| { | |
| "_name_or_path": "mlm-scratch", | |
| "add_prefix_space": true, | |
| "architectures": [ | |
| "RobertaForMaskedLM" | |
| ], | |
| "attention_probs_dropout_prob": 0.1, | |
| "bos_token": "<s>", | |
| "bos_token_id": 0, | |
| "cls_token": "<s>", | |
| "eos_token": "</s>", | |
| "eos_token_id": 2, | |
| "errors": "replace", | |
| "gradient_checkpointing": false, | |
| "hidden_act": "gelu", | |
| "hidden_dropout_prob": 0.1, | |
| "hidden_size": 768, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 3072, | |
| "layer_norm_eps": 1e-12, | |
| "mask_token": "<mask>", | |
| "max_position_embeddings": 512, | |
| "model_max_length": 512, | |
| "model_type": "roberta", | |
| "num_attention_heads": 12, | |
| "num_hidden_layers": 12, | |
| "pad_token": "<pad>", | |
| "pad_token_id": 1, | |
| "position_embedding_type": "absolute", | |
| "sep_token": "</s>", | |
| "transformers_version": "4.8.2", | |
| "type_vocab_size": 2, | |
| "unk_token": "<unk>", | |
| "use_cache": true, | |
| "vocab_size": 52000 | |
| } | |