| { | |
| "architectures": [ | |
| "CronformerModel" | |
| ], | |
| "attention_probs_dropout_prob": 0.1, | |
| "cron_vocab_size": 78, | |
| "hidden_act": "gelu", | |
| "hidden_dropout_prob": 0.1, | |
| "hidden_size": 768, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 3072, | |
| "lang_tokenizer": "distilbert/distilbert-base-uncased", | |
| "lang_vocab_size": 30522, | |
| "layer_norm_eps": 1e-12, | |
| "max_position_embeddings": 512, | |
| "model_type": "cronformer", | |
| "num_attention_heads": 12, | |
| "num_hidden_dec_layers": 4, | |
| "num_hidden_enc_layers": 6, | |
| "num_hidden_layers": 10, | |
| "pad_token_id": 0, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.42.4", | |
| "type_vocab_size": 2 | |
| } | |