kaldiio_read_write_ark_scp_numpy_arrays.py

1import kaldiio
2import numpy as np
3
4# 1. Writing data to an ark file and creating an scp file
5data = {
6    "uttid1": np.random.randn(10, 40).astype(np.float32),
7    "uttid2": np.random.randn(15, 40).astype(np.float32),
8}
9
10with kaldiio.WriteHelper('ark,scp:out.ark,out.scp') as writer:
11    for uttid, matrix in data.items():
12        writer(uttid, matrix)
13
14# 2. Reading data from an scp file
15# load_scp returns a dict-like object that loads data lazily
16scp_data = kaldiio.load_scp('out.scp')
17for uttid in scp_data:
18    matrix = scp_data[uttid]
19    print(f"Read {uttid} with shape {matrix.shape}")
20
21# 3. Reading data using ReadHelper (generator style)
22with kaldiio.ReadHelper('scp:out.scp') as reader:
23    for uttid, matrix in reader:
24        print(f"Read via ReadHelper: {uttid}")