{
  // Data Arguments
  //################
  "dataset_HF": "c-bone/alex_mp_20",
  "pretrained_tokenizer_dir": "_utils/HF-cif-tokenizer",
  "context_length": 1024,
  // Filters
  "remove_CIFs_above_context": false,
  "remove_CIFs_with_unk": true,
  // "activate_conditionality": "None",

  // Model Arguments
  //#################
  // Model Depth
  // n_positions has been tied to context_length
  "n_embd": 512,
  "n_layer": 8,
  "n_head": 8,
  // Dropout
  "residual_dropout": 0.1,
  "embedding_dropout": 0.1,
  "attention_dropout": 0.1,

  // Trainer Arguments
  //###################
  // Batching
  "train_batch_size": 32,
  "eval_batch_size": 32,
  "gradient_accumulation_steps": 1,
  // Learning Rate and Optimizer
  "learning_rate": 0.001,
  "lr_scheduler_type": "cosine_with_min_lr",
  "lr_scheduler_kwargs": {
    "min_lr_rate": 0.01
  },
  "warmup_ratio": 0.0,
  "adam_beta1": 0.9,
  "adam_beta2": 0.98,
  "grad_clip": 1.0,
  "weight_decay": 0.01,
  "optimizer": "muon",
  "muon_lr": 0.01,
  "muon_momentum": 0.95,
  // Logging
  "output_dir": "model_ckpts/alex_mp_20_text/",
  "save_total_limit": 2,
  "report_to": "wandb",
  "wandb_project_folder": "CrystaLLM-pi",
  "eval_strategy": "steps",
  "eval_steps": 10000,
  "logging_steps": 50,
  "save_strategy": "steps",
  "max_steps": 375000,
  "early_stopping_patience": 15,
  "early_stopping_threshold": 0.000005,
  // Utils
  "seed": 1,
  "data_seed": 1,
  "load_best_model_at_end": true,
  "metric_for_best_model": "eval_loss",
  "greater_is_better": false,
  "torch_compile": true,
  "deepspeed_config": "_config_files/deepspeed_default.json",

  // CodeCarbon Arguments
  //######################
  "codecarbon": true,
  "tracker_project": "CrystaLLM-pi"
}