自定义博客皮肤VIP专享

*博客头图:

格式为PNG、JPG,宽度*高度大于1920*100像素,不超过2MB,主视觉建议放在右侧,请参照线上博客头图

请上传大于1920*100像素的图片!

博客底图:

图片格式为PNG、JPG,不超过1MB,可上下左右平铺至整个背景

栏目图:

图片格式为PNG、JPG,图片宽度*高度为300*38像素,不超过0.5MB

主标题颜色:

RGB颜色,例如:#AFAFAF

Hover:

RGB颜色,例如:#AFAFAF

副标题颜色:

RGB颜色,例如:#AFAFAF

自定义博客皮肤

-+
  • 博客(130)
  • 资源 (4)
  • 收藏
  • 关注

原创 神经网络及其技巧

attentionhttps://blog.csdn.net/zimiao552147572/article/details/105893587/BahdanauAttention注意力机制:基于seq2seq的西班牙语到英语的机器翻译任务、解码器端的Attention注意力机制、seq2seq模型架构

2022-05-30 18:32:04 161 1

原创 musescore

ubuntu MIDI文件查看

2022-01-29 16:03:18 127

原创 pytorch 数据操作技巧

aimport torchalign = torch.FloatTensor([3,4,8]).to(torch.long)torch.repeat_interleave(torch.eye(3),align,dim=1)tensor([[1., 1., 1., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0., 0.],[0., 0., 0., 1., 1., 1., 1., 0., 0., 0., 0., 0., 0., 0., 0.],[0., .

2021-10-10 10:58:00 117

原创 pytorch_lightning

trainimport torchfrom torch.nn import functional as Ffrom torch import nnfrom pytorch_lightning.core.lightning import LightningModuleimport pytorch_lightning as plimport osimport torchimport torchvisionimport torch.nn as nnimport torch.nn.func.

2021-09-23 15:31:41 229

原创 textgrid

from praatio import tgiofrom pydub import AudioSegmentimport osimport globdef create_word_ctm(): with open("datasets/BBSpeech-1.0/metadata.csv",'r') as fp: lines1 = fp.readlines() words = {} phones = {} for i in lines1:

2021-09-10 17:00:49 476

原创 tensorflow rnn

tf.nn.rnn_cell.MultiRNNCell新版tf.contrib.rnn.MultiRNNCell  tf.nn.rnn_cell.MultiRNNCellstate_is_tuple:true,状态Ct和ht就是分开记录,放在一个tuple中,接受和返回的states是n-tuples,其中n=len(cells)# encoding:utf-8import tensorflow as tfbatch_size=10depth=128inputs=tf.Var.

2021-08-27 11:41:48 66

原创 几种Norm区别

BN那样跨样本、单通道,LN那样单样本、跨通道。INS单通道,单样本上的数据进行计算,

2021-07-30 20:04:09 241

原创 ffmpeg

16k to 8kffmpeg -i a.wav -acodec pcm_s16le -ac 1 -ar 8000 b.wav

2021-07-02 14:04:00 128

原创 tf.nn.ctc

batch_size=4, max_time=5, and labels_length=[5,4,3,2].import tensorflow as tflabels = tf.Variable([[4, 3, 1, 2, 5], [2, 3, 4, 1, 0], [1, 2, 3, 0, 0], [5, 4, 0, 0, 0]], tf.int32)with tf..

2021-05-27 13:43:33 89

原创 C++17 新特性总结 [转]

https://blog.csdn.net/baidu_41388533/article/details/110460442

2021-05-20 10:33:55 103

原创 pynini

编译bazel build ...注:bazel clean # 不会删除外部依赖bazel clean --expunge # 会删除外部依赖bazel clean --expunge --asyncbazel buildbazel build : # 在BUILD所在的package目录下执行,编译指定的targetbazel build :all # 编译该package下的所有targetbazel build … # 编译该package下的所有targetbaze.

2021-04-24 21:37:51 649

原创 tensorflow 梯度修剪

tf.clip_by_normimport tensorflow as tfw = tf.Variable(2,dtype=tf.float32)x = tf.Variable(3,dtype=tf.float32)loss = w*x*xoptimizer = tf.train.GradientDescentOptimizer(0.1)grads_and_vars = optimizer.compute_gradients(loss,[w,x])grads = tf.gradients.

2021-04-01 16:19:42 142

原创 wavenet_vocoder

运行./run.sh --stage 2 --stop-stage 2

2021-03-18 20:08:37 129

原创 warpctc

warpctc_pytorchimport torchimport warpctc_pytorch as warp_ctcfrom torch.autograd import Functionfrom torch.nn import Modulefrom ._warp_ctc import *__version__ = '0.1.1'def _assert_no_grad(tensor): assert not tensor.requires_grad, \ .

2021-02-20 18:28:09 124

原创 tensorflow-10

tf.keras.backend.image_data_format()

2021-02-19 18:11:33 68

原创 kaldi

docker操作https://blog.csdn.net/weixin_39946300/article/details/110802648?utm_medium=distribute.pc_relevant_download.none-task-blog-baidujs-1.nonecase&depth_1-utm_source=distribute.pc_relevant_download.none-task-blog-baidujs-1.nonecase

2021-01-23 18:31:02 87

原创 cuda安装

centos7cuda10sudo rpm -i cuda-repo-rhel7-10-0-local-10.0.130-410.48-1.0-1.x86_64.rpmsudo yum clean allsudo yum install cuda

2021-01-15 17:12:32 176

原创 cuda

Stream像素色彩空间转换,将一张7680x4320的8-bit BRGA图像转成同样尺寸的8-bit YUV完整代码(需要NVidia GPU,本文中的测试使用CUDA 10.0):#include <vector>#include <random>#include <iostream>#include <cuda.h>#include <cuda_runtime.h>#ifdef DEBUG#define CUDA.

2021-01-10 20:09:01 158

原创 C/C++ 调试工具

cppcheckcppcheck主要用于对C/C++源代码进行分析检查的一个开源工具,可以用来检测未使用的变量、越界访问、内存泄漏等问题。cppcheck --enable=all NAME.cppgcovgcov是伴随gcc发布的一个代码覆盖率检查工具,可以用来检查代码中各个语句的执行次数,查看代码执行逻辑,方便后期对代码的优化。在 gcc/g++ 的编译选项中添加两个选项 -fprofile-arcs -ftest-coverage运行可执行程序,生成两个包含代码覆盖信息的两个文.

2021-01-02 15:55:58 788

原创 C/C++ iterator

std::iteratortemplate< class Category,class T,class Distance = std::ptrdiff_t,class Pointer = T*,class Reference = T& > struct iterator;std::iterator 的模板参数:Category: 类型为 iterator_category,表示迭代器的种类,共有5类:T :类型为 value_type, 可以通过解除引用迭代器获得.

2021-01-02 15:34:43 303 1

原创 C/C++ 多线程

atomic#include <atomic>#include <thread>#include <assert.h>std::atomic<bool> x,y;std::atomic<int> z;void write_x_then_y(){ x.store(true,std::memory_order_relaxed); // 1 y.store(true,std::memory_order_relaxed).

2020-12-29 16:18:10 74

原创 C/C++ stdarg

#include <stdarg.h>#include <stdio.h>int AveInt(int,...);int main(){ printf("%d\n",AveInt(2,2,3)); printf("%s\n","--------------------------"); printf("%d\n",AveInt(4,2,4,6,8)); return 0;}int AveInt(int v,...){ int

2020-12-24 14:48:34 202

原创 yule-walker

1https://blog.csdn.net/naruhina/article/details/107013196https://zhuanlan.zhihu.com/p/64933417

2020-12-23 14:19:22 370

原创 scipy

lpcimport librosaimport scipy.io.wavfile as swavimport numpy as npimport matplotlib.pyplot as plt import scipy y, sr = librosa.load("test0.wav", offset=3,duration = 0.02) a = librosa.lpc(y, 2) #b = [0] + -1 * a[1:] b = [-2,1] .

2020-12-22 16:51:50 126

原创 nccl

expimport torchimport torch.distributed as distimport argparsefrom time import sleepfrom random import randintfrom torch.multiprocessing import Processdef initialize(rank, world_size, ip, port): dist.init_process_group(backend='nccl', init_m.

2020-12-22 10:16:03 253

原创 C/C++ map

#include <iostream>#include <map>#include <sstream>#include <string>#include <vector>#include <algorithm>#include <functional> using namespace std; struct FoundByValue { FoundByValue(string str):_s(

2020-12-05 14:19:00 139

原创 boost之thread_specific_ptr

#include <iostream>#include <time.h>#include <boost/random/uniform_int_distribution.hpp>#include <boost/random/mersenne_twister.hpp>#include <boost/thread.hpp>#include <boost/thread/tss.hpp> // thread_specific_ptr

2020-12-02 22:32:55 303

原创 docker

安装基本使用网络docker network lsdocker run -it --net host centos:centos7 /bin/bash

2020-11-02 14:35:32 188

原创 CPython

数据类型#include <stdio.h>#include <Python.h>using namespace std;PyObject* none_object(){ Py_RETURN_NONE; // 不需要自己return}int main() { // 初始化Python虚拟机 Py_Initialize(); // 判断Python虚拟机是否成功 if (Py_IsInitialized() == 0){ .

2020-10-24 00:01:14 231

原创 C/C++ 命令参数解析

exp1main.cpp#include <stdio.h>#include <stdlib.h>#include "argsh.h"int main(int argc, char* argv[]){ int tmp = 4; while( (tmp = getopt(argc, argv, "abck")) != -1 ) { printf("00000000\n"); printf("tmp=%.

2020-10-23 16:08:44 311

原创 cuda

exp1#include <cuda.h>#include <iostream>struct HostObject { __host__ int value() const { return 42; }};struct DeviceObject { __device__ int value() const { return 3; }};template <typename T>__host__ __device__i

2020-10-13 13:59:07 376 1

原创 paddlepaddle

paddlepaddle-v1.0.0编译g++-5(g++-6会报未识别编译选项)

2020-10-13 07:55:32 140

原创 C/C++ pthread

std::mutex#include <iostream>#include <chrono>#include <mutex>#include <thread>struct CriticalData{ std::mutex mut;};void deadLock(CriticalData& a, CriticalData& b){ std::unique_lock<std::mutex>guard1(a.

2020-09-30 14:44:26 230

原创 cmake

卸载xargs rm < install_manifest.txt

2020-09-26 20:59:22 128

原创 C/C++ std::function

#include <iostream>#include <functional>#include <memory>class classA{ typedef std::function<void(int i)> callback_t; public: classA() {} ~classA() {} void handle(int i) { std::cout << "classA::handle " &l

2020-09-24 15:31:34 164

原创 tensorflow编译

对tensorflow-gpu10源码进行编译报错ERROR: Config value cuda is not defined in any .rc file问题原因:bazel版本过高。解决办法:tf1.10的构建版本时需要使用bazel 0.15,换成了bazel 0.15版本构建问题解决。r1.11 bazel版本0.18 ...

2020-09-13 17:48:34 212

原创 C/C++ 结构体

#include <iostream>struct LogMessageEnvelope{ enum Severity{ kAssertFailed = -3, kError = -2, kWarning = -1, kInfo = 0 }; int severity;};int A(LogMessageEnvelope::Severity severity){ std::cout <

2020-08-26 14:46:52 159

原创 C/C++ value_type

#include <iostream>#include <vector>using namespace std;template<typename T>typename T::value_type top(const T &c){ if (!c.empty()){ return c.back(); } else return typename T::value_type();}int mai

2020-08-14 09:47:43 1119

原创 std::sort

#include<iostream>#include<vector>#include<algorithm>struct PathTrie{ int score; char character; int id;}*aa,*bb,*cc,*dd;bool prefix_compare(const PathTrie *x, const PathTrie *y) { if (x->score == y-&

2020-07-22 11:00:51 110

原创 wget

wget -c http://www.repository.voxforge1.org/downloads/SpeechCorpus/Trunk/Audio/Main/16kHz_16bit -l 1 -N -nd -c -e robots=off -A tgz -r -np -P /home/liyuan/.cache/paddle/dataset/speech/VoxForge/tgz

2020-07-21 09:42:21 76

10-batches-py.tar.gz

cifar10数据,用于神经网络训练的标准数据,包含5个文件, 该数据集共有60000张彩色图像,这些图像是32*32,分为10个类,每类6000张图。这里面有50000张用于训练,构成了5个训练批,每一批10000张图;另外10000用.

2020-07-17

VC6LineNumberAddin.rar

vc++6.0 regsvr32.exe *.dll 管理员运行 C:\Program Files (x86)\Microsoft Visual Studio\Common\MSDev98\AddIns

2020-04-15

fortan序列号sn.txt

fortan序列号.sn.txt,FortranPowerstation4

2011-06-27

键盘记录工具enregisterkey

This is a directe exe file!don`t need install

2011-03-30

空空如也

TA创建的收藏夹 TA关注的收藏夹

TA关注的人

提示
确定要删除当前文章?
取消 删除