{ "model": { "vocab_size": 3072, "max_seq_len": 84, "d_model": 384, "n_layers": 6, "n_heads": 6, "ffn_hidden": 1152, "dropout": 0.1, "pad_id": 0, "bos_id": 1, "eos_id": 2 }, "train": { "batch_size": 32, "learning_rate": 0.0003, "min_lr": 3e-05, "weight_decay": 0.1, "warmup_steps": 400, "max_steps": 12000, "eval_interval": 400, "save_interval": 1000, "grad_clip": 1.0, "device": "auto", "seed": 42, "data_dir": "data", "output_dir": "checkpoints" } }