[BUG]Some errors when get_feat_dim

[auzxb]

Montreal-Forced-Aligner/montreal_forced_aligner/corpus/acoustic_corpus.py", line 776, in get_feat_dim return int(feats) ValueError: invalid literal for int() with base 10: ''

[auzxb]

    with open(os.path.join(self.features_log_directory, "feat-to-dim.log"), "w") as log_file:
        subset_proc = subprocess.Popen(
            [
                thirdparty_binary("subset-feats"),
                "--n=1",
                feature_string,
                "ark:-",
            ],
            stderr=log_file,
            stdout=subprocess.PIPE,
        )
        dim_proc = subprocess.Popen(
            [thirdparty_binary("feat-to-dim"), "ark:-", "-"],
            stdin=subset_proc.stdout,
            stdout=subprocess.PIPE,
            stderr=log_file,
        )
        stdout, stderr = dim_proc.communicate()
        feats = stdout.decode("utf8").strip()
    return int(feats)

In my experiment, the number of my audios is 500,000. When running to this step, it will report an error. training:

• monophone: subset: 10000

When I set the parameter subset = 0, the problem is solved!

[HandsLing]

i got the same error, and i found i don't have a .lab file for each .wav file

[Sreyan88]

I get the same error! Has anyone found out a solution to this? Thanks!