{
"cells": [
{
"cell_type": "code",
"execution_count": 1,
"id": "ba198835",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Populating the interactive namespace from numpy and matplotlib\n"
]
}
],
"source": [
"%pylab inline"
]
},
{
"cell_type": "code",
"execution_count": 2,
"id": "95d8afc3",
"metadata": {},
"outputs": [],
"source": [
"import os\n",
"os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"\""
]
},
{
"cell_type": "code",
"execution_count": 4,
"id": "594ea11c",
"metadata": {},
"outputs": [],
"source": [
"import json\n",
"import scipy\n",
"import pandas as pd\n",
"\n",
"from suno_utils.utils.notebook import Audio\n",
"from suno_utils.tasks.asr import transcribe\n",
"from suno_utils.utils.numbers import interpolate_1d_array, safe_round\n",
"from suno_utils.utils.vad import predict_vad_speech"
]
},
{
"cell_type": "code",
"execution_count": 5,
"id": "89cf4e6f",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"audio = Audio.from_file(\"data/segment_1.mp3\")\n",
"audio.play()"
]
},
{
"cell_type": "code",
"execution_count": 6,
"id": "7abb2160",
"metadata": {},
"outputs": [],
"source": [
"with open(\"data/segment_1_annotation.json\") as f:\n",
" annotations_raw = json.load(f)\n",
"annotations = []\n",
"for e in annotations_raw[0][\"label\"]:\n",
" assert(len(e[\"labels\"]) == 1)\n",
" annotations.append((safe_round(e[\"start\"]), safe_round(e[\"end\"]), e[\"labels\"][0]))\n",
"annotations = sorted(annotations, key=lambda x: x[0])\n",
"# dedup\n",
"dedup_annots = []\n",
"prev_lang = None\n",
"for e in annotations:\n",
" lang = e[-1]\n",
" if lang == prev_lang:\n",
" dedup_annots[-1] = (dedup_annots[-1][0], e[1], lang[:2])\n",
" continue\n",
" dedup_annots.append((e[0], e[1], lang[:2]))\n",
" prev_lang = lang\n",
"# get boundaries\n",
"p_boundaries = []\n",
"for e1, e2 in zip(dedup_annots[:-1], dedup_annots[1:]):\n",
" p_boundaries.append(safe_round(np.mean([e1[1], e2[0]])))\n",
"# get text\n",
"p_text = []\n",
"for e in dedup_annots:\n",
" p_text.append((safe_round((e[1] + e[0]) / 2), e[-1]))"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "6764011b",
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "markdown",
"id": "c472f61b",
"metadata": {},
"source": [
"## Test using nemo"
]
},
{
"cell_type": "code",
"execution_count": 259,
"id": "d969a0d5",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"No GPU found even though requested, using CPU.\n",
"[NeMo W 2022-05-31 13:41:36 nemo_logging:349] /home/georg/venvs/ml/lib/python3.8/site-packages/torch/autocast_mode.py:162: UserWarning: User provided device_type of 'cuda', but CUDA is not available. Disabling\n",
" warnings.warn('User provided device_type of \\'cuda\\', but CUDA is not available. Disabling')\n",
" \n"
]
}
],
"source": [
"vocab, logits = transcribe(audio, mode=\"logits\")\n",
"probs = scipy.special.softmax(logits, axis=1)\n",
"frame_duration_ms = audio.duration_s / len(logits) * 1_000\n",
"vad_speech = predict_vad_speech(audio, frame_duration_ms=frame_duration_ms, vad_level=2)"
]
},
{
"cell_type": "code",
"execution_count": 260,
"id": "eabec775",
"metadata": {
"scrolled": false
},
"outputs": [],
"source": [
"# # average entries\n",
"# n_frames = int(len(probs) / 5)\n",
"# a = interpolate_1d_array(probs[:, -1], n_frames)\n",
"# b = interpolate_1d_array(vad_speech, n_frames)\n",
"# c = interpolate_1d_array(probs.max(axis=1), n_frames)\n",
"\n",
"# plt.figure(figsize=(12,3));\n",
"# pd.Series(a).rolling(1).mean().plot(label=\"vad speech\", alpha=0.8)\n",
"# pd.Series(b).rolling(1).mean().plot(label=\"asr speech\", alpha=0.8)\n",
"# pd.Series(c).rolling(1).mean().plot(label=\"asr confidence\", alpha=0.8)\n",
"# plt.legend(loc=\"lower center\")\n",
"# plt.show();"
]
},
{
"cell_type": "code",
"execution_count": 277,
"id": "c19cc8f5",
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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\n",
"text/plain": [
""
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"a = probs[:, :].max(axis=1).copy()\n",
"\n",
"n_smooth = 5\n",
"n_frames = int(len(a) / n_smooth)\n",
"a = interpolate_1d_array(a, n_frames)\n",
"\n",
"s = a\n",
"s = pd.Series(s).rolling(5, center=True).mean().values\n",
"\n",
"t = np.arange(0, audio.duration_s, audio.duration_s / len(s))\n",
"plt.figure(figsize=(12, 4))\n",
"plt.plot(t, s)\n",
"for e in p_boundaries:\n",
" plt.plot([e, e], [0, 1], \"--k\", alpha=0.3)\n",
"for e in p_text:\n",
" plt.text(e[0]-0.3, 0.05, e[1])\n",
"plt.xlabel(\"time (s)\")\n",
"plt.title(\"classify english vs not english\")\n",
"plt.show();"
]
},
{
"cell_type": "code",
"execution_count": 278,
"id": "a7ab4603",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"[]"
]
},
"execution_count": 278,
"metadata": {},
"output_type": "execute_result"
},
{
"data": {
"image/png": "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\n",
"text/plain": [
""
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"plt.plot(b)"
]
},
{
"cell_type": "code",
"execution_count": 271,
"id": "0b54bbbf",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"951"
]
},
"execution_count": 271,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"a = probs[:, :].max(axis=1).copy()\n",
"b = predict_vad_speech(audio, frame_duration_ms=frame_duration_ms, vad_level=3)\n",
"b = pd.Series(b).rolling(20, center=True, min_periods=1).mean().values\n",
"a[np.where(b <= 0.7)] = np.nan\n",
"\n",
"n_smooth = 5\n",
"n_frames = int(len(a) / n_smooth)\n",
"a = interpolate_1d_array(a, n_frames)\n",
"\n",
"s = a\n",
"s = pd.Series(s).rolling(5, center=True).mean().values\n",
"# s = interpolate_1d_array(s, n_frames)\n",
"\n",
"\n",
"t = np.arange(0, audio.duration_s, audio.duration_s / len(s))\n",
"plt.figure(figsize=(12, 4))\n",
"plt.plot(t, s)\n",
"for e in p_boundaries:\n",
" plt.plot([e, e], [0, 1], \"--k\", alpha=0.3)\n",
"for e in p_text:\n",
" plt.text(e[0]-0.3, 0.05, e[1])\n",
"plt.xlabel(\"time (s)\")\n",
"plt.title(\"classify english vs not english\")\n",
"plt.show();"
]
},
{
"cell_type": "code",
"execution_count": 274,
"id": "743c1b6e",
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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\n",
"text/plain": [
""
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"\n",
"\n",
"n_smooth = 5\n",
"n_frames = int(len(a) / n_smooth)\n",
"# b = interpolate_1d_array(b, n_frames)\n",
"\n",
"s = b#(a * b)\n",
"s = pd.Series(s).rolling(20, center=True, min_periods=1).mean().values\n",
"# s = interpolate_1d_array(s, n_frames)\n",
"\n",
"t = np.arange(0, audio.duration_s, audio.duration_s / len(s))\n",
"plt.figure(figsize=(12, 4))\n",
"plt.plot(t, s)\n",
"for e in p_boundaries:\n",
" plt.plot([e, e], [0, 1], \"--k\", alpha=0.3)\n",
"for e in p_text:\n",
" plt.text(e[0]-0.3, 0.05, e[1])\n",
"plt.xlabel(\"time (s)\")\n",
"plt.title(\"classify english vs not english\")\n",
"plt.show();"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "463fbdb1",
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"id": "cd0d539a",
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "markdown",
"id": "f336ca2b",
"metadata": {},
"source": [
"## try multilingual asr"
]
},
{
"cell_type": "code",
"execution_count": 10,
"id": "5695db11",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"audio = Audio.from_file(\"data/segment_1.mp3\")\n",
"audio_array = audio.convert(16_000, 2, 1).array.astype(np.float32) / np.iinfo(np.int16).max\n",
"audio.play()"
]
},
{
"cell_type": "markdown",
"id": "8ed64d89",
"metadata": {},
"source": [
"#### https://huggingface.co/voidful/wav2vec2-xlsr-multilingual-56"
]
},
{
"cell_type": "code",
"execution_count": 141,
"id": "631e8d1d",
"metadata": {},
"outputs": [],
"source": [
"import torchaudio\n",
"from datasets import load_dataset, load_metric\n",
"from transformers import (\n",
" Wav2Vec2ForCTC,\n",
" Wav2Vec2Processor,\n",
" AutoTokenizer, \n",
" AutoModelWithLMHead \n",
")\n",
"import torch\n",
"import re\n",
"import sys\n",
"import soundfile as sf\n",
"\n",
"model_name = \"voidful/wav2vec2-xlsr-multilingual-56\"\n",
"processor_name = \"voidful/wav2vec2-xlsr-multilingual-56\"\n",
"\n",
"import pickle\n",
"with open(\"lang_ids.pk\", 'rb') as output:\n",
" lang_ids = pickle.load(output)\n",
" \n",
"model = Wav2Vec2ForCTC.from_pretrained(model_name)\n",
"processor = Wav2Vec2Processor.from_pretrained(processor_name)\n",
"model.eval();"
]
},
{
"cell_type": "code",
"execution_count": 170,
"id": "7a0ed8a3",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"['heloma habechang yes i am back comsaing lahak wofulen behave ce youhaben wala ahadita anoveladahilinitao o mentaly and financialy ahinaman i tokabra for acopoe weks antiryamback and ino figting anyway alamunemana tis segment here on the chanel is not about ne so hirman omusainlad its me yorgetactive martin rules and wakanbacktmachao']"
]
},
"execution_count": 170,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"def predict(audio_array):\n",
" features = processor(audio_array, sampling_rate=16_000, padding=True, return_tensors=\"pt\")\n",
" input_values = features.input_values.to(device)\n",
" attention_mask = features.attention_mask.to(device)\n",
" with torch.no_grad():\n",
" logits = model(input_values, attention_mask=attention_mask).logits\n",
" decoded_results = []\n",
" for logit in logits:\n",
" pred_ids = torch.argmax(logit, dim=-1)\n",
" mask = pred_ids.ge(1).unsqueeze(-1).expand(logit.size())\n",
" vocab_size = logit.size()[-1]\n",
" voice_prob = torch.nn.functional.softmax((torch.masked_select(logit, mask).view(-1,vocab_size)),dim=-1)\n",
" comb_pred_ids = torch.argmax(voice_prob, dim=-1)\n",
" decoded_results.append(processor.decode(comb_pred_ids))\n",
" return logits, decoded_results\n",
"\n",
"def predict_lang_specific(audio_array, lang_code=\"en\"):\n",
" features = processor(audio_array, sampling_rate=16_000, padding=True, return_tensors=\"pt\")\n",
" input_values = features.input_values.to(device)\n",
" attention_mask = features.attention_mask.to(device)\n",
" with torch.no_grad():\n",
" logits = model(input_values, attention_mask=attention_mask).logits\n",
" decoded_results = []\n",
" for logit in logits:\n",
" pred_ids = torch.argmax(logit, dim=-1)\n",
" mask = ~pred_ids.eq(processor.tokenizer.pad_token_id).unsqueeze(-1).expand(logit.size())\n",
" vocab_size = logit.size()[-1]\n",
" voice_prob = torch.nn.functional.softmax((torch.masked_select(logit, mask).view(-1,vocab_size)),dim=-1)\n",
" filtered_input = pred_ids[pred_ids!=processor.tokenizer.pad_token_id].view(1,-1).to(device)\n",
" if len(filtered_input[0]) == 0:\n",
" decoded_results.append(\"\")\n",
" else:\n",
" lang_mask = torch.empty(voice_prob.shape[-1]).fill_(0)\n",
" lang_index = torch.tensor(sorted(lang_ids[lang_code]))\n",
" lang_mask.index_fill_(0, lang_index, 1)\n",
" lang_mask = lang_mask.to(device)\n",
" comb_pred_ids = torch.argmax(lang_mask*voice_prob, dim=-1)\n",
" decoded_results.append(processor.decode(comb_pred_ids))\n",
" \n",
" return logits, decoded_results\n",
"\n",
"logits, out = predict_lang_specific(audio_array, lang_code=\"en\")\n",
"out"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "9df96596",
"metadata": {},
"outputs": [],
"source": [
"# conclusion: doesn't seem to be much different between \"en\" and \"de\" decoding"
]
},
{
"cell_type": "markdown",
"id": "922dcb19",
"metadata": {},
"source": [
"#### https://huggingface.co/facebook/wav2vec2-large-robust-ft-swbd-300h"
]
},
{
"cell_type": "code",
"execution_count": 116,
"id": "4880ce84",
"metadata": {},
"outputs": [],
"source": [
"from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC\n",
"import torch\n",
"\n",
"# load model and processor\n",
"processor = Wav2Vec2Processor.from_pretrained(\"facebook/wav2vec2-large-robust-ft-swbd-300h\")\n",
"model = Wav2Vec2ForCTC.from_pretrained(\"facebook/wav2vec2-large-robust-ft-swbd-300h\")"
]
},
{
"cell_type": "code",
"execution_count": 117,
"id": "86bc2de5",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"\"heloma habat chung yes i am back comes ehind lahat hopefully n behave bca you how long wie a aadita uh uh um hum i i no a at i iitao o i mentally and financially uh inaan i took a break for a couple of weeks and here i am i'm back and you know fitthing anyway i lome you to mind uh this segment here on my chi it's not about me so himan comes behind lahat it's me your gat active martin rus and well come back to my channel\""
]
},
"execution_count": 117,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# tokenize\n",
"input_values = processor(\n",
" audio_array, return_tensors=\"pt\", padding=\"longest\", sampling_rate=16_000\n",
").input_values # Batch size 1\n",
"\n",
"# retrieve logits\n",
"logits = model(input_values).logits\n",
"\n",
"# take argmax and decode\n",
"predicted_ids = torch.argmax(logits, dim=-1)\n",
"transcription = processor.batch_decode(predicted_ids)\n",
"transcription[0].lower()"
]
},
{
"cell_type": "code",
"execution_count": 123,
"id": "27131d0d",
"metadata": {},
"outputs": [],
"source": [
"from suno_utils.utils.notebook import _collapse_to_numpy_array\n",
"logits = _collapse_to_numpy_array(logits)\n",
"probs = scipy.special.softmax(logits, axis=1)"
]
},
{
"cell_type": "code",
"execution_count": 129,
"id": "d7147f50",
"metadata": {},
"outputs": [
{
"data": {
"image/png": "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\n",
"text/plain": [
""
]
},
"metadata": {
"needs_background": "light"
},
"output_type": "display_data"
}
],
"source": [
"# s = logits[:, -1]\n",
"# s = probs[:, -1]\n",
"a = probs[:, :].max(axis=1)\n",
"\n",
"n_smooth = 5\n",
"n_frames = int(len(a) / n_smooth)\n",
"\n",
"a = interpolate_1d_array(a, n_frames)\n",
"b = interpolate_1d_array(b, n_frames)\n",
"\n",
"s = a#(a * b)\n",
"s = pd.Series(s).rolling(5, center=True).mean().values\n",
"# s = interpolate_1d_array(s, n_frames)\n",
"\n",
"t = np.arange(0, audio.duration_s, audio.duration_s / len(s))\n",
"plt.figure(figsize=(12, 4))\n",
"plt.plot(t, s)\n",
"for e in p_boundaries:\n",
" plt.plot([e, e], [0, 1], \"--k\", alpha=0.3)\n",
"for e in p_text:\n",
" plt.text(e[0]-0.3, 0.05, e[1])\n",
"plt.xlabel(\"time (s)\")\n",
"plt.title(\"classify english vs not english\")\n",
"plt.show();"
]
},
{
"cell_type": "markdown",
"id": "e9fe2b15",
"metadata": {},
"source": [
"#### https://huggingface.co/facebook/wav2vec2-xlsr-53-espeak-cv-ft"
]
},
{
"cell_type": "code",
"execution_count": 251,
"id": "31c5d38a",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"The tokenizer class you load from this checkpoint is not the same type as the class this function is called from. It may result in unexpected tokenization. \n",
"The tokenizer class you load from this checkpoint is 'Wav2Vec2PhonemeCTCTokenizer'. \n",
"The class this function is called from is 'Wav2Vec2CTCTokenizer'.\n"
]
}
],
"source": [
"from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC\n",
"import torch\n",
"\n",
"# load model and processor\n",
"processor = Wav2Vec2Processor.from_pretrained(\"facebook/wav2vec2-xlsr-53-espeak-cv-ft\")\n",
"model = Wav2Vec2ForCTC.from_pretrained(\"facebook/wav2vec2-xlsr-53-espeak-cv-ft\")"
]
},
{
"cell_type": "code",
"execution_count": 252,
"id": "5f58cf7e",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"It is strongly recommended to pass the ``sampling_rate`` argument to this function. Failing to do so can result in silent errors that might be hard to debug.\n",
"[NeMo W 2022-05-31 13:34:54 nemo_logging:349] /home/georg/venvs/ml/lib/python3.8/site-packages/transformers/feature_extraction_utils.py:158: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at ../torch/csrc/utils/tensor_new.cpp:210.)\n",
" tensor = as_tensor(value)\n",
" \n"
]
},
{
"data": {
"text/plain": [
"['heloːmahabɛtʃhaŋjesaɪɛmbaːkkamustaxaɪnlahaːhoːfuliːnəɡbiheːbkʌjohabaŋwalaxɔdiːtɔɑyhypaɪjahonovɛladahilinitajɔɔraɪmɛmtəliːɛndfinanʃeliːaxaɪanamanaɪtokabɹeːkfɔakɔpləpwiːksɛnthiːɹaɪjɛmaɪmbakɛnnɔfaɪteŋɛniweːalamjunemaːnɛdissɛɡmɛnthiːɔnmetʃanlesnɑːtabaʊtniːsoːhaɪrmankomusthaɪlahdɛtsmiːjɔɡeːtɛktɪvmatɪnɹuːsɛndwɔlkəmbaktəmeːttʃaːnal']"
]
},
"execution_count": 252,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"# tokenize\n",
"input_values = processor(audio_array, return_tensors=\"pt\").input_values\n",
"\n",
"# retrieve logits\n",
"with torch.no_grad():\n",
" logits = model(input_values).logits\n",
"\n",
"# take argmax and decode\n",
"predicted_ids = torch.argmax(logits, dim=-1)\n",
"transcription = processor.batch_decode(predicted_ids)\n",
"transcription"
]
},
{
"cell_type": "code",
"execution_count": 253,
"id": "e8fd81d1",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"audio.play()"
]
},
{
"cell_type": "code",
"execution_count": 256,
"id": "e12de6a0",
"metadata": {},
"outputs": [],
"source": [
"from suno_utils.audio.conversion import _collapse_to_numpy_array\n",
"logits = _collapse_to_numpy_array(logits)\n",
"probs = scipy.special.softmax(logits, axis=1)"
]
},
{
"cell_type": "code",
"execution_count": 257,
"id": "fb7fe846",
"metadata": {},
"outputs": [
{
"data": {
"text/plain": [
"(1900, 392)"
]
},
"execution_count": 257,
"metadata": {},
"output_type": "execute_result"
}
],
"source": [
"probs.shape"
]
},
{
"cell_type": "code",
"execution_count": 218,
"id": "05c02c68",
"metadata": {},
"outputs": [],
"source": [
"# https://arxiv.org/pdf/2109.11680.pdf\n",
"# conclusion: not clear what to do with phoneme output"
]
},
{
"cell_type": "markdown",
"id": "82da37bd",
"metadata": {},
"source": [
"#### https://huggingface.co/speechbrain/lang-id-voxlingua107-ecapa"
]
},
{
"cell_type": "code",
"execution_count": 344,
"id": "f6fe841b",
"metadata": {},
"outputs": [],
"source": [
"from speechbrain.pretrained import EncoderClassifier\n",
"\n",
"language_id = EncoderClassifier.from_hparams(source=\"speechbrain/lang-id-voxlingua107-ecapa\", savedir=\"sb_model_1\")"
]
},
{
"cell_type": "code",
"execution_count": 345,
"id": "0a072a08",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"tensor([-16.7989, -8.9914, -13.4097]) tensor([-0.1595]) tensor([96]) ['tl: Tagalog']\n"
]
}
],
"source": [
"logits, a, b, lang_pred = language_id.classify_batch(torch.Tensor(audio_array))\n",
"print(logits[0,:3], a, b, lang_pred)"
]
},
{
"cell_type": "code",
"execution_count": 346,
"id": "2f7e7922",
"metadata": {},
"outputs": [],
"source": [
"audio = Audio.from_file(\"data/segment_1.mp3\").convert(16_000, 2, 1)\n",
"array_segments = [] \n",
"for s, e in zip([0] + p_boundaries, p_boundaries + [audio.duration_s]):\n",
" array_segment = audio.get_slice(s, e).array.astype(np.float32) / np.iinfo(np.int16).max\n",
" array_segments.append(torch.Tensor(array_segment))"
]
},
{
"cell_type": "code",
"execution_count": 347,
"id": "0a762f69",
"metadata": {},
"outputs": [],
"source": [
"# Audio.from_array_float(array_segments[1], 16_000).play()"
]
},
{
"cell_type": "code",
"execution_count": 348,
"id": "95e96bc7",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"En -- max: 0.443 ['la: Latin'] -- p_En: 0.0 -- p_Fi: 0.004\n",
"Fi -- max: 0.472 ['ko: Korean'] -- p_En: 0.0 -- p_Fi: 0.005\n",
"En -- max: 0.491 ['tl: Tagalog'] -- p_En: 0.073 -- p_Fi: 0.491\n",
"Fi -- max: 0.916 ['ar: Arabic'] -- p_En: 0.0 -- p_Fi: 0.032\n",
"En -- max: 0.304 ['lo: Lao'] -- p_En: 0.0 -- p_Fi: 0.003\n",
"Fi -- max: 0.973 ['tl: Tagalog'] -- p_En: 0.0 -- p_Fi: 0.973\n",
"En -- max: 0.472 ['tl: Tagalog'] -- p_En: 0.008 -- p_Fi: 0.472\n",
"Fi -- max: 0.647 ['tk: Turkmen'] -- p_En: 0.001 -- p_Fi: 0.018\n",
"En -- max: 0.488 ['tl: Tagalog'] -- p_En: 0.315 -- p_Fi: 0.488\n",
"Fi -- max: 0.391 ['yi: Yiddish'] -- p_En: 0.037 -- p_Fi: 0.029\n",
"En -- max: 0.25 ['tl: Tagalog'] -- p_En: 0.148 -- p_Fi: 0.25\n",
"Fi -- max: 0.958 ['tl: Tagalog'] -- p_En: 0.0 -- p_Fi: 0.958\n",
"En -- max: 0.656 ['en: English'] -- p_En: 0.656 -- p_Fi: 0.049\n"
]
}
],
"source": [
"from suno_utils.audio.conversion import _collapse_to_numpy_array\n",
"\n",
"assert(len(p_text) == len(array_segments))\n",
"\n",
"for (_, lang), array_segment in zip(p_text, array_segments):\n",
" logits, a, b, lang_pred = language_id.classify_batch(array_segment)\n",
" logits = _collapse_to_numpy_array(logits)\n",
" probs = scipy.special.softmax(logits, axis=0)\n",
" print(lang, \"--\", \"max:\", round(probs[b], 3), lang_pred, \"--\", \"p_En:\", round(probs[20], 3), \"--\",\"p_Fi:\", round(probs[96], 3))"
]
},
{
"cell_type": "markdown",
"id": "cd4e515e",
"metadata": {},
"source": [
"#### https://huggingface.co/speechbrain/lang-id-commonlanguage_ecapa"
]
},
{
"cell_type": "code",
"execution_count": 64,
"id": "1062c074",
"metadata": {},
"outputs": [
{
"data": {
"application/vnd.jupyter.widget-view+json": {
"model_id": "",
"version_major": 2,
"version_minor": 0
},
"text/plain": [
"Downloading: 0%| | 0.00/1.67k [00:00 probs[96] else \"p-Fi\"\n",
" print(lang, o)\n",
"# print(lang, \"--\", \"max:\", round(probs[b], 3), lang_pred, \"--\", \"p_En:\", round(probs[20], 3), \"--\",\"p_Fi:\", round(probs[96], 3))"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "9ab46bde",
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"id": "4867f12b",
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "markdown",
"id": "b9294b22",
"metadata": {},
"source": [
"## Playground"
]
},
{
"cell_type": "code",
"execution_count": 7,
"id": "854ad64b",
"metadata": {},
"outputs": [],
"source": [
"import random\n",
"import os\n",
"import torch\n",
"import scipy\n",
"from suno_utils.utils.notebook import Audio\n",
"from suno_utils.audio.conversion import get_audio_properties\n",
"\n",
"VOXLINGUA_DIR = \"/mnt/data-ssd-1/data/voxlingua107/\""
]
},
{
"cell_type": "code",
"execution_count": 111,
"id": "2b70ebc4",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"15565 english samples\n",
"37241 tagalog samples\n"
]
}
],
"source": [
"d = VOXLINGUA_DIR + \"train/en/\"\n",
"en_filepaths = [d + fn for fn in os.listdir(d)]\n",
"print(len(en_filepaths), \"english samples\")\n",
"\n",
"d = VOXLINGUA_DIR + \"train/tl/\"\n",
"tl_filepaths = [d + fn for fn in os.listdir(d)]\n",
"print(len(tl_filepaths), \"tagalog samples\")"
]
},
{
"cell_type": "code",
"execution_count": 9,
"id": "8d9f6858",
"metadata": {},
"outputs": [],
"source": [
"# en_details_df = pd.DataFrame([get_audio_properties(fp) for fp in en_filepaths])\n",
"# tl_details_df = pd.DataFrame([get_audio_properties(fp) for fp in tl_filepaths])\n",
"\n",
"## ~2-20s slices\n",
"## 50h english, 90h tagalog"
]
},
{
"cell_type": "code",
"execution_count": 10,
"id": "1186ecb7",
"metadata": {},
"outputs": [],
"source": [
"en_sample_fps = en_filepaths.copy()\n",
"tl_sample_fps = tl_filepaths.copy()\n",
"random.seed(6006)\n",
"random.shuffle(en_sample_fps)\n",
"random.shuffle(tl_sample_fps)"
]
},
{
"cell_type": "code",
"execution_count": 11,
"id": "bf96e699",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"for fp in en_sample_fps[:5]:\n",
" Audio.from_file(fp).play()"
]
},
{
"cell_type": "code",
"execution_count": 12,
"id": "6531613f",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
}
],
"source": [
"for fp in tl_sample_fps[:5]:\n",
" Audio.from_file(fp).play()"
]
},
{
"cell_type": "code",
"execution_count": 13,
"id": "a3fca0d9",
"metadata": {},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"[NeMo W 2022-06-06 15:02:18 nemo_logging:349] /home/georg/venvs/ml/lib/python3.8/site-packages/huggingface_hub/utils/_deprecation.py:39: FutureWarning: Pass library_name=False as keyword args. From version 0.8 passing these as positional arguments will result in an error\n",
" warnings.warn(\n",
" \n"
]
}
],
"source": [
"from suno_utils.audio.conversion import _collapse_to_numpy_array\n",
"from speechbrain.pretrained import EncoderClassifier\n",
"\n",
"\n",
"clf_lang_1 = EncoderClassifier.from_hparams(\"speechbrain/lang-id-voxlingua107-ecapa\")\n",
"clf_lang_2 = EncoderClassifier.from_hparams(\"TalTechNLP/voxlingua107-epaca-tdnn\")"
]
},
{
"cell_type": "code",
"execution_count": 58,
"id": "14a5e2b3",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"English: 100.0% (m1), 97.6% (m2)\n",
"Tagalog: 0.0% (m1), 57.9% (m2)\n",
"\n",
"English: 100.0% (m1), 96.6% (m2)\n",
"Tagalog: 0.0% (m1), 54.0% (m2)\n",
"\n",
"English: 100.0% (m1), 98.7% (m2)\n",
"Tagalog: 0.0% (m1), 59.2% (m2)\n",
"\n",
"English: 100.0% (m1), 95.3% (m2)\n",
"Tagalog: 0.0% (m1), 55.4% (m2)\n",
"\n",
"English: 99.9% (m1), 97.4% (m2)\n",
"Tagalog: 0.0% (m1), 56.0% (m2)\n",
"\n"
]
}
],
"source": [
"for fp in en_sample_fps[:5]:\n",
" audio_arr = torch.Tensor(Audio.from_file(fp).convert(16_000, 2, 1).array_float)\n",
" logits_1, _, _, label_1 = clf_lang_1.classify_batch(audio_arr)\n",
" logits_1 = _collapse_to_numpy_array(logits_1)\n",
" probs_1 = np.exp(logits_1)\n",
"# probs_1 = scipy.special.softmax(logits_1, axis=0)\n",
" scores_2, _, _, label_2 = clf_lang_2.classify_batch(audio_arr)\n",
" scores_2 = _collapse_to_numpy_array(scores_2)\n",
" print(\"English: {}% (m1), {}% (m2)\".format(round(probs_1[20] * 100, 1), round(scores_2[20] * 100, 1)))\n",
" print(\"Tagalog: {}% (m1), {}% (m2)\".format(round(probs_1[96] * 100, 1), round(scores_2[96] * 100, 1)))\n",
" print()"
]
},
{
"cell_type": "code",
"execution_count": 59,
"id": "bdfb030d",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"English: 0.0% (m1), 42.7% (m2)\n",
"Tagalog: 100.0% (m1), 91.4% (m2)\n",
"\n",
"English: 0.0% (m1), 45.1% (m2)\n",
"Tagalog: 100.0% (m1), 92.2% (m2)\n",
"\n",
"English: 0.0% (m1), 40.6% (m2)\n",
"Tagalog: 100.0% (m1), 91.1% (m2)\n",
"\n",
"English: 0.0% (m1), 51.0% (m2)\n",
"Tagalog: 98.8% (m1), 92.9% (m2)\n",
"\n",
"English: 0.0% (m1), 44.1% (m2)\n",
"Tagalog: 99.8% (m1), 92.0% (m2)\n",
"\n"
]
}
],
"source": [
"for fp in tl_sample_fps[:5]:\n",
" audio_arr = torch.Tensor(Audio.from_file(fp).convert(16_000, 2, 1).array_float)\n",
" logits_1, _, _, label_1 = clf_lang_1.classify_batch(audio_arr)\n",
" logits_1 = _collapse_to_numpy_array(logits_1)\n",
"# probs_1 = scipy.special.softmax(logits_1, axis=0)\n",
" probs_1 = np.exp(logits_1)\n",
" scores_2, _, _, label_2 = clf_lang_2.classify_batch(audio_arr)\n",
" scores_2 = _collapse_to_numpy_array(scores_2)\n",
" print(\"English: {}% (m1), {}% (m2)\".format(round(probs_1[20] * 100, 1), round(scores_2[20] * 100, 1)))\n",
" print(\"Tagalog: {}% (m1), {}% (m2)\".format(round(probs_1[96] * 100, 1), round(scores_2[96] * 100, 1)))\n",
" print()"
]
},
{
"cell_type": "code",
"execution_count": 57,
"id": "04ffec0c",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"En\n",
"English: 0.0% (m1), 65.5% (m2)\n",
"Tagalog: 0.4% (m1), 67.7% (m2)\n",
"----------\n"
]
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Fi\n",
"English: 0.0% (m1), 48.3% (m2)\n",
"Tagalog: 0.5% (m1), 49.7% (m2)\n",
"----------\n"
]
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"En\n",
"English: 7.3% (m1), 86.7% (m2)\n",
"Tagalog: 49.1% (m1), 75.4% (m2)\n",
"----------\n"
]
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Fi\n",
"English: 0.0% (m1), 67.6% (m2)\n",
"Tagalog: 3.2% (m1), 73.6% (m2)\n",
"----------\n"
]
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"En\n",
"English: 0.0% (m1), 61.0% (m2)\n",
"Tagalog: 0.3% (m1), 60.5% (m2)\n",
"----------\n"
]
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Fi\n",
"English: 0.0% (m1), 49.2% (m2)\n",
"Tagalog: 97.3% (m1), 92.2% (m2)\n",
"----------\n"
]
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"En\n",
"English: 0.8% (m1), 61.8% (m2)\n",
"Tagalog: 47.2% (m1), 63.8% (m2)\n",
"----------\n"
]
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Fi\n",
"English: 0.1% (m1), 75.0% (m2)\n",
"Tagalog: 1.8% (m1), 76.6% (m2)\n",
"----------\n"
]
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"En\n",
"English: 31.5% (m1), 61.9% (m2)\n",
"Tagalog: 48.8% (m1), 89.0% (m2)\n",
"----------\n"
]
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Fi\n",
"English: 3.7% (m1), 76.5% (m2)\n",
"Tagalog: 2.9% (m1), 70.1% (m2)\n",
"----------\n"
]
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"En\n",
"English: 14.8% (m1), 75.9% (m2)\n",
"Tagalog: 25.0% (m1), 70.7% (m2)\n",
"----------\n"
]
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Fi\n",
"English: 0.0% (m1), 59.4% (m2)\n",
"Tagalog: 95.8% (m1), 84.4% (m2)\n",
"----------\n"
]
},
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"En\n",
"English: 65.6% (m1), 77.6% (m2)\n",
"Tagalog: 4.9% (m1), 68.1% (m2)\n",
"----------\n"
]
}
],
"source": [
"audio = Audio.from_file(\"data/segment_1.mp3\").convert(16_000, 2, 1)\n",
"array_segments = [] \n",
"for s, e, (_, true_lang) in zip([0] + p_boundaries, p_boundaries + [audio.duration_s], p_text):\n",
" a = audio.get_slice(s, e)\n",
" a.play()\n",
" audio_arr = torch.Tensor(a.array_float)\n",
" logits_1, _, _, label_1 = clf_lang_1.classify_batch(audio_arr)\n",
" logits_1 = _collapse_to_numpy_array(logits_1)\n",
" probs_1 = np.exp(logits_1)\n",
"# probs_1 = scipy.special.softmax(logits_1, axis=0)\n",
" scores_2, _, _, label_2 = clf_lang_2.classify_batch(audio_arr)\n",
" scores_2 = _collapse_to_numpy_array(scores_2)\n",
" print(true_lang)\n",
" print(\"English: {}% (m1), {}% (m2)\".format(round(probs_1[20] * 100, 1), round(scores_2[20] * 100, 1)))\n",
" print(\"Tagalog: {}% (m1), {}% (m2)\".format(round(probs_1[96] * 100, 1), round(scores_2[96] * 100, 1)))\n",
" print(\"-\"*10)"
]
},
{
"cell_type": "code",
"execution_count": 23,
"id": "b73b1a95",
"metadata": {},
"outputs": [
{
"data": {
"text/html": [
"\n",
" \n",
" "
],
"text/plain": [
""
]
},
"metadata": {},
"output_type": "display_data"
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"English: 82.9% (m1), 63.5% (m2)\n",
"Tagalog: 2.3% (m1), 79.0% (m2)\n"
]
}
],
"source": [
"a = Audio.from_file(\"hard_en.mp3\").convert(16_000, 2, 1)\n",
"a.play()\n",
"audio_arr = torch.Tensor(a.array_float)\n",
"logits_1, _, _, label_1 = clf_lang_1.classify_batch(torch.Tensor(audio_arr))\n",
"logits_1 = _collapse_to_numpy_array(logits_1)\n",
"probs_1 = scipy.special.softmax(logits_1, axis=0)\n",
"scores_2, _, _, label_2 = clf_lang_2.classify_batch(torch.Tensor(audio_arr))\n",
"scores_2 = _collapse_to_numpy_array(scores_2)\n",
"print(\"English: {}% (m1), {}% (m2)\".format(round(probs_1[20] * 100, 1), round(scores_2[20] * 100, 1)))\n",
"print(\"Tagalog: {}% (m1), {}% (m2)\".format(round(probs_1[96] * 100, 1), round(scores_2[96] * 100, 1)))"
]
},
{
"cell_type": "code",
"execution_count": 31,
"id": "ae671a8a",
"metadata": {},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"English: 83.0% (m1), 63.5% (m2)\n",
"Tagalog: 2.3% (m1), 79.0% (m2)\n"
]
}
],
"source": [
"audio_arr = clf_lang_1.load_audio(\"hard_en.mp3\")\n",
"logits_1, _, _, label_1 = clf_lang_1.classify_batch(audio_arr)\n",
"logits_1 = _collapse_to_numpy_array(logits_1)\n",
"probs_1 = scipy.special.softmax(logits_1, axis=0)\n",
"audio_arr = clf_lang_2.load_audio(\"hard_en.mp3\")\n",
"scores_2, _, _, label_2 = clf_lang_2.classify_batch(audio_arr)\n",
"scores_2 = _collapse_to_numpy_array(scores_2)\n",
"print(\"English: {}% (m1), {}% (m2)\".format(round(probs_1[20] * 100, 1), round(scores_2[20] * 100, 1)))\n",
"print(\"Tagalog: {}% (m1), {}% (m2)\".format(round(probs_1[96] * 100, 1), round(scores_2[96] * 100, 1)))"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "c6bd7a52",
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"id": "6569d60e",
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"id": "cc19ea2d",
"metadata": {},
"outputs": [],
"source": []
},
{
"cell_type": "code",
"execution_count": null,
"id": "fe5fd72e",
"metadata": {},
"outputs": [],
"source": []
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3 (ipykernel)",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.8.10"
}
},
"nbformat": 4,
"nbformat_minor": 5
}