training_cfg.json

{
  "model_name": "/home/ubuntu/code/llm-eval/models/single_label/deberta-v3-large/exp5/checkpoint-1500",
  "lr": 3e-07,
  "epochs": 1,
  "weight_decay": 0.01,
  "train_bs": 16,
  "eval_bs": 16,
  "gradient_accumulation_steps": 4,
  "gradient_checkpointing": true,
  "eval_datasets": [
    {"path": "./data/all_eval/osu_eval.csv", "name": "osu"},
    {"path": "./data/all_eval/deepset_hallucination_eval_only_all_agree.csv", "name": "deepset_hall"},
    {"path": "./data/all_eval/deepset_flower_hallucinations_eval_only_all_agree.csv", "name": "deepset_flower"}
  ],
  "train_dataset_path": "./data/all_training/all_train_v2.csv",
  "exclude_datasets": ["nq"],
  "tf32": true,
  "bf16": false,
  "half_precision": true,
  "optim": "adamw_bnb_8bit"
}