Jerry's Blog

PyTorch without CUDA:conda install pytorch torchvision cpuonly -c pytorchPyTorch with CUDA 10.1:conda install pytorch torchvision cudatoolkit=10.1 -c pytorch

if configarp - a

R matrix is the matrix from x_t to prediction x_t+1: R = A*C +S which has the dimension of (sensor_numbers, sensor_nukmbers)Statistics:In file:A: Layer2_time1109B: Layer2_time1110C:Layer3_time1107D:Layer3_time1106A = np.array([ [1.131289, 0.2404

robot->addSensor(std::make_shared<SpeedSensor>(1,SpeedSensor::Translational,SpeedSensor::XYZ), Attachment(-1));robot->addSensor(std::make_shared<RelativePositionSensor>(1,1,RelativePositionSensor::XY), Attachment(0));

ffmpeg -i hexabot2.mkv -ss 00:00:39 -t 00:02:28 -async 1 cut.mp4

conda list -e > requirements.txtconda env export > <environment-name>.ymlCreate full conda environment from yml file:conda env create -f .yml

pip install -r requirements.txt# install [train] version of python package with the rllib dependenciespip install -e .[train]# make sure you can run tests (and verify they are passing)pip install -e .[test]make test

Use XQuartz version: XQuartz 2.7.8, else it may have libGL error!set XQuartz preferences --> Security --> Tick all connectionsin /etc/ssh/sshd_configset x11forwardingto yes (there are two x11forwarding, all set from no to yes)The scripts used to.

//The way to assign Matrix：First transform double[*] to double[*,*] using this ArraytoMatrix method; Matrix qq = new Matrix(Utility.ArrayToMatrix(1, reals.Length, reals)); Console.WriteLine(qq); //The right way to use a

using System;using System.Collections;using System.Collections.Generic;using System.IO;using LinearAlgebra.MatrixAlgebra;using LinearAlgebra.VectorAlgebra;using LinearAlgebra.LinearEquations;namespace LinearAlgebra{ class Program {

Saturation Problem solved with Better Results!In the original diamond controller code, the update parameter resulted by the tle error is set to default 1.0:My change of code is:'gradient'.map(g).mapP(0.1,clip)using EE = .1/(v.norm_sqr() + .001); (c

bash script to backup all conda environment lists#!bin/bashNOW=$(date "+%Y-%m-%d")CONDA_BASE=$(conda info --base)CONDA_FUNCTION="etc/profile.d/conda.sh"CONDA="$CONDA_BASE/$CONDA_FUNCTION"source $CONDAmkdir ./condaenvs-$NOWENVS=$(conda env list |

Crowpi 2 githubCrowpi 2 image google drive link

First, execute python to check the default python on your raspberry Pi. Press Ctrl-Z to exit.If it is python3, you can skip this section.If it is python2, you need execute the following commands to set default python to python3.Enter directory /usr/bin

pip install ./cmake_exampleTraceback (most recent call last):File “/home/jerry/anaconda3/bin/pip”, line 7, in from pip._internal.cli.main import mainFile “/usr/lib/python3/dist-packages/pip/init.py”, line 29, in from pip.utils import get_installed_d.

https://github.com/google-research/motion_imitation

cmake_minimum_required(VERSION 2.8.4)project(kalman)set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")find_package(Eigen3 REQUIRED)include_directories(${EIGEN3_INCLUDE_DIR})set(SOURCE_FILES kalman.cpp kalman-test.cpp)add_executable(kalman-test $

/* compile this with g++ -Wall -lm -L. -lselforg -o example matrixexample.cpp or use the Makefile*/#include <iostream>#include <math.h>#include <selforg/matrix.h>using namespace matrix;using namespace std;int main(){ /////////.

#include <iostream>#include <eigen3/Eigen/Dense>#include <fstream>#include <vector>using namespace Eigen;using namespace std;void saveData(string filename, MatrixXd matrix){ //dynamic, doubles const static IOFormat CSV

#include <eigen3/Eigen/Core>#include <iostream>#include <eigen3/Eigen/Dense>using namespace Eigen;using namespace std;int main(){ Matrix <float, 4 , 3> matrixA; matrixA.setOnes(); cout<< matrixA <<

/** track options */ TrackRobot* TrackOpt = new TrackRobot(false,false,false, track_rob); TrackOpt->conf.displayTraceDur = 2000; TrackOpt->conf.displayTraceThickness = 0.01; // 0.01 or 0. agent->setTrackOptions( *TrackOpt )

add & commit You can propose changes (add it to the Index) using git add <filename> git add * This is the first step in the basic git workflow. To actually commit these changes use...

d = {'k1':1,'k2':2,'k3':3}for item in d: print(item)>> k1>> k2>> k3# Create a dictionary view objectd.items()# Dictionary unpackingfor k,v in d.items(): v = v+1 print(k) print(v) >> k1>> 2>> k.

a = 10type(a)a = 'I said :"stupid! 'type(a)>> strprint('Use
to print a new line')s = 'Hello World'# Grab everything but the last letters[:-1]# Concatenate strings!s + ' concatenate me!'#We can use the multiplication symbol to create r

3If you look into the original ResNet Paper (http://openaccess.thecvf.com/content_cvpr_2016/papers/He_Deep_Residual_Learning_CVPR_2016_paper.pdf) they use strided convolutions to downsample the image. The main path is downsampled automatically using these

Abstracts:DenseNet breaks away from the fixed thinking of deepening the number of network layers (ResNet) and widening the network structure (Inception) to improve network performance. From the perspective of features, through feature reuse and bypass (By

An article on batch processing, detailing the specific advantages of using BN1. The relationship between weight initialization and preprocessing methods in neural networksIf you have done the dnn experiment, you may find that preprocessing the data, such

Now, Residual Connections are everywhere: Not only in image recognition, they’re in transformers.Problem: people knew if you can increase depth pof a neural network, you can make it perform better, make it generalize better, you can reach lower training

We need to report testset performanceWhy validation set is not enough and we need a seperate test set?We don’t strictly need a test set, it might be okay if some cases to have training and validation sets onluy, the purpose of the test set is to help us

Classifying images from Fashion MNIST using feedforward neural networksimport torchimport torchvisionimport numpy as npimport matplotlib.pyplot as pltimport torch.nn as nnimport torch.nn.functional as Ffrom torchvision.datasets import FashionMNISTf

import torchimport torchvisionimport numpy as npimport matplotlib.pyplot as pltimport torch.nn as nnimport torch.nn.functional as Ffrom torchvision.datasets import MNISTfrom torchvision.transforms import ToTensorfrom torchvision.utils import make_g

Image Classification with Logistic Regression Code# Uncomment and run the commands below if imports fail# !conda install numpy pytorch torchvision cpuonly -c pytorch -y# !pip install matplotlib --upgrade --quiet!pip install jovian --upgrade --quiet#

Working with MNIST & Logistic Regression in PyTorch# Importsimport torchimport torchvisionfrom torchvision.datasets import MNIST# Download training datasetdataset = MNIST(root='data/', download=True)When this statement is executed for the firs

ViewController.swiftimport UIKitimport WebKitclass ViewController: UIViewController, WKNavigationDelegate { var webView: WKWebView! override func loadView() { webView = WKWebView() webView.navigationDelegate = self

Laplace Approximation:Fit a Gaussian to p(w|D)log p(w|D) = log p(w,D) +const wrt. wQuadratic in w if GaussianFind mode and fund 2nd derivative“Energy” E(w) = -log( p(w,D) )w* = argmin E(w) (L2 regularization term or MAP fit)Hessian Hi

import numpy as np import pandas as pd import osimport matplotlib.pyplot as plimport seaborn as snsimport warningsdata = pd.read_csv('/kaggle/input/vehicle-dataset-from-cardekho/car data.csv')import torchinputs = np.asarray(data.values[:,1:3], dty

#full codeimport numpy as npall_means=[]all_stds=[]for i in range(1,6): stat = np.loadtxt(str(i)+'.txt',skiprows=1) print('Layer',i) means =[] stds=[] for j in range(5): cover = stat[100*j:100*(j+1),-1] steps = stat.

ultrasonic sensor :gpio wiring:https://thepihut.com/blogs/raspberry-pi-tutorials/hc-sr04-ultrasonic-range-sensor-on-the-raspberry-pipython code:https://tutorials-raspberrypi.com/raspberry-pi-ultrasonic-sensor-hc-sr04/

convert python2 to python3: (.ipynb)nbconvertconvert .ipynb to markdown,html,pdf

follow https://blog.csdn.net/lclfans1983/article/details/108238796#1.%20%E7%83%A7%E5%BD%95Jetson%20Nano%E9%95%9C%E5%83%8FWhen pytorch: 1. install other necessary packages: search from this:https://ubuntu.pkgs.org/18.04/ubuntu-universe-amd64/libopenblas-d

sudo python blinking_lcd (25green,23,24 )import timeimport RPi.GPIO as GPIO# define LED pinled_pin = 25# set GPIO mode to GPIO.BOARDGPIO.setmode(GPIO.BCM)# set puin as inputGPIO.setup(led_pin, GPIO.OUT)try: while True: # turn on LED

Youtube Explorer Hat Motorhardware wiring (ppt) with code of the youtube aboveCode:python motor.py>>> import RPi.GPIO as IO>>> import time>>> from time import sleep>>> IO.setwarnings(False)>>> IO.setmod

2004 LCD software part: 2004 LCD Hardware Wiring YoutubeImportant how to test method/class ina python file:>>> import os>>> os.chdir('./')>>> import lcd>>> lcd.lcd_init()>>> lcd.lcd_byte(lcd.LCD_LINE_1,lc

http://ubuntuhandbook.org/index.php/2020/07/install-qt4-ubuntu-20-04/

This is by far the most creative game I seen so far:https://samhogan.itch.io/cell-machine

1. try different landscapetaller hills : to get up to taller hills, one need to have momentum by first go into the lower hills and use inertia–> advanced thinking, need more levels of layertry plot grid on with each bin, notice how frequent the robot

how to change Mac read-only file permission:Restart in Recovery mode (Command-R at startup), open Terminal (from the Utilities menu), and disable SIP filesystem protection with:csrutil enable --without fsRestart normally, open Terminal, and remount the

https://github.com/deepmind/open_spiel/blob/master/docs/install.mdI use a docker file—activate with:docker run -it --entrypoint /bin/bash openspiel(edc25d1fd803 / openspiel:latest)

brew tap nagakiran/depsbrew tap brewsci/scienceThe location of user’s packages folder can be found using:python -m site --user-site

clang -dynamiclib -o libtest.dylib file1.o file2.o -L/some/library/path -lname_of_library_without_lib_prefix

pydrake package is installed in /opt/drakeconda activate python3.8export PYTHONPATH=/opt/drake/lib/python3.8/site-packages:$PYTHONPATHpython3 -c ‘import pydrake; print(pydrake.file)’https://drake.mit.edu/python_bindings.html#python-bindings-binary...

website:https://docs.github.com/en/github/authenticating-to-github/generating-a-new-ssh-key-and-adding-it-to-the-ssh-agenthttps://docs.github.com/en/github/authenticating-to-github/adding-a-new-ssh-key-to-your-github-accountcommon issue:Users/jerry/.ss

https://github.com/miyosuda/evolution_and_ai/tree/master/multi_agent_emergenceMy conda env: hide-and-seekAlso,It looks like a recent release of pyglet cloud break gym. I downgraded pyglet to 1.3.2 for the colabs, and it should work now.

tf1 to tf2:http://tf2up.mlconda to requirement.txtconda list -e > requirements.txtconda delete envconda env remove --name myenvmy virtual environment:PARL—for PARL high level RL (Tensorflow, gym)pygame – customised game (not have gym–)openai (t

values = np.array([Q[state,a] for a in actions])don’t forget this [] inside np.array()!!

Ipad :raspivid -o Desktop/video.h264 -t 600000 (600s)macbookopen quick time player --> movie recording --> record -->name the fileimac1. launchroscoreroslaunch turtlebot3_bringup turtlebot3_robot.launch2. run algorithmcatkin_makesource

google colab on yolov4:https://darknet.gong.im/

https://github.com/hunglc007/tensorflow-yolov4-tflitechange xxxx.names at this config file:https://github.com/hunglc007/tensorflow-yolov4-tflite/blob/master/core/config.py#L14Codeconda activate yolov4-tflitepython save_model.py --weights ./data/y

because guilogger is using qt4 to compile and only qt4 can have qt3 support which is also needed.1.chage the version of your qt to qt4 (/usr/bin or usr/local)2.we need to install guilogger from a new directory and using original gui logger package, since

Install ode (open dynamic engine) first !!!—- latest(2017) version

yolov4 有的dataset课程的ball messi—用conv.137 transfer learningdrone github的头盔。csdn搜到的

Difficulty:After installing tensorflow using virtualenv, we cannot install opencv in the same virtual environment using:pip3 install python3-opencvwill result in command not foundthen I searched pypip.com for wheel for opencv with aarch64, result in