{ "D": 512, "L": 8, "H": 8, "T": 512, "VOCAB": 8192, "model_type": "atom-gpt", "architecture": "decoder-only transformer, RMSNorm, RoPE, SwiGLU, tied embeddings", "step": 5968, "best_val": 1.2423416799114597, "eos_token": "<|endoftext|>", "eos_token_id": 0, "prompt_template": "### Instruction:\nWrite a short story.\nWords: ...\nFeatures: ...\n### Story:\n" }