config.json

{
  "architectures": [
    "EsmCForPreTraining"
  ],
  "attention_bias": false,
  "attention_dropout": 0.0,
  "attention_layer_norm_bias": true,
  "bos_token_id": 1,
  "dtype": "float32",
  "eos_token_id": 2,
  "feedforward_bias": false,
  "final_layer_norm_bias": false,
  "head": null,
  "hidden_act": "swiglu",
  "hidden_dropout": 0.0,
  "hidden_size": 2560,
  "id2label": null,
  "initializer_range": 0.02,
  "intermediate_size": 6912,
  "label2id": null,
  "layer_norm_eps": 1e-05,
  "lm_head": {
    "act": null,
    "bias": true,
    "dropout": 0.0,
    "hidden_size": null,
    "layer_norm_eps": 1e-05,
    "loss_weight": null,
    "num_labels": 37,
    "output_name": null,
    "transform": "nonlinear",
    "transform_act": "gelu"
  },
  "mask_token_id": 4,
  "max_position_embeddings": 2048,
  "model_type": "esmc",
  "null_token_id": 5,
  "num_attention_heads": 40,
  "num_hidden_layers": 80,
  "num_labels": 1,
  "pad_token_id": 0,
  "qk_layer_norm": true,
  "qk_layer_norm_bias": false,
  "residue_scaling_factor": 1.4907119849998598,
  "tie_word_embeddings": false,
  "transformers_version": "5.7.0",
  "unk_token_id": 3,
  "vocab_size": 37
}