roberta with impiria

config.json

@@ -7,24 +7,23 @@
   "bos_token_id": 0,
   "classifier_dropout": null,
   "eos_token_id": 2,
-  "gradient_checkpointing": false,
   "hidden_act": "gelu",
   "hidden_dropout_prob": 0.1,
-  "hidden_size": 768,
+  "hidden_size": 1024,
   "initializer_range": 0.02,
-  "intermediate_size": 3072,
-  "layer_norm_eps": 1e-05,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-12,
   "max_2d_position_embeddings": 1024,
-  "max_position_embeddings": 514,
+  "max_position_embeddings": 512,
   "model_type": "roberta",
-  "num_attention_heads": 12,
-  "num_hidden_layers": 12,
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "output_past": true,
   "pad_token_id": 1,
   "position_embedding_type": "absolute",
-  "tokenizer_class": "RobertaTokenizer",
   "torch_dtype": "float32",
   "transformers_version": "4.23.1",
   "type_vocab_size": 2,
   "use_cache": true,
-  "vocab_size": 50265
+  "vocab_size": 30522
 }

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1a3ef97341c62a1f5dc8cec331375b6a66d287060c363d98f1353b5b9b83d726
-size 511247909
+oid sha256:01159ba7af127218df697cc6da8ad1cb8d5b1fa8e8f888e2d4b04623902382e8
+size 1357491877