quantized_artifacts/artifact_manifest.json

{
  "base_model": "nvidia/canary-qwen-2.5b",
  "artifact_type": "optimum-quanto-qint8-llm-state-dict-plus-quantization-map",
  "created_at": "2026-06-04T17:28:19-0700",
  "scope": "SALM LLM component only",
  "notes": [
    "The SALM perception/audio encoder, tokenizer, and config are loaded from the upstream base model.",
    "This artifact is intended for this service wrapper and reload path.",
    "It is not advertised as a generic Transformers AutoModel drop-in."
  ]
}