| { |
| "dataset": "VoxLingua107", |
| "split": "dev", |
| "utterances": 1609, |
| "languages": 33, |
| "metric": "utterance-level accuracy (%)", |
| "entrypoint": "espnet2.bin.lid_inference", |
| "implementation_revision": "ca777d59eb49519e3827e40cc34997f2123d6682", |
| "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", |
| "manifest_sha256": "6ad3c8fc6c74fab2720fe186525fc3bd20c4ab7b1be35d227c5c7f5456ba6c54", |
| "environment": { |
| "python": "3.12.3 (main, Jun 19 2026, 12:46:00) [GCC 13.3.0]", |
| "torch": "2.11.0+cu128", |
| "torch_cuda_build": "12.8", |
| "torchaudio": "2.11.0+cu128", |
| "transformers": "5.14.1", |
| "espnet_s3prl": "2026.5.14", |
| "soundfile": "0.14.0" |
| }, |
| "repo_id": "espnet/geolid_combined_shared_trainable", |
| "model_sha256": "8a8f9dd767c141dcae0cb34a7780d7f593bc67f560267cd4d2d2fadd54f6bd16", |
| "correct": 1501, |
| "accuracy_percent": 93.2877563704164, |
| "paper_accuracy_percent": 94.4, |
| "source_hf_revision_before_layout_migration": "0c7ee08dc9ab7d9ac651c7858fd6ce40849475b6", |
| "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 |
| } |
| } |
|
|
