icefall/egs/librispeech/encodec/local/convert_h5_to_txt.py
2023-09-22 13:49:29 +08:00

22 lines
642 B
Python
Executable File

import h5py
import numpy as np
from tqdm import tqdm
with h5py.File("download/DiscreteAudioToken/encodec/feats.h5") as reader, open(
"download/DiscreteAudioToken/encodec/out_quantized", "a"
) as writer:
utt_ids = sorted(reader.keys())
for utt_id in tqdm(utt_ids):
data = np.array(reader[utt_id])
data_line = str(utt_id) + " "
for i in range(8):
data_line += (
str(data.T[i].tolist())
.replace(",", "")
.replace("[", "")
.replace("]", "")
)
data_line += " "
writer.write(data_line.strip() + "\n")