HeAR-s1.1 / maeb_run_config.json
matthewagi's picture
Release HeAR-s1.1
a5776be verified
{
"checkpoint": "/home/ubuntu/maeb_audio_only_eval/models/ckpt_final.pt",
"embedding_head": "student",
"device_request": "cuda",
"resolved_device": "cuda",
"target_sr": 16000,
"clip_seconds": 2.0,
"crop": "center",
"full_clip": false,
"sliding_window": true,
"window_hop_seconds": 2.0,
"window_pool": "mean",
"batch_size_request": 0,
"batch_size": 8192,
"num_proc": null,
"compile": true,
"compile_mode": "default",
"compile_dynamic": true,
"runtime": {
"tf32": true,
"cudnn_benchmark": true,
"matmul_precision": "high",
"amp": true,
"amp_dtype": "bfloat16",
"channels_last": true
},
"classification_max_iter": 1000,
"classification_zero_division": 0,
"classification_defaults_patched": true,
"classification_evaluator_updates": 11,
"dataset_transform_compat_patches": 5,
"audio_task_compat_patches": [
"VoxPopuliLanguageID"
],
"benchmark_name": "MAEB(beta, audio-only)",
"task_source": "MAEB(beta, audio-only)",
"tasks": [
"JamAltArtistA2ARetrieval",
"BeijingOpera",
"BirdCLEF",
"CREMA_D",
"CommonLanguageAgeDetection",
"GTZANGenre",
"IEMOCAPGender",
"MInDS14",
"MridinghamTonic",
"VoxCelebSA",
"VoxPopuliLanguageID",
"CREMA_DClustering",
"VehicleSoundClustering",
"VoxPopuliGenderClustering",
"SIBFLEURS",
"CREMADPairClassification",
"NMSQAPairClassification",
"VoxPopuliAccentPairClassification",
"GTZANAudioReranking"
]
}