training_args.json

{
  "run_id": "20260906_020241",
  "model_name": "microsoft/deberta-v3-small",
  "num_labels": 11,
  "classes": [
    "QA",
    "Summary",
    "Coding",
    "Reasoning",
    "Agentic",
    "Creative",
    "Translation",
    "Classification",
    "Extraction",
    "Long-context",
    "Fast/simple"
  ],
  "n_per_class": 220,
  "train_size": 1936,
  "eval_size": 484,
  "max_seq_len": 96,
  "batch_size": 16,
  "epochs": 3,
  "learning_rate": 2e-05,
  "weight_decay": 0.01,
  "warmup_ratio": 0.1,
  "seed": 42,
  "device": "cpu"
}