File size: 1,797 Bytes
44894b6 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 | {
"schema": "visualears-preprocessor-v2-nemo-fbank",
"description": "NeMo AudioToMelSpectrogramPreprocessor-compatible browser feature pipeline for shenava_koochik_1_0. normalize=NA, preemphasis=0.97, torch.stft center=True, Slaney mel filters, requires processed_signal_length.",
"sample_rate": 16000,
"n_fft": 512,
"win_length": 400,
"hop_length": 160,
"n_mels": 80,
"window": "hann_periodic_false",
"window_formula": "0.5 - 0.5*cos(2*pi*i/(win_length-1))",
"center": true,
"center_pad": 256,
"pad_mode": "reflect",
"preemphasis": 0.97,
"mel_scale": "slaney/librosa.filters.mel(htk=False,norm='slaney')",
"mel_filters_file": "mel_filters_slaney_80x257.json",
"spectrum": "magnitude_power_2_no_fft_normalization",
"log": "natural",
"log_zero_guard_type": "add",
"log_zero_guard_value": 5.960464477539063e-08,
"normalize": "NA",
"fixed_frames": 2005,
"pad_value": 0.0,
"frame_count_formula": "max(1, min(fixed_frames, floor(num_samples / hop_length) + 1))",
"inputs": [
{
"name": "processed_signal",
"dtype": "float16",
"shape": [
"batch",
80,
2005
]
},
{
"name": "processed_signal_length",
"dtype": "int64",
"shape": [
"batch"
]
}
],
"outputs": [
{
"name": "logits",
"dtype": "float16",
"shape": [
"batch",
252,
1025
]
},
{
"name": "encoded_lengths",
"dtype": "int64",
"shape": [
"batch"
]
}
],
"output_stride": 8,
"usable_steps_formula": "min(encoded_lengths[0], logits_steps)",
"ms_per_output_step": 80,
"blank_id": 1024,
"ctc_decode": "greedy argmax; drop blank, repeats, and SentencePiece special tokens; join pieces; '▁' -> space"
}
|