{ "architecture": { "embedding_dim": 256, "vocab_size": 128, "context_size": 256, "num_heads": 4, "num_layers": 4 }, "training": { "batch_size": 64, "training_data_path": "datasets/corpora/shakespeare.txt", "learning_rate": 3e-4, "num_steps": 500, "val_interval": 100, "save_folder": "checkpoints/candidates" }, "generation": { "top_k": 10, "top_p": 1, "temp": 0.8 }, "restore_path": "checkpoints/model.pt" }