| { | |
| "architectures": [ | |
| "LeanAlbertForMaskedLM" | |
| ], | |
| "attention_probs_dropout_prob": 0, | |
| "bos_token_id": 2, | |
| "classifier_dropout_prob": 0.1, | |
| "share_large_matrices": 1, | |
| "embedding_size": 128, | |
| "eos_token_id": 3, | |
| "hidden_act": "gelu_fused", | |
| "hidden_act_gated": true, | |
| "hidden_dropout_prob": 0, | |
| "hidden_size": 4096, | |
| "initializer_range": 0.02, | |
| "inner_group_num": 1, | |
| "intermediate_size": 16384, | |
| "layer_norm_eps": 1e-12, | |
| "position_embedding_type": "rotary", | |
| "model_type": "lean_albert", | |
| "num_attention_heads": 64, | |
| "num_hidden_groups": 32, | |
| "num_hidden_layers": 32, | |
| "num_memory_blocks": 0, | |
| "pad_token_id": 0, | |
| "type_vocab_size": 2, | |
| "vocab_size": 50005 | |
| } |