hparams.json

{
  "optimizer":"AdamW",
  "learning_rate":0.0005,
  "adam_epsilon":0.00001,
  "warmup_steps":25,
  "weight_decay":0,
  "train_batch_size":64,
  "eval_batch_size":64,
  "use_scheduler":true,
  "metrics":null,
  "model_name_or_path":"allegro/herbert-base-cased",
  "num_classes":7,
  "finetune_last_n_layers":4,
  "config_kwargs":{
    "classifier_dropout":0.2
  },
  "task_model_kwargs":{
    "adam_epsilon":0.00001,
    "eval_batch_size":64,
    "learning_rate":0.0005,
    "optimizer":"AdamW",
    "train_batch_size":64,
    "use_scheduler":true,
    "warmup_steps":25,
    "weight_decay":0
  },
  "model_compile_kwargs":null,
  "evaluation_mode":"unit",
  "tagging_scheme":null,
  "ignore_index":-100,
  "downstream_model_type":"AutoModelForTokenClassification"
}