{
  "dataset": "VoxLingua107",
  "split": "dev",
  "repo_id": "espnet/geolid_combined_shared_trainable",
  "source_hf_revision": "613d18645d1be3bd585a2cbae0485350a2a4c027",
  "implementation_revision": "301372e4da311d2a3d9574a5a43b3037ae6e67c1",
  "model": "ESPnetLIDModel",
  "model_sha256": "8a8f9dd767c141dcae0cb34a7780d7f593bc67f560267cd4d2d2fadd54f6bd16",
  "config_sha256": "b06ed46b21a44a48ee0860407d1fa772344fedae302380f74ae2b02df7757abb",
  "manifest_sha256": "6ad3c8fc6c74fab2720fe186525fc3bd20c4ab7b1be35d227c5c7f5456ba6c54",
  "utterances": 1609,
  "languages": 33,
  "metric": "utterance-level accuracy (%)",
  "correct": 1501,
  "accuracy_percent": 93.2877563704164,
  "paper_accuracy_percent": 94.4,
  "entrypoint": "espnet2.bin.lid_inference",
  "device": "cuda",
  "dtype": "float32",
  "batch_size": 1,
  "num_workers": 8,
  "use_preprocessor": false,
  "fix_duration": false,
  "input_order": "wav.scp sorted by utterance ID; standard ESPnet streaming iterator",
  "environment": {
    "python": "3.12.3",
    "torch": "2.11.0+cu128",
    "torchaudio": "2.11.0+cu128",
    "transformers": "5.14.1",
    "espnet-s3prl": "2026.5.14",
    "soundfile": "0.14.0"
  },
  "finished_at": "2026-10-23T00:55:52.840208+00:00",
  "seconds": 161.17425368726254,
  "predictions_file": "evaluation_predictions.jsonl",
  "predictions_sha256": "40264d2766f60dfce717ebc638ef76bc85380709693ca3e32178829e5df5453e",
  "comparison": {
    "baseline_implementation_revision": "ca777d59eb49519e3827e40cc34997f2123d6682",
    "baseline_correct": 1501,
    "matched_predictions": 1609,
    "changed_predictions": 0,
    "baseline_predictions_sha256": "40264d2766f60dfce717ebc638ef76bc85380709693ca3e32178829e5df5453e"
  },
  "per_language": {
    "ara": {
      "total": 100,
      "correct": 95
    },
    "aze": {
      "total": 34,
      "correct": 31
    },
    "cmn": {
      "total": 23,
      "correct": 22
    },
    "dan": {
      "total": 99,
      "correct": 93
    },
    "deu": {
      "total": 82,
      "correct": 77
    },
    "ell": {
      "total": 5,
      "correct": 4
    },
    "eng": {
      "total": 80,
      "correct": 73
    },
    "est": {
      "total": 85,
      "correct": 80
    },
    "fas": {
      "total": 100,
      "correct": 89
    },
    "fin": {
      "total": 93,
      "correct": 91
    },
    "fra": {
      "total": 100,
      "correct": 96
    },
    "hrv": {
      "total": 4,
      "correct": 3
    },
    "hun": {
      "total": 4,
      "correct": 4
    },
    "hye": {
      "total": 100,
      "correct": 96
    },
    "isl": {
      "total": 26,
      "correct": 26
    },
    "ita": {
      "total": 49,
      "correct": 49
    },
    "jpn": {
      "total": 42,
      "correct": 41
    },
    "lav": {
      "total": 95,
      "correct": 94
    },
    "lit": {
      "total": 13,
      "correct": 12
    },
    "mkd": {
      "total": 4,
      "correct": 4
    },
    "nld": {
      "total": 100,
      "correct": 95
    },
    "nno": {
      "total": 3,
      "correct": 3
    },
    "nor": {
      "total": 24,
      "correct": 13
    },
    "pol": {
      "total": 2,
      "correct": 2
    },
    "por": {
      "total": 1,
      "correct": 1
    },
    "rus": {
      "total": 29,
      "correct": 29
    },
    "slv": {
      "total": 9,
      "correct": 9
    },
    "spa": {
      "total": 55,
      "correct": 47
    },
    "srp": {
      "total": 7,
      "correct": 5
    },
    "swe": {
      "total": 100,
      "correct": 94
    },
    "tur": {
      "total": 72,
      "correct": 71
    },
    "ukr": {
      "total": 17,
      "correct": 17
    },
    "urd": {
      "total": 52,
      "correct": 35
    }
  },
  "batch_dependence": {
    "s3prl_frame_alignment": "S3PRL reports ceil(samples/320) frames and repeats the final frame only at the batch boundary. Shorter items can instead expose padded-query frames from Wav2Vec2.",
    "ecapa_tdnn": "Temporal convolutions and unmasked SE global means depend on the padded sequence length before the masked final pooling.",
    "mvn_and_final_pooling_mask_padding": true,
    "reported_measurement_batch_size": 1
  },
  "resolved_config_sha256": "fe2e8acd705871a4dea823f80c7afabfcb89b2cb40cf51161178b6f3209b72f5",
  "languages_sha256": "285a350acd7dc5b109e950e54560b8bab70d282bae12b25328327d28a2b5226d",
  "log_file": "evaluation.log",
  "log_sha256": "d6831bc2d1e154c487052b31b10adc9072db4c8d045e4a1c6897125aa7a4c6c1",
  "prediction_differences_file": "evaluation_prediction_differences.json",
  "previous_evaluation_file": "evaluation_previous.json",
  "source_config_url": "https://huggingface.co/espnet/geolid_combined_shared_trainable/blob/613d18645d1be3bd585a2cbae0485350a2a4c027/config.yaml",
  "loader": "Unmodified standard ESPnet CLI loader; strict loading was not forced"
}