training_config.json

{
    "model_name": "roberta-base",
    "output_dir": "./ckpts/finetune_relu_roberta_from_gelu_sst5",
    "learning_rate": 1e-05,
    "batch_size": 32,
    "num_epochs": 20,
    "max_length": 128,
    "warmup_steps": 500,
    "weight_decay": 0.01,
    "seed": 42,
    "hidden_act": "relu",
    "timestamp": "2025-01-13 19:56:34",
    "cuda_available": true,
    "cuda_device": "NVIDIA A100 80GB PCIe"
}