| { |
| "architectures": [ |
| "GPTBertForMaskedLM", |
| "GPTBertForCausalLM", |
| "GPTBertForSequenceClassification" |
| ], |
| "attention_probs_dropout_prob": 0.1, |
| "auto_map": { |
| "AutoConfig": "configuration_gpt_bert.GPTBertConfig", |
| "AutoModel": "modeling_gpt_bert.GPTBertForMaskedLM", |
| "AutoModelForCausalLM": "modeling_gpt_bert.GPTBertForCausalLM", |
| "AutoModelForMaskedLM": "modeling_gpt_bert.GPTBertForMaskedLM", |
| "AutoModelForSequenceClassification": "modeling_gpt_bert.GPTBertForSequenceClassification" |
| }, |
| "bos_token_id": 1, |
| "classifier_dropout": 0.1, |
| "classifier_layer_norm_eps": 1e-05, |
| "eos_token_id": 2, |
| "force_causal_mask": true, |
| "hidden_dropout_prob": 0.1, |
| "hidden_size": 384, |
| "intermediate_size": 1280, |
| "layer_norm_eps": 1e-05, |
| "mask_token_id": 4, |
| "max_position_embeddings": 512, |
| "model_type": "gpt_bert", |
| "num_attention_heads": 6, |
| "num_hidden_layers": 12, |
| "num_labels": 2, |
| "pad_token_id": 3, |
| "position_bucket_size": 32, |
| "vocab_size": 8192 |
| } |
|
|
