config.json

{
  "architectures": [
    "Isoformer"
  ],
  "auto_map": {
    "AutoConfig": "isoformer_config.IsoformerConfig",
    "AutoModelForMaskedLM": "modeling_isoformer.Isoformer"
  },
  "enformer_attn_dim_key": 64,
  "enformer_attn_dropout": 0.05,
  "enformer_depth": 11,
  "enformer_dim": 1536,
  "enformer_dim_divisible_by": 128,
  "enformer_dropout_rate": 0.4,
  "enformer_heads": 8,
  "enformer_num_downsamples": 7,
  "enformer_output_heads": 0,
  "enformer_pos_dropout": 0.01,
  "enformer_target_length": 896,
  "enformer_use_checkpointing": false,
  "enformer_use_convnext": false,
  "enformer_use_tf_gamma": false,
  "esm_add_bias_fnn": true,
  "esm_attention_probs_dropout_prob": 0.0,
  "esm_emb_layer_norm_before": false,
  "esm_hidden_dropout_prob": 0.0,
  "esm_hidden_size": 640,
  "esm_intermediate_size": 2560,
  "esm_mask_token_id": 32,
  "esm_max_position_embeddings": 1026,
  "esm_num_attention_heads": 20,
  "esm_num_hidden_layers": 30,
  "esm_pad_token_id": 1,
  "esm_position_embedding_type": "rotary",
  "esm_tie_word_embeddings": false,
  "esm_token_dropout": true,
  "esm_use_cache": false,
  "esm_vocab_size": 33,
  "model_type": "isoformer",
  "nt_add_bias_fnn": false,
  "nt_attention_probs_dropout_prob": 0.0,
  "nt_emb_layer_norm_before": false,
  "nt_hidden_dropout_prob": 0.0,
  "nt_hidden_size": 768,
  "nt_intermediate_size": 3072,
  "nt_mask_token_id": 2,
  "nt_max_position_embeddings": 2050,
  "nt_num_attention_heads": 16,
  "nt_num_hidden_layers": 24,
  "nt_pad_token_id": 1,
  "nt_position_embedding_type": "rotary",
  "nt_tie_word_embeddings": false,
  "nt_token_dropout": false,
  "nt_use_cache": 0.0,
  "nt_vocab_size": 4107,
  "num_heads_omics_cross_attention": 8,
  "num_protein_tokens_per_seq": 2048,
  "num_tokens_per_seq_nuctf": 2048,
  "num_tokens_per_seq_nuctf_rna": 2048,
  "pool_window_end": 770,
  "pool_window_start": 765,
  "torch_dtype": "float32",
  "transformers_version": "4.29.2"
}