{
  "activation": "gelu",
  "architectures": [
    "ProteomeLMForMaskedLM"
  ],
  "attention_dropout": 0.1,
  "batch_size": 32,
  "beta1": 0.9,
  "beta2": 0.999,
  "dataloader_num_workers": 8,
  "db_path": "/home/malbrank/training/",
  "dim": 512,
  "dropout": 0.1,
  "dtype": "bfloat16",
  "eval_accumulation_steps": 1,
  "eval_epochs": 1,
  "gradient_accumulation_steps": 1,
  "gradient_checkpointing": false,
  "hidden_dim": 512,
  "initializer_range": 0.02,
  "input_size": 1152,
  "learning_rate": 0.0003,
  "logging_steps": 10,
  "loss_type": "mse",
  "mask_fraction": 0.5,
  "max_grad_norm": 1.0,
  "max_position_embeddings": 512,
  "max_steps": 1000000,
  "min_taxid_size": 200,
  "model_type": "distilbert",
  "n_heads": 8,
  "n_inner": null,
  "n_layers": 6,
  "namedir": "ProteomeLM-Mini-Kuma",
  "num_epochs": 2000,
  "output_dir": "/home/malbrank/ProteomeLM/output/",
  "pad_token_id": 0,
  "qa_dropout": 0.1,
  "save_epochs": 15,
  "scheduler": "cosine",
  "seq_classif_dropout": 0.2,
  "sinusoidal_pos_embds": false,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.51.1",
  "use_cache": true,
  "use_one_gpu": "0",
  "vocab_size": 30522,
  "wandb_project": "proteomelm",
  "warmup_steps": 500,
  "weight_decay": 0.01
}