{"alphabet_size": 4107, "pad_token_id": 1, "mask_token_id": 2, "max_positions": 2048, "embed_scale": 1.0, "emb_layer_norm_before": false, "attention_heads": 16, "key_size": 64, "embed_dim": 1024, "ffn_embed_dim": 4096, "num_layers": 29, "positional_embedding": null, "lm_head": "roberta", "add_bias_kv": false, "add_bias_ffn": false, "use_rotary_embedding": true, "ffn_activation_name": "swish", "use_glu_in_ffn": true, "mask_before_attention": false, "token_dropout": false, "masking_ratio": 0.0, "masking_prob": 0.0, "use_gradient_checkpointing": false, "k_for_kmers": 6} |