Delete Libtorch C++ Infer
Browse files
Libtorch C++ Infer/VITS-LibTorch.cpp
DELETED
|
@@ -1,121 +0,0 @@
|
|
| 1 |
-
#include <iostream>
|
| 2 |
-
#include <torch/torch.h>
|
| 3 |
-
#include <torch/script.h>
|
| 4 |
-
#include <string>
|
| 5 |
-
#include <vector>
|
| 6 |
-
#include <locale>
|
| 7 |
-
#include <codecvt>
|
| 8 |
-
#include <direct.h>
|
| 9 |
-
#include <fstream>
|
| 10 |
-
typedef int64_t int64;
|
| 11 |
-
namespace Shirakana {
|
| 12 |
-
|
| 13 |
-
struct WavHead {
|
| 14 |
-
char RIFF[4];
|
| 15 |
-
long int size0;
|
| 16 |
-
char WAVE[4];
|
| 17 |
-
char FMT[4];
|
| 18 |
-
long int size1;
|
| 19 |
-
short int fmttag;
|
| 20 |
-
short int channel;
|
| 21 |
-
long int samplespersec;
|
| 22 |
-
long int bytepersec;
|
| 23 |
-
short int blockalign;
|
| 24 |
-
short int bitpersamples;
|
| 25 |
-
char DATA[4];
|
| 26 |
-
long int size2;
|
| 27 |
-
};
|
| 28 |
-
|
| 29 |
-
int conArr2Wav(int64 size, int16_t* input, const char* filename) {
|
| 30 |
-
WavHead head = { {'R','I','F','F'},0,{'W','A','V','E'},{'f','m','t',' '},16,
|
| 31 |
-
1,1,22050,22050 * 2,2,16,{'d','a','t','a'},
|
| 32 |
-
0 };
|
| 33 |
-
head.size0 = size * 2 + 36;
|
| 34 |
-
head.size2 = size * 2;
|
| 35 |
-
std::ofstream ocout;
|
| 36 |
-
char* outputData = (char*)input;
|
| 37 |
-
ocout.open(filename, std::ios::out | std::ios::binary);
|
| 38 |
-
ocout.write((char*)&head, 44);
|
| 39 |
-
ocout.write(outputData, (int32_t)(size * 2));
|
| 40 |
-
ocout.close();
|
| 41 |
-
return 0;
|
| 42 |
-
}
|
| 43 |
-
|
| 44 |
-
inline std::wstring to_wide_string(const std::string& input)
|
| 45 |
-
{
|
| 46 |
-
std::wstring_convert<std::codecvt_utf8<wchar_t>> converter;
|
| 47 |
-
return converter.from_bytes(input);
|
| 48 |
-
}
|
| 49 |
-
|
| 50 |
-
inline std::string to_byte_string(const std::wstring& input)
|
| 51 |
-
{
|
| 52 |
-
std::wstring_convert<std::codecvt_utf8<wchar_t>> converter;
|
| 53 |
-
return converter.to_bytes(input);
|
| 54 |
-
}
|
| 55 |
-
}
|
| 56 |
-
|
| 57 |
-
#define val const auto
|
| 58 |
-
int main()
|
| 59 |
-
{
|
| 60 |
-
torch::jit::Module Vits;
|
| 61 |
-
std::string buffer;
|
| 62 |
-
std::vector<int64> text;
|
| 63 |
-
std::vector<int16_t> data;
|
| 64 |
-
while(true)
|
| 65 |
-
{
|
| 66 |
-
while (true)
|
| 67 |
-
{
|
| 68 |
-
std::cin >> buffer;
|
| 69 |
-
if (buffer == "end")
|
| 70 |
-
return 0;
|
| 71 |
-
if(buffer == "model")
|
| 72 |
-
{
|
| 73 |
-
std::cin >> buffer;
|
| 74 |
-
Vits = torch::jit::load(buffer);
|
| 75 |
-
continue;
|
| 76 |
-
}
|
| 77 |
-
if (buffer == "endinfer")
|
| 78 |
-
{
|
| 79 |
-
Shirakana::conArr2Wav(data.size(), data.data(), "temp\\tmp.wav");
|
| 80 |
-
data.clear();
|
| 81 |
-
std::cout << "endofinfe";
|
| 82 |
-
continue;
|
| 83 |
-
}
|
| 84 |
-
if (buffer == "line")
|
| 85 |
-
{
|
| 86 |
-
std::cin >> buffer;
|
| 87 |
-
while (buffer.find("endline")==std::string::npos)
|
| 88 |
-
{
|
| 89 |
-
text.push_back(std::atoi(buffer.c_str()));
|
| 90 |
-
std::cin >> buffer;
|
| 91 |
-
}
|
| 92 |
-
val InputTensor = torch::from_blob(text.data(), { 1,static_cast<int64>(text.size()) }, torch::kInt64);
|
| 93 |
-
std::array<int64, 1> TextLength{ static_cast<int64>(text.size()) };
|
| 94 |
-
val InputTensor_length = torch::from_blob(TextLength.data(), { 1 }, torch::kInt64);
|
| 95 |
-
std::vector<torch::IValue> inputs;
|
| 96 |
-
inputs.push_back(InputTensor);
|
| 97 |
-
inputs.push_back(InputTensor_length);
|
| 98 |
-
if (buffer.length() > 7)
|
| 99 |
-
{
|
| 100 |
-
std::array<int64, 1> speakerIndex{ (int64)atoi(buffer.substr(7).c_str()) };
|
| 101 |
-
inputs.push_back(torch::from_blob(speakerIndex.data(), { 1 }, torch::kLong));
|
| 102 |
-
}
|
| 103 |
-
val output = Vits.forward(inputs).toTuple()->elements()[0].toTensor().multiply(32276.0F);
|
| 104 |
-
val outputSize = output.sizes().at(2);
|
| 105 |
-
val floatOutput = output.data_ptr<float>();
|
| 106 |
-
int16_t* outputTmp = (int16_t*)malloc(sizeof(float) * outputSize);
|
| 107 |
-
if (outputTmp == nullptr) {
|
| 108 |
-
throw std::exception("内存不足");
|
| 109 |
-
}
|
| 110 |
-
for (int i = 0; i < outputSize; i++) {
|
| 111 |
-
*(outputTmp + i) = (int16_t) * (floatOutput + i);
|
| 112 |
-
}
|
| 113 |
-
data.insert(data.end(), outputTmp, outputTmp+outputSize);
|
| 114 |
-
free(outputTmp);
|
| 115 |
-
text.clear();
|
| 116 |
-
std::cout << "endofline";
|
| 117 |
-
}
|
| 118 |
-
}
|
| 119 |
-
}
|
| 120 |
-
//model S:\VSGIT\ShirakanaTTSUI\build\x64\Release\Mods\AtriVITS\AtriVITS_LJS.pt
|
| 121 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Libtorch C++ Infer/toLibTorch.ipynb
DELETED
|
@@ -1,142 +0,0 @@
|
|
| 1 |
-
{
|
| 2 |
-
"cells": [
|
| 3 |
-
{
|
| 4 |
-
"cell_type": "code",
|
| 5 |
-
"execution_count": null,
|
| 6 |
-
"metadata": {},
|
| 7 |
-
"outputs": [],
|
| 8 |
-
"source": [
|
| 9 |
-
"%matplotlib inline\n",
|
| 10 |
-
"import matplotlib.pyplot as plt\n",
|
| 11 |
-
"import IPython.display as ipd\n",
|
| 12 |
-
"\n",
|
| 13 |
-
"import os\n",
|
| 14 |
-
"import json\n",
|
| 15 |
-
"import math\n",
|
| 16 |
-
"import torch\n",
|
| 17 |
-
"from torch import nn\n",
|
| 18 |
-
"from torch.nn import functional as F\n",
|
| 19 |
-
"from torch.utils.data import DataLoader\n",
|
| 20 |
-
"\n",
|
| 21 |
-
"import ../commons\n",
|
| 22 |
-
"import ../utils\n",
|
| 23 |
-
"from ../data_utils import TextAudioLoader, TextAudioCollate, TextAudioSpeakerLoader, TextAudioSpeakerCollate\n",
|
| 24 |
-
"from ../models import SynthesizerTrn\n",
|
| 25 |
-
"from ../text.symbols import symbols\n",
|
| 26 |
-
"from ../text import text_to_sequence\n",
|
| 27 |
-
"\n",
|
| 28 |
-
"from scipy.io.wavfile import write\n",
|
| 29 |
-
"\n",
|
| 30 |
-
"\n",
|
| 31 |
-
"def get_text(text, hps):\n",
|
| 32 |
-
" text_norm = text_to_sequence(text, hps.data.text_cleaners)\n",
|
| 33 |
-
" if hps.data.add_blank:\n",
|
| 34 |
-
" text_norm = commons.intersperse(text_norm, 0)\n",
|
| 35 |
-
" text_norm = torch.LongTensor(text_norm)\n",
|
| 36 |
-
" return text_norm"
|
| 37 |
-
]
|
| 38 |
-
},
|
| 39 |
-
{
|
| 40 |
-
"cell_type": "code",
|
| 41 |
-
"execution_count": null,
|
| 42 |
-
"metadata": {},
|
| 43 |
-
"outputs": [],
|
| 44 |
-
"source": [
|
| 45 |
-
"#############################################################\n",
|
| 46 |
-
"# #\n",
|
| 47 |
-
"# Single Speakers #\n",
|
| 48 |
-
"# #\n",
|
| 49 |
-
"#############################################################"
|
| 50 |
-
]
|
| 51 |
-
},
|
| 52 |
-
{
|
| 53 |
-
"cell_type": "code",
|
| 54 |
-
"execution_count": null,
|
| 55 |
-
"metadata": {},
|
| 56 |
-
"outputs": [],
|
| 57 |
-
"source": [
|
| 58 |
-
"hps = utils.get_hparams_from_file(\"configs/XXX.json\") #将\"\"内的内容修改为你的模型路径与config路径\n",
|
| 59 |
-
"net_g = SynthesizerTrn(\n",
|
| 60 |
-
" len(symbols),\n",
|
| 61 |
-
" hps.data.filter_length // 2 + 1,\n",
|
| 62 |
-
" hps.train.segment_size // hps.data.hop_length,\n",
|
| 63 |
-
" **hps.model).cuda()\n",
|
| 64 |
-
"_ = net_g.eval()\n",
|
| 65 |
-
"\n",
|
| 66 |
-
"_ = utils.load_checkpoint(\"/path/to/model.pth\", net_g, None)"
|
| 67 |
-
]
|
| 68 |
-
},
|
| 69 |
-
{
|
| 70 |
-
"cell_type": "code",
|
| 71 |
-
"execution_count": null,
|
| 72 |
-
"metadata": {},
|
| 73 |
-
"outputs": [],
|
| 74 |
-
"source": [
|
| 75 |
-
"stn_tst = get_text(\"こんにちは\", hps)\n",
|
| 76 |
-
"with torch.no_grad():\n",
|
| 77 |
-
" x_tst = stn_tst.cuda().unsqueeze(0)\n",
|
| 78 |
-
" x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).cuda()\n",
|
| 79 |
-
" traced_mod = torch.jit.trace(net_g,(x_tst, x_tst_lengths,sid))\n",
|
| 80 |
-
" torch.jit.save(traced_mod,\"OUTPUTLIBTORCHMODEL.pt\")\n",
|
| 81 |
-
" audio = net_g.infer(x_tst, x_tst_lengths, noise_scale=.667, noise_scale_w=0.8, length_scale=1)[0][0,0].data.cpu().float().numpy()\n",
|
| 82 |
-
"ipd.display(ipd.Audio(audio, rate=hps.data.sampling_rate, normalize=False))"
|
| 83 |
-
]
|
| 84 |
-
},
|
| 85 |
-
{
|
| 86 |
-
"cell_type": "code",
|
| 87 |
-
"execution_count": null,
|
| 88 |
-
"metadata": {},
|
| 89 |
-
"outputs": [],
|
| 90 |
-
"source": [
|
| 91 |
-
"#############################################################\n",
|
| 92 |
-
"# #\n",
|
| 93 |
-
"# Multiple Speakers #\n",
|
| 94 |
-
"# #\n",
|
| 95 |
-
"#############################################################"
|
| 96 |
-
]
|
| 97 |
-
},
|
| 98 |
-
{
|
| 99 |
-
"cell_type": "code",
|
| 100 |
-
"execution_count": null,
|
| 101 |
-
"metadata": {},
|
| 102 |
-
"outputs": [],
|
| 103 |
-
"source": [
|
| 104 |
-
"hps = utils.get_hparams_from_file(\"./configs/XXX.json\") #将\"\"内的内容修改为你的模型路径与config路径\n",
|
| 105 |
-
"net_g = SynthesizerTrn(\n",
|
| 106 |
-
" len(symbols),\n",
|
| 107 |
-
" hps.data.filter_length // 2 + 1,\n",
|
| 108 |
-
" hps.train.segment_size // hps.data.hop_length,\n",
|
| 109 |
-
" n_speakers=hps.data.n_speakers,\n",
|
| 110 |
-
" **hps.model).cuda()\n",
|
| 111 |
-
"_ = net_g.eval()\n",
|
| 112 |
-
"\n",
|
| 113 |
-
"_ = utils.load_checkpoint(\"/path/to/model.pth\", net_g, None)"
|
| 114 |
-
]
|
| 115 |
-
},
|
| 116 |
-
{
|
| 117 |
-
"cell_type": "code",
|
| 118 |
-
"execution_count": null,
|
| 119 |
-
"metadata": {},
|
| 120 |
-
"outputs": [],
|
| 121 |
-
"source": [
|
| 122 |
-
"stn_tst = get_text(\"こんにちは\", hps)\n",
|
| 123 |
-
"with torch.no_grad():\n",
|
| 124 |
-
" x_tst = stn_tst.cuda().unsqueeze(0)\n",
|
| 125 |
-
" x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).cuda()\n",
|
| 126 |
-
" sid = torch.LongTensor([4]).cuda()\n",
|
| 127 |
-
" traced_mod = torch.jit.trace(net_g,(x_tst, x_tst_lengths,sid))\n",
|
| 128 |
-
" torch.jit.save(traced_mod,\"OUTPUTLIBTORCHMODEL.pt\")\n",
|
| 129 |
-
" audio = net_g.infer(x_tst, x_tst_lengths, sid=sid, noise_scale=.667, noise_scale_w=0.8, length_scale=1)[0][0,0].data.cpu().float().numpy()\n",
|
| 130 |
-
"ipd.display(ipd.Audio(audio, rate=hps.data.sampling_rate, normalize=False))"
|
| 131 |
-
]
|
| 132 |
-
}
|
| 133 |
-
],
|
| 134 |
-
"metadata": {
|
| 135 |
-
"language_info": {
|
| 136 |
-
"name": "python"
|
| 137 |
-
},
|
| 138 |
-
"orig_nbformat": 4
|
| 139 |
-
},
|
| 140 |
-
"nbformat": 4,
|
| 141 |
-
"nbformat_minor": 2
|
| 142 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|