mirror of
https://github.com/snakers4/silero-vad.git
synced 2026-02-05 18:09:22 +08:00
delete onnx from utils
This commit is contained in:
adamnsandle
283
silero-vad.ipynb
283
silero-vad.ipynb
@@ -1,30 +1,39 @@
|
||||
{
|
||||
"cells": [
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"# Jit example"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-14T13:43:24.487521Z",
|
||||
"start_time": "2020-12-14T13:43:23.780570Z"
|
||||
"end_time": "2020-12-15T11:54:25.940761Z",
|
||||
"start_time": "2020-12-15T11:54:25.933842Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"# imports\n",
|
||||
"import glob\n",
|
||||
"import torch\n",
|
||||
"from IPython.display import Audio\n",
|
||||
"torch.set_num_threads(1)\n",
|
||||
"from utils import (init_jit_model, get_speech_ts, \n",
|
||||
"\n",
|
||||
"from utils import (init_jit_model, get_speech_ts,\n",
|
||||
" save_audio, read_audio, \n",
|
||||
" state_generator, single_audio_stream, init_onnx_model)"
|
||||
" state_generator, single_audio_stream)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"# Full audio example"
|
||||
"## Full audio"
|
||||
]
|
||||
},
|
||||
{
|
||||
@@ -32,8 +41,8 @@
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-14T13:43:24.492506Z",
|
||||
"start_time": "2020-12-14T13:43:24.489440Z"
|
||||
"end_time": "2020-12-15T11:54:27.939388Z",
|
||||
"start_time": "2020-12-15T11:54:27.936636Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
@@ -50,14 +59,13 @@
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-14T13:43:24.760714Z",
|
||||
"start_time": "2020-12-14T13:43:24.493992Z"
|
||||
"end_time": "2020-12-15T11:54:28.415177Z",
|
||||
"start_time": "2020-12-15T11:54:28.231677Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"model = init_jit_model('files/joint_VAD_just_RU_jit_cut_q.pth.tar', 'cpu') # from yml file\n",
|
||||
"model = init_onnx_model('files/joint_VAD_just_RU.onnx')"
|
||||
"model = init_jit_model('files/model.jit', 'cpu')"
|
||||
]
|
||||
},
|
||||
{
|
||||
@@ -65,14 +73,13 @@
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-14T13:43:24.793384Z",
|
||||
"start_time": "2020-12-14T13:43:24.762311Z"
|
||||
"end_time": "2020-12-15T11:54:28.560822Z",
|
||||
"start_time": "2020-12-15T11:54:28.549811Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"Audio('files/test_audio_2.wav')\n",
|
||||
"wav = read_audio('files/test_audio_2.wav')"
|
||||
"wav = read_audio('files/en.wav')"
|
||||
]
|
||||
},
|
||||
{
|
||||
@@ -80,13 +87,13 @@
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-14T13:43:25.320324Z",
|
||||
"start_time": "2020-12-14T13:43:24.808594Z"
|
||||
"end_time": "2020-12-15T11:54:30.088721Z",
|
||||
"start_time": "2020-12-15T11:54:29.019358Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"speech_timestamps = get_speech_ts(wav, model, num_steps=4) # kill extractor"
|
||||
"speech_timestamps = get_speech_ts(wav, model, num_steps=4) # get speech timestamps from full audio file"
|
||||
]
|
||||
},
|
||||
{
|
||||
@@ -94,8 +101,8 @@
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-14T13:43:25.324901Z",
|
||||
"start_time": "2020-12-14T13:43:25.321759Z"
|
||||
"end_time": "2020-12-15T11:54:30.198484Z",
|
||||
"start_time": "2020-12-15T11:54:30.188311Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
@@ -108,13 +115,13 @@
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-14T13:43:25.344065Z",
|
||||
"start_time": "2020-12-14T13:43:25.326162Z"
|
||||
"end_time": "2020-12-15T11:54:30.816893Z",
|
||||
"start_time": "2020-12-15T11:54:30.782667Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"save_audio('only_speech.wav', collect_speeches(speech_timestamps, wav), 16000)\n",
|
||||
"save_audio('only_speech.wav', collect_speeches(speech_timestamps, wav), 16000) # merge all speech chunks to one audio\n",
|
||||
"Audio('only_speech.wav')"
|
||||
]
|
||||
},
|
||||
@@ -122,7 +129,7 @@
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"# Single stream example"
|
||||
"## Single audio stream"
|
||||
]
|
||||
},
|
||||
{
|
||||
@@ -130,15 +137,14 @@
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-14T13:43:25.778585Z",
|
||||
"start_time": "2020-12-14T13:43:25.496583Z"
|
||||
"end_time": "2020-12-15T11:54:31.886189Z",
|
||||
"start_time": "2020-12-15T11:54:31.572194Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"model = init_jit_model('files/joint_VAD_just_RU_jit_cut_q.pth.tar', 'cpu') # from yml file\n",
|
||||
"#model = init_onnx_model('files/joint_VAD_just_RU.onnx')\n",
|
||||
"audio = 'files/test_audio_6.wav'"
|
||||
"model = init_jit_model('files/model.jit', 'cpu')\n",
|
||||
"wav = 'files/en.wav'"
|
||||
]
|
||||
},
|
||||
{
|
||||
@@ -146,13 +152,13 @@
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-14T13:43:29.402604Z",
|
||||
"start_time": "2020-12-14T13:43:25.780037Z"
|
||||
"end_time": "2020-12-15T11:54:35.624279Z",
|
||||
"start_time": "2020-12-15T11:54:32.049532Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"for i in single_audio_stream(model, audio):\n",
|
||||
"for i in single_audio_stream(model, wav):\n",
|
||||
" if i:\n",
|
||||
" print(i)"
|
||||
]
|
||||
@@ -161,7 +167,7 @@
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"# Multiple stream example"
|
||||
"## Multiple audio stream"
|
||||
]
|
||||
},
|
||||
{
|
||||
@@ -169,14 +175,13 @@
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-14T13:43:29.674262Z",
|
||||
"start_time": "2020-12-14T13:43:29.403972Z"
|
||||
"end_time": "2020-12-15T11:40:13.406225Z",
|
||||
"start_time": "2020-12-15T11:40:13.206354Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"model = init_jit_model('files/joint_VAD_just_RU_jit_cut_q.pth.tar', 'cpu') # from yml file\n",
|
||||
"model = init_onnx_model('files/joint_VAD_just_RU.onnx')"
|
||||
"model = init_jit_model('files/model.jit', 'cpu')"
|
||||
]
|
||||
},
|
||||
{
|
||||
@@ -184,14 +189,14 @@
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-14T13:43:29.678449Z",
|
||||
"start_time": "2020-12-14T13:43:29.675519Z"
|
||||
"end_time": "2020-12-15T11:41:08.470917Z",
|
||||
"start_time": "2020-12-15T11:41:08.467369Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"audios_for_stream = glob.glob('files/test*.wav')\n",
|
||||
"len(audios_for_stream)"
|
||||
"audios_for_stream = glob.glob('files/*.wav')\n",
|
||||
"len(audios_for_stream) # total 4 audios"
|
||||
]
|
||||
},
|
||||
{
|
||||
@@ -199,29 +204,211 @@
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-14T13:43:40.236387Z",
|
||||
"start_time": "2020-12-14T13:43:29.679274Z"
|
||||
"end_time": "2020-12-15T11:41:25.685356Z",
|
||||
"start_time": "2020-12-15T11:41:16.222672Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"for i in state_generator(model, audios_for_stream, audios_in_stream=2):\n",
|
||||
"for i in state_generator(model, audios_for_stream, audios_in_stream=2): # 2 audio stream\n",
|
||||
" if i:\n",
|
||||
" print(i)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": 2,
|
||||
"execution_count": null,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": []
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"# Onnx example"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-14T13:46:49.812052Z",
|
||||
"start_time": "2020-12-14T13:46:49.586637Z"
|
||||
"end_time": "2020-12-15T11:55:45.597504Z",
|
||||
"start_time": "2020-12-15T11:55:45.582356Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"!cp ../silero-models-research/model_saves/joint_VAD_just_RU_jit_cut_q.pth.tar files/"
|
||||
"# imports\n",
|
||||
"import glob\n",
|
||||
"import torch\n",
|
||||
"from IPython.display import Audio\n",
|
||||
"torch.set_num_threads(1)\n",
|
||||
"import onnxruntime\n",
|
||||
"\n",
|
||||
"from utils import (get_speech_ts, save_audio, read_audio, \n",
|
||||
" state_generator, single_audio_stream)\n",
|
||||
"\n",
|
||||
"def init_onnx_model(model_path: str):\n",
|
||||
" return onnxruntime.InferenceSession(model_path)\n",
|
||||
"\n",
|
||||
"def validate_onnx(model, inputs):\n",
|
||||
" with torch.no_grad():\n",
|
||||
" ort_inputs = {'input': inputs.cpu().numpy()}\n",
|
||||
" outs = model.run(None, ort_inputs)\n",
|
||||
" outs = [torch.Tensor(x) for x in outs]\n",
|
||||
" return outs"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Full audio"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-15T11:55:56.874376Z",
|
||||
"start_time": "2020-12-15T11:55:56.782230Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"model = init_onnx_model('files/model.onnx')\n",
|
||||
"wav = read_audio('files/en.wav')"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-15T11:56:12.159463Z",
|
||||
"start_time": "2020-12-15T11:56:11.446991Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"speech_timestamps = get_speech_ts(wav, model, num_steps=4, run_function=validate_onnx) # get speech timestamps from full audio file"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-15T11:56:20.488863Z",
|
||||
"start_time": "2020-12-15T11:56:20.485485Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"speech_timestamps"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-15T11:56:27.908128Z",
|
||||
"start_time": "2020-12-15T11:56:27.870978Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"save_audio('only_speech.wav', collect_speeches(speech_timestamps, wav), 16000) # merge all speech chunks to one audio\n",
|
||||
"Audio('only_speech.wav')"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Single audio stream"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-15T11:58:09.012892Z",
|
||||
"start_time": "2020-12-15T11:58:08.940907Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"model = init_onnx_model('files/model.onnx')\n",
|
||||
"wav = 'files/en.wav'"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-15T11:58:11.562186Z",
|
||||
"start_time": "2020-12-15T11:58:09.949825Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"for i in single_audio_stream(model, wav, run_function=validate_onnx):\n",
|
||||
" if i:\n",
|
||||
" print(i)"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"## Multiple audio stream"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"model = init_onnx_model('files/model.onnx')"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-15T11:59:09.381687Z",
|
||||
"start_time": "2020-12-15T11:59:09.378552Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"audios_for_stream = glob.glob('files/*.wav')\n",
|
||||
"len(audios_for_stream) # total 4 audios"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "code",
|
||||
"execution_count": null,
|
||||
"metadata": {
|
||||
"ExecuteTime": {
|
||||
"end_time": "2020-12-15T11:59:27.712905Z",
|
||||
"start_time": "2020-12-15T11:59:21.608435Z"
|
||||
}
|
||||
},
|
||||
"outputs": [],
|
||||
"source": [
|
||||
"for i in state_generator(model, audios_for_stream, audios_in_stream=2, run_function=validate_onnx): # 2 audio stream\n",
|
||||
" if i:\n",
|
||||
" print(i)"
|
||||
]
|
||||
},
|
||||
{
|
||||
|
||||
Reference in New Issue
Block a user