INFERENCE_VERIFICATION.json

{
  "model_name": "OpenSML-150M",
  "weights_sha256": "cbd3e3fb4ada74d7264371b79cee7598f513b7b950f1f141ef9f1a43bd2e1b4e",
  "strict_loading": "passed",
  "prompt": "Say hello in one sentence.",
  "generation": {
    "text": " I am a student.",
    "token_ids": [
      345,
      768,
      261,
      2560,
      17,
      1
    ],
    "stop_reason": "eos",
    "prompt_tokens": 13,
    "generated_tokens": 6
  },
  "precision": "FP32",
  "attention": "vanilla",
  "ffn": "reference",
  "scope": "Short local loading/generation check; no benchmark rerun or cross-framework parity claim.",
  "parameter_count": 150439168,
  "tested_mlx_runtime": "0.32.3.dev20260920+7241f12",
  "cached_full_prefix_argmax": "matched on checked prompt",
  "context_overflow": "rejected",
  "dependency_scope": "PyPI requirements are a baseline; smoke check used the recorded local development runtime."
}