flexEAT-base_epoch30_pretrain

import torchaudio
import torch
import soundfile as sf
import numpy as np
from transformers import AutoModel

model_id = "HTill/flexEAT-base_epoch30_pretrain"
model = AutoModel.from_pretrained(model_id, trust_remote_code=True).eval().cuda()

source_file = "/path/to/input.wav"
target_file = "/path/to/output.npy"
norm_mean = -4.268
norm_std = 4.569

# Load and resample audio
wav, sr = sf.read(source_file)
waveform = torch.tensor(wav).float().cuda()
if sr != 16000:
    waveform = torchaudio.functional.resample(waveform, sr, 16000)

# Normalize and convert to mel-spectrogram
waveform = waveform - waveform.mean()
mel = torchaudio.compliance.kaldi.fbank(
    waveform.unsqueeze(0),
    htk_compat=True,
    sample_frequency=16000,
    use_energy=False,
    window_type='hanning',
    num_mel_bins=128,
    dither=0.0,
    frame_shift=10
).unsqueeze(0)

# Normalize
mel = (mel - norm_mean) / (norm_std * 2)
mel = mel.unsqueeze(0).cuda()  # shape: [1, 1, T, F]

# Extract features
with torch.no_grad():
    feat = model.extract_features(mel)

feat = feat.squeeze(0).cpu().numpy()
np.save(target_file, feat)
print(f"Feature shape: {feat.shape}")
print(f"Saved to: {target_file}")