| { | |
| "_name_or_path": "AI4Protein/VenusPLM-300M", | |
| "act_fn": "silu", | |
| "architectures": [ | |
| "TransformerForMaskedLM" | |
| ], | |
| "attention_probs_dropout_prob": 0.1, | |
| "attn_impl": "sdpa", | |
| "auto_map": { | |
| "AutoConfig": "configuration_transformer.TransformerConfig", | |
| "AutoModelForMaskedLM": "modeling_transformer.TransformerForMaskedLM" | |
| }, | |
| "decoder_start_token_id": 1, | |
| "decoder_vocab_size": 36, | |
| "embedding_layer_norm": true, | |
| "embedding_shrinking": false, | |
| "flash_attention": true, | |
| "hidden_dropout_prob": 0.1, | |
| "hidden_size": 1024, | |
| "initializer_sigma": 0.01976423537605237, | |
| "intermediate_size": 4096, | |
| "layer_norm_eps": 1e-05, | |
| "layernorm_type": "rmsnorm", | |
| "mask_token_id": 24, | |
| "model_type": "transformer", | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 24, | |
| "pad_token_id": 0, | |
| "position_embedding_type": "rotary", | |
| "rotary_b": 10000, | |
| "token_dropout": false, | |
| "torch_dtype": "float32", | |
| "transformers_version": "4.45.0", | |
| "vocab_size": 36 | |
| } | |