data/Whisper-Sidecar-data-metadata/stat.py

import glob

tsv_files = glob.glob("/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/*.tsv")
head = "\t".join(["id", "audio", "n_frames", "prompt", "tgt_text", "codec", "with_speech", "language", "speakers", "genders"])

tsv_files = [
    "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/de_train.tsv",
    "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/de-en-2mix_train.tsv",
    "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/de-en-3mix_train.tsv",
    # "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/de-en-2mix_train_targetLingual.tsv",
    # "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/de-en-3mix_train_targetLingual.tsv",
    "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/librispeech_train_fixed.tsv",
    "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/librispeech2mix_train_fixed.tsv",
    "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/librispeech3mix_train_fixed.tsv",
    "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/libri2mix_train_fixed.tsv",
    "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/libri3mix_train_fixed.tsv",
    # "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/de_test.tsv",
    # "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/de-en-2mix_test.tsv",
    # "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/de-en-3mix_test.tsv",
    # "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/de-en-2mix_test_targetLingual.tsv",
    # "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/de-en-3mix_test_targetLingual.tsv",
    # "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/librispeech_test_fixed.tsv",
    # "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/librispeech2mix_test_fixed.tsv",
    # "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/librispeech3mix_test_fixed.tsv",
    # "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/libri2mix_test_fixed.tsv",
    # "/valleblob/v-lingmeng/speech/data/Whisper-Sidecar-data-metadata/data_for_wavllm/libri3mix_test_fixed.tsv",
]

all_count = 0.0
for tsv in tsv_files:
    if "test" in tsv:
        continue
    with open(tsv, "r") as f:
        lines = f.readline()
        lines = f.readlines()
    print(tsv)
    print(len(lines))
    count = 0.0
    max_sec = 0
    new_lines = []
    for line in lines:
        new_line = line.strip().split("\t")
        # if float(new_line[2])/16000 > 40:
        #     continue
        new_lines.append("\t".join(new_line))
        # max_sec = max_sec if max_sec > float(new_line[2])/16000 else float(new_line[2])/16000
        count += float(new_line[2])/16000
    # with open(tsv, "w") as f:
    #     f.write(head + "\n" + "\n".join(new_lines)) 

    print(count/60/60)
    print(max_sec)
    print()
    all_count += count/60/60

print(all_count)