File size: 21,322 Bytes
9742bb8 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 |
{
"nbformat": 4,
"nbformat_minor": 0,
"metadata": {
"colab": {
"provenance": [],
"gpuType": "T4"
},
"kernelspec": {
"name": "python3",
"display_name": "Python 3"
},
"language_info": {
"name": "python"
},
"accelerator": "GPU",
"gpuClass": "standard"
},
"cells": [
{
"cell_type": "markdown",
"source": [
"# Running MMS-LID inference in Colab"
],
"metadata": {
"id": "Rhm7khm6GskV"
}
},
{
"cell_type": "markdown",
"source": [
"## Step 1: Clone fairseq-py and install latest version"
],
"metadata": {
"id": "2GfxksHDGyJv"
}
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "Cj2x80SegRzr",
"outputId": "c81e367d-ec5f-4b17-b375-6980d6291c43"
},
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"fatal: destination path 'fairseq' already exists and is not an empty directory.\n",
"/content\n",
"/content/fairseq\n",
"Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
"Obtaining file:///content/fairseq\n",
" Installing build dependencies ... \u001b[?25l\u001b[?25hdone\n",
" Checking if build backend supports build_editable ... \u001b[?25l\u001b[?25hdone\n",
" Getting requirements to build editable ... \u001b[?25l\u001b[?25hdone\n",
" Preparing editable metadata (pyproject.toml) ... \u001b[?25l\u001b[?25hdone\n",
"Requirement already satisfied: cffi in /usr/local/lib/python3.10/dist-packages (from fairseq==0.12.2) (1.15.1)\n",
"Requirement already satisfied: cython in /usr/local/lib/python3.10/dist-packages (from fairseq==0.12.2) (0.29.34)\n",
"Requirement already satisfied: hydra-core<1.1,>=1.0.7 in /usr/local/lib/python3.10/dist-packages (from fairseq==0.12.2) (1.0.7)\n",
"Requirement already satisfied: omegaconf<2.1 in /usr/local/lib/python3.10/dist-packages (from fairseq==0.12.2) (2.0.6)\n",
"Requirement already satisfied: numpy>=1.21.3 in /usr/local/lib/python3.10/dist-packages (from fairseq==0.12.2) (1.22.4)\n",
"Requirement already satisfied: regex in /usr/local/lib/python3.10/dist-packages (from fairseq==0.12.2) (2022.10.31)\n",
"Requirement already satisfied: sacrebleu>=1.4.12 in /usr/local/lib/python3.10/dist-packages (from fairseq==0.12.2) (2.3.1)\n",
"Requirement already satisfied: torch>=1.13 in /usr/local/lib/python3.10/dist-packages (from fairseq==0.12.2) (2.0.1+cu118)\n",
"Requirement already satisfied: tqdm in /usr/local/lib/python3.10/dist-packages (from fairseq==0.12.2) (4.65.0)\n",
"Requirement already satisfied: bitarray in /usr/local/lib/python3.10/dist-packages (from fairseq==0.12.2) (2.7.3)\n",
"Requirement already satisfied: torchaudio>=0.8.0 in /usr/local/lib/python3.10/dist-packages (from fairseq==0.12.2) (2.0.2+cu118)\n",
"Requirement already satisfied: scikit-learn in /usr/local/lib/python3.10/dist-packages (from fairseq==0.12.2) (1.2.2)\n",
"Requirement already satisfied: packaging in /usr/local/lib/python3.10/dist-packages (from fairseq==0.12.2) (23.1)\n",
"Requirement already satisfied: antlr4-python3-runtime==4.8 in /usr/local/lib/python3.10/dist-packages (from hydra-core<1.1,>=1.0.7->fairseq==0.12.2) (4.8)\n",
"Requirement already satisfied: PyYAML>=5.1.* in /usr/local/lib/python3.10/dist-packages (from omegaconf<2.1->fairseq==0.12.2) (6.0)\n",
"Requirement already satisfied: typing-extensions in /usr/local/lib/python3.10/dist-packages (from omegaconf<2.1->fairseq==0.12.2) (4.5.0)\n",
"Requirement already satisfied: portalocker in /usr/local/lib/python3.10/dist-packages (from sacrebleu>=1.4.12->fairseq==0.12.2) (2.7.0)\n",
"Requirement already satisfied: tabulate>=0.8.9 in /usr/local/lib/python3.10/dist-packages (from sacrebleu>=1.4.12->fairseq==0.12.2) (0.8.10)\n",
"Requirement already satisfied: colorama in /usr/local/lib/python3.10/dist-packages (from sacrebleu>=1.4.12->fairseq==0.12.2) (0.4.6)\n",
"Requirement already satisfied: lxml in /usr/local/lib/python3.10/dist-packages (from sacrebleu>=1.4.12->fairseq==0.12.2) (4.9.2)\n",
"Requirement already satisfied: filelock in /usr/local/lib/python3.10/dist-packages (from torch>=1.13->fairseq==0.12.2) (3.12.0)\n",
"Requirement already satisfied: sympy in /usr/local/lib/python3.10/dist-packages (from torch>=1.13->fairseq==0.12.2) (1.11.1)\n",
"Requirement already satisfied: networkx in /usr/local/lib/python3.10/dist-packages (from torch>=1.13->fairseq==0.12.2) (3.1)\n",
"Requirement already satisfied: jinja2 in /usr/local/lib/python3.10/dist-packages (from torch>=1.13->fairseq==0.12.2) (3.1.2)\n",
"Requirement already satisfied: triton==2.0.0 in /usr/local/lib/python3.10/dist-packages (from torch>=1.13->fairseq==0.12.2) (2.0.0)\n",
"Requirement already satisfied: cmake in /usr/local/lib/python3.10/dist-packages (from triton==2.0.0->torch>=1.13->fairseq==0.12.2) (3.25.2)\n",
"Requirement already satisfied: lit in /usr/local/lib/python3.10/dist-packages (from triton==2.0.0->torch>=1.13->fairseq==0.12.2) (16.0.5)\n",
"Requirement already satisfied: pycparser in /usr/local/lib/python3.10/dist-packages (from cffi->fairseq==0.12.2) (2.21)\n",
"Requirement already satisfied: scipy>=1.3.2 in /usr/local/lib/python3.10/dist-packages (from scikit-learn->fairseq==0.12.2) (1.10.1)\n",
"Requirement already satisfied: joblib>=1.1.1 in /usr/local/lib/python3.10/dist-packages (from scikit-learn->fairseq==0.12.2) (1.2.0)\n",
"Requirement already satisfied: threadpoolctl>=2.0.0 in /usr/local/lib/python3.10/dist-packages (from scikit-learn->fairseq==0.12.2) (3.1.0)\n",
"Requirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.10/dist-packages (from jinja2->torch>=1.13->fairseq==0.12.2) (2.1.2)\n",
"Requirement already satisfied: mpmath>=0.19 in /usr/local/lib/python3.10/dist-packages (from sympy->torch>=1.13->fairseq==0.12.2) (1.3.0)\n",
"Building wheels for collected packages: fairseq\n",
" Building editable for fairseq (pyproject.toml) ... \u001b[?25l\u001b[?25hdone\n",
" Created wheel for fairseq: filename=fairseq-0.12.2-0.editable-cp310-cp310-linux_x86_64.whl size=9219 sha256=b6289e3715902d34fd7c54490679210a5be155dd4416754f0e8c376f193b5ac4\n",
" Stored in directory: /tmp/pip-ephem-wheel-cache-o62sj_ry/wheels/c6/d7/db/bc419b1daa8266aa8de2a7c4d29f62dbfa814e8701fe4695a2\n",
"Successfully built fairseq\n",
"Installing collected packages: fairseq\n",
" Attempting uninstall: fairseq\n",
" Found existing installation: fairseq 0.12.2\n",
" Uninstalling fairseq-0.12.2:\n",
" Successfully uninstalled fairseq-0.12.2\n",
"Successfully installed fairseq-0.12.2\n",
"Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
"Requirement already satisfied: tensorboardX in /usr/local/lib/python3.10/dist-packages (2.6)\n",
"Requirement already satisfied: numpy in /usr/local/lib/python3.10/dist-packages (from tensorboardX) (1.22.4)\n",
"Requirement already satisfied: packaging in /usr/local/lib/python3.10/dist-packages (from tensorboardX) (23.1)\n",
"Requirement already satisfied: protobuf<4,>=3.8.0 in /usr/local/lib/python3.10/dist-packages (from tensorboardX) (3.20.3)\n"
]
}
],
"source": [
"import os\n",
"\n",
"!git clone https://github.com/pytorch/fairseq\n",
"\n",
"# Change current working directory\n",
"!pwd\n",
"%cd \"/content/fairseq\"\n",
"!pip install --editable ./ \n",
"!pip install tensorboardX\n"
]
},
{
"cell_type": "markdown",
"source": [
"## 2. Download MMS-LID model\n",
"\n"
],
"metadata": {
"id": "cyk4JvZOHSw3"
}
},
{
"cell_type": "code",
"source": [
"available_models = [\"l126\", \"l256\", \"l512\", \"l1024\", \"l2048\", \"l4017\"]\n",
"\n",
"# We will use L126 model which can recognize 126 languages \n",
"model_name = available_models[0] # l126\n",
"print(f\"Using model - {model_name}\")\n",
"print(f\"Visit https://dl.fbaipublicfiles.com/mms/lid/mms1b_{model_name}_langs.html to check all the languages supported by this model.\")\n",
"\n",
"! mkdir -p /content/models_lid\n",
"!wget -P /content/models_lid/{model_name} 'https://dl.fbaipublicfiles.com/mms/lid/mms1b_{model_name}.pt'\n",
"!wget -P /content/models_lid/{model_name} 'https://dl.fbaipublicfiles.com/mms/lid/dict/l126/dict.lang.txt'\n",
"\n"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "3uZ9WG85gZId",
"outputId": "93f456ab-7aa1-47ac-a054-c0e3417b2e5e"
},
"execution_count": 5,
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"Using model - l126\n",
"Visit https://dl.fbaipublicfiles.com/mms/lid/mms1b_l126_langs.html to check all the languages supported by this model.\n",
"--2023-05-25 18:18:45-- https://dl.fbaipublicfiles.com/mms/lid/mms1b_l126.pt\n",
"Resolving dl.fbaipublicfiles.com (dl.fbaipublicfiles.com)... 52.84.251.15, 52.84.251.114, 52.84.251.27, ...\n",
"Connecting to dl.fbaipublicfiles.com (dl.fbaipublicfiles.com)|52.84.251.15|:443... connected.\n",
"HTTP request sent, awaiting response... 200 OK\n",
"Length: 3856229421 (3.6G) [binary/octet-stream]\n",
"Saving to: ‘/content/models_lid/l126/mms1b_l126.pt’\n",
"\n",
"mms1b_l126.pt 100%[===================>] 3.59G 198MB/s in 24s \n",
"\n",
"2023-05-25 18:19:09 (155 MB/s) - ‘/content/models_lid/l126/mms1b_l126.pt’ saved [3856229421/3856229421]\n",
"\n",
"--2023-05-25 18:19:09-- https://dl.fbaipublicfiles.com/mms/lid/dict/l126/dict.lang.txt\n",
"Resolving dl.fbaipublicfiles.com (dl.fbaipublicfiles.com)... 52.84.251.15, 52.84.251.114, 52.84.251.27, ...\n",
"Connecting to dl.fbaipublicfiles.com (dl.fbaipublicfiles.com)|52.84.251.15|:443... connected.\n",
"HTTP request sent, awaiting response... 200 OK\n",
"Length: 882 [text/plain]\n",
"Saving to: ‘/content/models_lid/l126/dict.lang.txt’\n",
"\n",
"dict.lang.txt 100%[===================>] 882 --.-KB/s in 0s \n",
"\n",
"2023-05-25 18:19:09 (183 MB/s) - ‘/content/models_lid/l126/dict.lang.txt’ saved [882/882]\n",
"\n"
]
}
]
},
{
"cell_type": "markdown",
"source": [
"## 3. Prepare manifest files\n",
"Create a folder on path '/content/audio_samples/' and upload your .wav audio files that you need to recognize e.g. '/content/audio_samples/abc.wav' , '/content/audio_samples/def.wav' etc...\n",
"\n",
"Note: You need to make sure that the audio data you are using has a sample rate of 16kHz You can easily do this with FFMPEG like the example below that converts .mp3 file to .flac and fixing the audio sample rate\n",
"\n",
"Here, we use three examples - one audio file from English, Hindi, Chinese each. "
],
"metadata": {
"id": "3p5-TQvKHXjO"
}
},
{
"cell_type": "code",
"source": [
"! mkdir -p /content/audio_samples/\n",
"for key in [\"en_us\", \"hi_in\", \"cmn_hans_cn\"]:\n",
" !wget -O /content/audio_samples/tmp.mp3 https://datasets-server.huggingface.co/assets/google/fleurs/--/{key}/train/0/audio/audio.mp3\n",
" !ffmpeg -hide_banner -loglevel error -y -i /content/audio_samples/tmp.mp3 -ar 16000 /content/audio_samples/{key}.wav\n",
"\n",
"! mkdir -p /content/audio_samples/\n"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "cnim4bokprbB",
"outputId": "89026a92-0518-49c2-9c84-98f0966caeac"
},
"execution_count": 6,
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"--2023-05-25 18:19:09-- https://datasets-server.huggingface.co/assets/google/fleurs/--/en_us/train/0/audio/audio.mp3\n",
"Resolving datasets-server.huggingface.co (datasets-server.huggingface.co)... 34.200.186.24, 44.197.252.161, 54.165.66.147, ...\n",
"Connecting to datasets-server.huggingface.co (datasets-server.huggingface.co)|34.200.186.24|:443... connected.\n",
"HTTP request sent, awaiting response... 200 OK\n",
"Length: 20853 (20K) [audio/mpeg]\n",
"Saving to: ‘/content/audio_samples/tmp.mp3’\n",
"\n",
"/content/audio_samp 100%[===================>] 20.36K 92.8KB/s in 0.2s \n",
"\n",
"2023-05-25 18:19:11 (92.8 KB/s) - ‘/content/audio_samples/tmp.mp3’ saved [20853/20853]\n",
"\n",
"--2023-05-25 18:19:12-- https://datasets-server.huggingface.co/assets/google/fleurs/--/hi_in/train/0/audio/audio.mp3\n",
"Resolving datasets-server.huggingface.co (datasets-server.huggingface.co)... 34.200.186.24, 44.197.252.161, 54.165.66.147, ...\n",
"Connecting to datasets-server.huggingface.co (datasets-server.huggingface.co)|34.200.186.24|:443... connected.\n",
"HTTP request sent, awaiting response... 200 OK\n",
"Length: 26361 (26K) [audio/mpeg]\n",
"Saving to: ‘/content/audio_samples/tmp.mp3’\n",
"\n",
"/content/audio_samp 100%[===================>] 25.74K 116KB/s in 0.2s \n",
"\n",
"2023-05-25 18:19:13 (116 KB/s) - ‘/content/audio_samples/tmp.mp3’ saved [26361/26361]\n",
"\n",
"--2023-05-25 18:19:13-- https://datasets-server.huggingface.co/assets/google/fleurs/--/cmn_hans_cn/train/0/audio/audio.mp3\n",
"Resolving datasets-server.huggingface.co (datasets-server.huggingface.co)... 34.200.186.24, 44.197.252.161, 54.165.66.147, ...\n",
"Connecting to datasets-server.huggingface.co (datasets-server.huggingface.co)|34.200.186.24|:443... connected.\n",
"HTTP request sent, awaiting response... 200 OK\n",
"Length: 23877 (23K) [audio/mpeg]\n",
"Saving to: ‘/content/audio_samples/tmp.mp3’\n",
"\n",
"/content/audio_samp 100%[===================>] 23.32K 105KB/s in 0.2s \n",
"\n",
"2023-05-25 18:19:14 (105 KB/s) - ‘/content/audio_samples/tmp.mp3’ saved [23877/23877]\n",
"\n"
]
}
]
},
{
"cell_type": "code",
"source": [
"! mkdir -p /content/manifest/\n",
"import os\n",
"with open(\"/content/manifest/dev.tsv\", \"w\") as ftsv, open(\"/content/manifest/dev.lang\", \"w\") as flang:\n",
" ftsv.write(\"/\\n\")\n",
"\n",
" for fl in os.listdir(\"/content/audio_samples/\"):\n",
" if not fl.endswith(\".wav\"):\n",
" continue\n",
" audio_path = f\"/content/audio_samples/{fl}\"\n",
" # duration should be number of samples in audio. For inference, using a random value should be fine. \n",
" duration = 1234 \n",
" ftsv.write(f\"{audio_path}\\t{duration}\\n\")\n",
" flang.write(\"eng\\n\") # This is the \"true\" language for the audio. For inference, using a random value should be fine. \n"
],
"metadata": {
"id": "C2QcjRT-BArW"
},
"execution_count": 7,
"outputs": []
},
{
"cell_type": "markdown",
"source": [
"# 4: Run Inference and transcribe your audio(s)\n"
],
"metadata": {
"id": "44UvHjmMI28Z"
}
},
{
"cell_type": "code",
"source": [
"import os\n",
"\n",
"os.environ[\"PYTHONPATH\"] = \"/content/fairseq\"\n",
"os.environ[\"PREFIX\"] = \"INFER\"\n",
"os.environ[\"HYDRA_FULL_ERROR\"] = \"1\"\n",
"os.environ[\"USER\"] = \"mms_lid_user\"\n",
"\n",
"!python3 examples/mms/lid/infer.py /content/models_lid/{model_name} --path /content/models_lid/{model_name}/mms1b_l126.pt \\\n",
" --task audio_classification --infer-manifest /content/manifest/dev.tsv --output-path /content/manifest/"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "J8N1RKtBiw5V",
"outputId": "09d3fe43-26a4-4f9b-c56d-d38b6d45cdab"
},
"execution_count": 8,
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"2023-05-25 18:19:19.545731: I tensorflow/core/platform/cpu_feature_guard.cc:182] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.\n",
"To enable the following instructions: AVX2 AVX512F FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.\n",
"2023-05-25 18:19:21.567795: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT\n",
"| loading model from /content/models_lid/l126/mms1b_l126.pt\n",
"2023-05-25 18:19:29 | INFO | fairseq.tasks.audio_classification | Using dict_path : /content/models_lid/l126/dict.lang.txt\n",
"2023-05-25 18:19:29 | INFO | root | === Number of labels = 126\n",
"2023-05-25 18:20:01 | INFO | fairseq.data.audio.raw_audio_dataset | loaded 3, skipped 0 samples\n",
"2023-05-25 18:20:01 | INFO | fairseq.tasks.fairseq_task | can_reuse_epoch_itr = True\n",
"2023-05-25 18:20:01 | INFO | fairseq.tasks.fairseq_task | reuse_dataloader = True\n",
"2023-05-25 18:20:01 | INFO | fairseq.tasks.fairseq_task | rebuild_batches = True\n",
"2023-05-25 18:20:01 | INFO | fairseq.tasks.fairseq_task | batches will be rebuilt for each epoch\n",
"2023-05-25 18:20:01 | INFO | fairseq.tasks.fairseq_task | creating new batches for epoch 1\n",
"/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py:560: UserWarning: This DataLoader will create 4 worker processes in total. Our suggested max number of worker in current system is 2, which is smaller than what this DataLoader is going to create. Please be aware that excessive worker creation might get DataLoader running slow or even freeze, lower the worker number to avoid potential slowness/freeze if necessary.\n",
" warnings.warn(_create_warning_msg(\n",
"3it [00:07, 2.61s/it]\n",
"Outputs will be located at - /content/manifest//predictions.txt\n"
]
}
]
},
{
"cell_type": "code",
"source": [
"print(\"----- INPUT FILES -----\")\n",
"! tail -n +2 /content/manifest/dev.tsv\n",
"\n",
"print(\"\\n----- TOP-K PREDICTONS WITH SCORE -----\")\n",
"! cat /content/manifest//predictions.txt"
],
"metadata": {
"colab": {
"base_uri": "https://localhost:8080/"
},
"id": "5f7FROqiC-2z",
"outputId": "3a28ceee-dbb7-4810-f9ca-d11b14a8340b"
},
"execution_count": 9,
"outputs": [
{
"output_type": "stream",
"name": "stdout",
"text": [
"----- INPUT FILES -----\n",
"/content/audio_samples/hi_in.wav\t1234\n",
"/content/audio_samples/en_us.wav\t1234\n",
"/content/audio_samples/cmn_hans_cn.wav\t1234\n",
"\n",
"----- TOP-K PREDICTONS WITH SCORE -----\n",
"[[\"hin\", 0.9931250810623169], [\"urd\", 0.005808886140584946], [\"snd\", 0.0005312535213306546]]\n",
"[[\"eng\", 0.9989539980888367], [\"fas\", 0.00036296260077506304], [\"haw\", 7.031611312413588e-05]]\n",
"[[\"cmn\", 0.9996059536933899], [\"bod\", 0.0002111078501911834], [\"kor\", 9.211552242049947e-05]]\n"
]
}
]
},
{
"cell_type": "code",
"source": [],
"metadata": {
"id": "TzHHmno5DZC4"
},
"execution_count": null,
"outputs": []
}
]
}
|