{
  "schema": "transcribe-golden-manifest-v1",
  "family": "canary",
  "variant": "canary-1b-flash",
  "source_model": {
    "hf_repo": "nvidia/canary-1b-flash",
    "hf_revision": "a9a55e0295e7dd50d0c8c2a19491900a0daf24f3"
  },
  "reference": {
    "kind": "nemo",
    "source": "https://github.com/NVIDIA-NeMo/NeMo",
    "revision": "v2.7.2",
    "entrypoint": "scripts/dump_reference_canary_nemo.py"
  },
  "expected_dtype": "float32",
  "dtype_source": "config",
  "frontend": {
    "sample_rate": 16000,
    "n_mels": 128,
    "hop_length": 160,
    "fft_size": 512,
    "win_length": 400,
    "window": "hann_periodic",
    "normalization": "per_feature",
    "preemphasis": 0.97,
    "dither": 0.0
  },
  "tokenizer_summary": {
    "type": "sentencepiece",
    "vocab_size": null,
    "special_tokens": {
      "bos": 1,
      "eos": 2,
      "pad": 0
    }
  },
  "capabilities": {
    "languages": [
      "en",
      "de",
      "es",
      "fr"
    ],
    "language_detection": false,
    "translation": true,
    "translation_target_languages": ["en", "de", "es", "fr"],
    "translation_pairs": ["en>de", "de>en", "en>es", "es>en", "en>fr", "fr>en"],
    "timestamps": [
      "word",
      "segment"
    ],
    "streaming": false,
    "speaker_diarization": false
  },
  "multitask_default": {
    "source_lang": "en",
    "target_lang": "en",
    "task": "asr",
    "pnc": "yes",
    "toggle_timestamps": "no",
    "beam": 1
  },
  "tolerance_file": "tests/tolerances/canary.json",
  "cases": [
    "jfk"
  ],
  "transcript_compare": "normalized"
}
