reran finetuning, saving best model, adjusted config

config.json

@@ -1,10 +1,11 @@
 {
-  "_name_or_path": "tweeteval/roberta-base-rt/",
+  "_name_or_path": "maxpe/twitter-roberta-base_semeval18_emodetection",
   "architectures": [
-    "RobertaForMaskedLM"
+    "RobertaForSequenceClassification"
   ],
   "attention_probs_dropout_prob": 0.1,
   "bos_token_id": 0,
+  "classifier_dropout": null,
   "eos_token_id": 2,
   "gradient_checkpointing": false,
   "hidden_act": "gelu",
@@ -19,7 +20,8 @@
   "num_hidden_layers": 12,
   "pad_token_id": 1,
   "position_embedding_type": "absolute",
-  "transformers_version": "4.4.2",
+  "problem_type": "multi_label_classification",
+  "transformers_version": "4.11.3",
   "type_vocab_size": 1,
   "use_cache": true,
   "vocab_size": 50265

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:10215a75f426a95b6ed3bfdff2a28e1d0d202810035f5068d917986e494f5030
+oid sha256:e7918c97f1d8b886aaf61fbe2720b67f5eed6daa7007b56103b02e1114e36be3
 size 498699629

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f7c82ea1dd1a6785c5a7b20c3e4d380c3dbe7918d87235b8da4925f55685b2c7
-size 2351
+oid sha256:8cba9ab0879b87b03585e92262820c27c75ccf19489645970d82fca00d46eee3
+size 2927