{ "_name_or_path": "ltg/norbert3-base", "architectures": [ "NorbertModel" ], "attention_probs_dropout_prob": 0.1, "auto_map": { "AutoConfig": "ltg/norbert3-base--configuration_norbert.NorbertConfig", "AutoModel": "ltg/norbert3-base--modeling_norbert.NorbertModel", "AutoModelForMaskedLM": "ltg/norbert3-base--modeling_norbert.NorbertForMaskedLM", "AutoModelForMultipleChoice": "ltg/norbert3-base--modeling_norbert.NorbertForMultipleChoice", "AutoModelForQuestionAnswering": "ltg/norbert3-base--modeling_norbert.NorbertForQuestionAnswering", "AutoModelForSequenceClassification": "ltg/norbert3-base--modeling_norbert.NorbertForSequenceClassification", "AutoModelForTokenClassification": "ltg/norbert3-base--modeling_norbert.NorbertForTokenClassification" }, "hidden_dropout_prob": 0.1, "hidden_size": 768, "intermediate_size": 2048, "layer_norm_eps": 1e-07, "max_position_embeddings": 512, "num_attention_heads": 12, "num_hidden_layers": 12, "output_all_encoded_layers": true, "position_bucket_size": 32, "torch_dtype": "float32", "transformers_version": "4.42.3", "vocab_size": 50000 }