2e8e0ce97e5b256cf408655f5be12c26d586d24f445980e
+size 5916

runs/Dec14_17-34-13_129-146-102-49/events.out.tfevents.1671039262.129-146-102-49.1708481.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:81c01da1392f5c746aff6b2bfec81fb06caf30f2426707f2e49925b861cad984
+size 4347