evaluation/published-consumer/linux/cuda-0.json

{
  "ready": {
    "kind": "caption_stream_ready",
    "backend": "cuda",
    "device_index": 0,
    "model_sha256": "f447bfe0d62200be97e0ce36e0883eb4499a42baded0840b15e9a1b68be17f83",
    "runtime_info": {
      "model_sha256": "f447bfe0d62200be97e0ce36e0883eb4499a42baded0840b15e9a1b68be17f83",
      "tensor_count": 938,
      "resident_weight_payload_bytes": 2335448192,
      "allocated_weight_buffer_bytes": 2335498752,
      "arithmetic_profile": "f16_hidden_bf16_cache_f32_acc_no_q8_explicit_rounding_ggml_gelu_erf_v1",
      "precision_sentinel": {
        "identity": "audio8-q4_0-f16-f32-device-sentinel-v2",
        "device": {
          "backend": "CUDA",
          "device_index": 0,
          "name": "CUDA0",
          "description": "NVIDIA GeForce RTX 5080 Laptop GPU",
          "device_id": "0000:64:00.0",
          "integrated": false,
          "memory_free_bytes": 16376201216,
          "memory_total_bytes": 16609247232,
          "caps": {
            "async": true,
            "host_buffer": true,
            "events": true,
            "mmap": true
          },
          "feature_api_available": true,
          "features": {
            "ARCHS": "1200",
            "BLACKWELL_NATIVE_FP4": "1",
            "FA_QUANTS": "q4_0-q4_0,q8_0-q8_0,f16-f16,bf16-bf16",
            "USE_GRAPHS": "1"
          }
        },
        "shape": [
          3,
          64,
          1
        ],
        "input": "F16",
        "accumulator": "F32",
        "counter_delta": 1,
        "counter_scope": "device_kernel_invocations_including_graph_replay",
        "max_abs_error": 0,
        "scratch_bytes": 640,
        "cpu_fallback": false,
        "full_model_fidelity_verified": false,
        "passed": true
      },
      "device": {
        "backend": "CUDA",
        "device_index": 0,
        "name": "CUDA0",
        "description": "NVIDIA GeForce RTX 5080 Laptop GPU",
        "device_id": "0000:64:00.0",
        "integrated": false,
        "memory_free_bytes": 13725401088,
        "memory_total_bytes": 16609247232,
        "caps": {
          "async": true,
          "host_buffer": true,
          "events": true,
          "mmap": true
        },
        "feature_api_available": true,
        "features": {
          "ARCHS": "1200",
          "BLACKWELL_NATIVE_FP4": "1",
          "FA_QUANTS": "q4_0-q4_0,q8_0-q8_0,f16-f16,bf16-bf16",
          "USE_GRAPHS": "1"
        }
      }
    },
    "silero_model_sha256": "1a153a22f4509e292a94e67d6f9b85e8deb25b4988682b7e174c65279d8788e3",
    "vad_backend": "tract_cpu",
    "policy": "silero_v5_threshold0.5_silence2000ms_preroll400ms_v1",
    "vad_state_reset_on_asr_reset": false,
    "startup": "realtime18",
    "delay_ms": 240,
    "batch_windows": 1,
    "dtype": "float16",
    "cache_dtype": "bfloat16",
    "activation_quantization": false,
    "cpu_asr_fallback": false,
    "input_mode": "stdin",
    "model_loaded_once": true,
    "load_seconds": 2.582996022,
    "measured_word_aligned_latency": false,
    "production_accuracy_approved": false
  },
  "final": {
    "kind": "caption_stream_complete",
    "status": "complete",
    "audio_samples": 93680,
    "audio_seconds": 5.855,
    "segments": 1,
    "selected_samples": 90864,
    "unconsumed_selected_samples": 0,
    "emissions": 85,
    "text_bytes": 88,
    "model_seconds": 3.0049685310000003,
    "reset_seconds": 0.005830358,
    "vad_seconds": 0.059070129000000006,
    "read_seconds": 3.2423218840000003,
    "sleep_seconds": 0,
    "elapsed_seconds": 6.314729831,
    "work_rtf": 0.5243158015371477,
    "peak_cache_bytes": 210169856,
    "max_retained_input_samples": 6912,
    "input_sha256": "22d472b40f913206b6917115d136306be88df3b278e77f68be5910423250fa4d",
    "generated_ids_sha256": "4cab5d89ace01105314a3ff9ea386dd217c2cb319c74d0295a34ead46d3c44f7",
    "all_source_audio_observed": true,
    "measured_word_aligned_latency": false,
    "production_accuracy_approved": false,
    "model_sha256": "f447bfe0d62200be97e0ce36e0883eb4499a42baded0840b15e9a1b68be17f83"
  },
  "generated_token_ids": [
    151665,
    151665,
    151665,
    151665,
    151665,
    151665,
    151665,
    151665,
    151665,
    151665,
    151665,
    151665,
    151666,
    12275,
    13,
    151665,
    151665,
    151665,
    151666,
    3406,
    2044,
    374,
    279,
    151665,
    151665,
    151665,
    151665,
    151665,
    151666,
    38471,
    273,
    315,
    279,
    151665,
    151666,
    6149,
    151665,
    151665,
    151665,
    151665,
    151665,
    151665,
    151666,
    6846,
    11,
    323,
    151665,
    151666,
    582,
    525,
    151665,
    151665,
    151666,
    15713,
    311,
    151665,
    151665,
    151665,
    151665,
    151666,
    10565,
    151665,
    151666,
    806,
    151665,
    151665,
    151665,
    151665,
    151665,
    151665,
    151666,
    41482,
    13,
    151665,
    151665,
    151665,
    151665,
    151665,
    151665,
    151665,
    151665,
    151665,
    151665,
    151665,
    151665
  ],
  "text": "Mr. Quilter is the apostle of the middle classes, and we are glad to welcome his gospel.",
  "segments": [
    {
      "kind": "caption_segment_end",
      "segment": 0,
      "start_sample": 2816,
      "end_sample": 93680,
      "final_eof": true,
      "unconsumed_samples": 0,
      "windows": 85,
      "seconds": 6.314727546,
      "text": "Mr. Quilter is the apostle of the middle classes, and we are glad to welcome his gospel."
    }
  ],
  "batch_work_rtf": 0.5243158015371477,
  "first_readable_seconds": 1.294302616,
  "clock_and_token_oracle_passed": true,
  "feeder": {
    "chunks": 293,
    "samples": 93680,
    "duration_seconds": 5.855090425116941,
    "max_write_seconds": 5.851895548403263e-05,
    "max_delivery_lateness_seconds": 0.0009123341180385935,
    "process_started_monotonic": 1189195.64397091,
    "feeder_started_monotonic": 1189198.455315203,
    "ready_received_monotonic": 1189198.455064014,
    "clock_scope": "Feeder starts after host receipt of native readiness. Native and observer clocks are separately recorded; no microphone capture clock."
  },
  "command": [
    "/bin/sh",
    "/dev/shm/audio8-q3-public-linux-consumer-v1/run-cuda.sh",
    "-",
    "en",
    "0"
  ],
  "invoked_shipped_launcher": true,
  "memory": {
    "os_peak_process_rss_bytes": 651280384,
    "os_peak_memory_footprint_bytes": null,
    "footprint_available": false,
    "rss_source_unit": "KiB (1024 bytes)",
    "elapsed_seconds": 9.38,
    "user_seconds": 5.12,
    "system_seconds": 0.6,
    "scope": "Fresh native child process including loading and inference; excludes Python harness. RSS and footprint are separate measurements and must not be summed."
  }
}