config.json

{
  "_name_or_path": "../pre_trained_models/esm2_t33_650M_UR50D",
  "architectures": [
    "BertRnnDist"
  ],
  "attention_probs_dropout_prob": 0.0,
  "classifier_dropout": null,
  "cnn_dropout": 0.1,
  "cnn_filters": 512,
  "emb_layer_norm_before": false,
  "esmfold_config": null,
  "hidden_act": "gelu",
  "hidden_dropout_prob": 0.0,
  "hidden_size": 1280,
  "initializer_range": 0.02,
  "intermediate_size": 5120,
  "is_folding_model": false,
  "layer_norm_eps": 1e-05,
  "length": 50,
  "mask_token_id": 32,
  "max_position_embeddings": 1026,
  "model_type": "bert",
  "num_attention_heads": 20,
  "num_hidden_layers": 33,
  "num_rnn_layer": 2,
  "pad_token_id": 1,
  "position_embedding_type": "rotary",
  "rnn": "lstm",
  "rnn_dropout": 0.1,
  "rnn_hidden": 768,
  "token_dropout": true,
  "torch_dtype": "float32",
  "transformers_version": "4.24.0",
  "type_vocab_size": 2,
  "use_cache": true,
  "vocab_list": null,
  "vocab_size": 33
}