{ "cells": [ { "cell_type": "code", "execution_count": 2, "id": "147aea9e", "metadata": {}, "outputs": [], "source": [ "import numpy as np\n", "import pandas as pd\n", "\n", "%matplotlib inline\n", "\n", "import os\n", "import json\n", "import shutil\n", "\n", "import portion\n", "import tempfile\n", "\n", "from suno_utils.utils.conversion import (\n", " convert_audio,\n", " get_audio_dur_s,\n", " get_audio_slice,\n", " read_safe_wav,\n", " write_wav,\n", ")\n", "from suno_utils.datasets.supreme_court.slicer import _get_vad_gaps" ] }, { "cell_type": "code", "execution_count": null, "id": "3b5250a0", "metadata": {}, "outputs": [], "source": [] }, { "cell_type": "code", "execution_count": 3, "id": "d56235fa", "metadata": {}, "outputs": [], "source": [ "\n", "FILE_PATH = \"/home/ubuntu/code/meets/output.wav\"\n", "#FILE_PATH = \"/my/path.mp3\"\n", "SAMPLE_RATE = 16_000" ] }, { "cell_type": "code", "execution_count": 4, "id": "59f43d83", "metadata": { "scrolled": false }, "outputs": [], "source": [ "with tempfile.TemporaryDirectory() as temp_dir:\n", " # create wav file of known sample rate\n", " tmp_vad_wav_fp = temp_dir + FILE_PATH.split(\"/\")[-1].split(\".\")[0] + \".wav\"\n", " convert_audio(FILE_PATH, tmp_vad_wav_fp, sample_rate=SAMPLE_RATE)\n", " audio_bytes = read_safe_wav(tmp_vad_wav_fp, sample_rate=SAMPLE_RATE)\n", " audio_dur_s = get_audio_dur_s(tmp_vad_wav_fp)\n", " # TODO: use code from utils.notebook to do this as array instead of temp file\n", " # TODO: create higher sample_rate bytes for slicing if available\n", "\n", " \n", "# find pauses\n", "gap_intervals = _get_vad_gaps(audio_bytes)" ] }, { "cell_type": "code", "execution_count": 5, "id": "fa768fda", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "[5.3,5.78)" ] }, "execution_count": 5, "metadata": {}, "output_type": "execute_result" } ], "source": [ "gap_intervals[0]" ] }, { "cell_type": "code", "execution_count": 6, "id": "39ca6b92", "metadata": {}, "outputs": [], "source": [ "# get valid slice intervals\n", "min_dur_s = 2.0\n", "max_dur_s = 15.0\n", "slice_intervals = []\n", "offs_s = gap_intervals.lower\n", "for _ in range(int(audio_dur_s // min_dur_s) + 1):\n", " if audio_dur_s - offs_s <= min_dur_s:\n", " break\n", " gaps_found = gap_intervals & portion.closedopen(offs_s + min_dur_s, offs_s + max_dur_s)\n", " if gaps_found.empty:\n", " remaining_interval = (gap_intervals & portion.closedopen(offs_s + min_dur_s, np.inf))\n", " if not remaining_interval.empty:\n", " offs_s = remaining_interval.lower\n", " continue\n", " break\n", " slice_intervals.append((offs_s, gaps_found.upper))\n", " offs_s = gaps_found.upper\n", " \n", "\n", "# create byte slice objects\n", "slices = []\n", "for start_s, end_s in slice_intervals:\n", " slice_audio_bytes = get_audio_slice(\n", " audio_bytes, from_s=start_s, to_s=end_s, sample_rate=SAMPLE_RATE\n", " )\n", " slices.append((slice_audio_bytes, (start_s, end_s)))" ] }, { "cell_type": "code", "execution_count": 7, "id": "ee18c012", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "bytes" ] }, "execution_count": 7, "metadata": {}, "output_type": "execute_result" } ], "source": [ "type(slices[0][0])" ] }, { "cell_type": "code", "execution_count": 8, "id": "c9efbd19", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "(5.3, 18.36)" ] }, "execution_count": 8, "metadata": {}, "output_type": "execute_result" } ], "source": [ "slices[0][1]" ] }, { "cell_type": "code", "execution_count": 9, "id": "8bbe6038", "metadata": {}, "outputs": [], "source": [ "from suno_utils.utils.notebook import play_audio" ] }, { "cell_type": "code", "execution_count": 10, "id": "71c2a224", "metadata": {}, "outputs": [ { "data": { "text/html": [ "\n", " \n", " " ], "text/plain": [ "" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "play_audio(slices[55][0])" ] }, { "cell_type": "code", "execution_count": 11, "id": "6af74234", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "" ] }, "execution_count": 11, "metadata": {}, "output_type": "execute_result" }, { "data": { "image/png": "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", "text/plain": [ "
" ] }, "metadata": { "needs_background": "light" }, "output_type": "display_data" } ], "source": [ "audio_lens = [s[1][1] - s[1][0] for s in slices]\n", "audio_lens = pd.Series(audio_lens)\n", "audio_lens.hist(bins=30)" ] }, { "cell_type": "code", "execution_count": 43, "id": "425eb731", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "0.9298838920504986" ] }, "execution_count": 43, "metadata": {}, "output_type": "execute_result" } ], "source": [ "audio_lens.sum() / audio_dur_s" ] }, { "cell_type": "code", "execution_count": 13, "id": "3bed459b", "metadata": {}, "outputs": [], "source": [ "data_dir = \"/home/ubuntu/code/glockenspiel/audio-labeler/data/slices2\"\n", "all_files = []\n", "for bts, interval in slices:\n", " st, ed = interval\n", " fname = \"{0:08.2f}__to__{0:08.2f}\".format(st, ed).replace(\".\", \"-\") + \".wav\"\n", " fpath = os.path.join(data_dir, fname)\n", " with open(fpath, \"wb\") as fi:\n", " write_wav(fi, bts)\n", " all_files.append(fpath)" ] }, { "cell_type": "code", "execution_count": 44, "id": "d827ea04", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "176" ] }, "execution_count": 44, "metadata": {}, "output_type": "execute_result" } ], "source": [ "len(all_files)" ] }, { "cell_type": "code", "execution_count": 15, "id": "f77e9d70", "metadata": {}, "outputs": [ { "data": { "text/plain": [ "'/home/ubuntu/code/glockenspiel/audio-labeler/data/slices2/01142-64__to__01142-64.wav'" ] }, "execution_count": 15, "metadata": {}, "output_type": "execute_result" } ], "source": [ "all_files[81]" ] }, { "cell_type": "code", "execution_count": 16, "id": "18f3b201", "metadata": {}, "outputs": [ { "name": "stderr", "output_type": "stream", "text": [ "[NeMo W 2022-03-22 22:59:31 optimizers:55] Apex was not found. Using the lamb or fused_adam optimizer will error out.\n", "################################################################################\n", "### WARNING, path does not exist: KALDI_ROOT=/mnt/matylda5/iveselyk/Tools/kaldi-trunk\n", "### (please add 'export KALDI_ROOT=' in your $HOME/.profile)\n", "### (or run as: KALDI_ROOT= python .py)\n", "################################################################################\n", "\n" ] } ], "source": [ "\n", "\n", "import nemo\n", "import nemo.collections.asr as nemo_asr\n", "import tqdm" ] }, { "cell_type": "code", "execution_count": 31, "id": "8f10bc31", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "[NeMo I 2022-03-22 23:27:40 cloud:66] Downloading from: https://api.ngc.nvidia.com/v2/models/nvidia/nemo/stt_en_conformer_ctc_large/versions/1.6.0/files/stt_en_conformer_ctc_large.nemo to /home/ubuntu/.cache/torch/NeMo/NeMo_1.7.2/stt_en_conformer_ctc_large/010120d9959425c7862c9843960b3235/stt_en_conformer_ctc_large.nemo\n", "[NeMo I 2022-03-22 23:28:02 common:704] Instantiating model from pre-trained checkpoint\n", "[NeMo I 2022-03-22 23:28:05 mixins:146] Tokenizer SentencePieceTokenizer initialized with 128 tokens\n" ] }, { "name": "stderr", "output_type": "stream", "text": [ "[NeMo W 2022-03-22 23:28:05 modelPT:148] If you intend to do training or fine-tuning, please call the ModelPT.setup_training_data() method and provide a valid configuration file to setup the train data loader.\n", " Train config : \n", " manifest_filepath: /data/NeMo_ASR_SET/English/v2.0/train/tarred_audio_manifest.json\n", " sample_rate: 16000\n", " batch_size: 32\n", " shuffle: true\n", " num_workers: 8\n", " pin_memory: true\n", " use_start_end_token: false\n", " trim_silence: false\n", " max_duration: 20.0\n", " min_duration: 0.1\n", " shuffle_n: 2048\n", " is_tarred: true\n", " tarred_audio_filepaths: /data/NeMo_ASR_SET/English/v2.0/train/audio__OP_0..4095_CL_.tar\n", " \n", "[NeMo W 2022-03-22 23:28:05 modelPT:155] If you intend to do validation, please call the ModelPT.setup_validation_data() or ModelPT.setup_multiple_validation_data() method and provide a valid configuration file to setup the validation data loader(s). \n", " Validation config : \n", " manifest_filepath:\n", " - /data/ASR/LibriSpeech/librispeech_withsp2/manifests/librivox-dev-other.json\n", " - /data/ASR/LibriSpeech/librispeech_withsp2/manifests/librivox-dev-clean.json\n", " - /data/ASR/LibriSpeech/librispeech_withsp2/manifests/librivox-test-other.json\n", " - /data/ASR/LibriSpeech/librispeech_withsp2/manifests/librivox-test-clean.json\n", " sample_rate: 16000\n", " batch_size: 16\n", " shuffle: false\n", " num_workers: 8\n", " pin_memory: true\n", " use_start_end_token: false\n", " is_tarred: false\n", " tarred_audio_filepaths: na\n", " \n", "[NeMo W 2022-03-22 23:28:05 modelPT:161] Please call the ModelPT.setup_test_data() or ModelPT.setup_multiple_test_data() method and provide a valid configuration file to setup the test data loader(s).\n", " Test config : \n", " manifest_filepath:\n", " - /data/ASR/LibriSpeech/librispeech_withsp2/manifests/librivox-test-other.json\n", " - /data/ASR/LibriSpeech/librispeech_withsp2/manifests/librivox-dev-clean.json\n", " - /data/ASR/LibriSpeech/librispeech_withsp2/manifests/librivox-dev-other.json\n", " - /data/ASR/LibriSpeech/librispeech_withsp2/manifests/librivox-test-clean.json\n", " sample_rate: 16000\n", " batch_size: 16\n", " shuffle: false\n", " num_workers: 8\n", " pin_memory: true\n", " use_start_end_token: false\n", " is_tarred: false\n", " tarred_audio_filepaths: na\n", " \n" ] }, { "name": "stdout", "output_type": "stream", "text": [ "[NeMo I 2022-03-22 23:28:05 features:255] PADDING: 0\n", "[NeMo I 2022-03-22 23:28:05 features:272] STFT using torch\n", "[NeMo I 2022-03-22 23:28:06 save_restore_connector:157] Model EncDecCTCModelBPE was successfully restored from /home/ubuntu/.cache/torch/NeMo/NeMo_1.7.2/stt_en_conformer_ctc_large/010120d9959425c7862c9843960b3235/stt_en_conformer_ctc_large.nemo.\n" ] } ], "source": [ "asr_model = nemo_asr.models.EncDecCTCModelBPE.from_pretrained(model_name=\"stt_en_conformer_ctc_large\").cpu()" ] }, { "cell_type": "code", "execution_count": 40, "id": "96aefc30", "metadata": {}, "outputs": [ { "data": { "application/vnd.jupyter.widget-view+json": { "model_id": "e09e9731c3c8408a9df41ac6e833cc25", "version_major": 2, "version_minor": 0 }, "text/plain": [ "Transcribing: 0%| | 0/44 [00:00