johannah

# to get mujoco-py to install with headless rendering without sudo access

##############################
# 1) first download and install mujoco according to instructions from https://github.com/deepmind/mujoco/

##############################
# 2) add environment variables to your bashrc
# importantly, this should be done before installing mujoco-py
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/home/$USER/.mujoco/mujoco210/bin
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/lib/nvidia

johannah

'''
This script saves each topic in a bagfile as a csv.
Accepts a filename as an optional argument. Operates on all bagfiles in current directory if no argument provided

Originally Written by Nick Speal in May 2013 at McGill University's Aerospace Mechatronics Laboratory

Modified by J.Hansen in 2021
'''

import rosbag, sys, csv

johannah

import numpy as np
import os
import sys

# hours_per_time_step should be greater than 1       
hours_per_step = 2

class PainModel():
    def __init__(self, seed):
        # pain -1 is bad

johannah

# from KK
from copy import deepcopy
import time
import os
import torch 
import torch.nn as nn
from torch.nn.utils.clip_grad import clip_grad_norm
import torch.nn.functional as F
import torch.optim as optim
from torch.autograd import Variable

johannah

# from KK
import torch 
from torch.autograd import Variable
import numpy as np
import matplotlib.pyplot as plt
import torch.nn.init as init
from IPython import embed

dtype = torch.FloatTensor
input_size, hidden_size, output_size = 1,128,1

johannah

import os, sys
from glob import glob
from IPython import embed
from subprocess import call
# example <a href='catalog.html?dataset=NOAA/CBOFS/MODELS/201710/nos.cbofs.stations.nowcast.20171031.t18z.nc'><tt>nos.cbofs.stations.nowcast.20171031.t18z.nc</tt></a></td>^M
fpaths = glob('/localdata/jhansen/thredds/opendap.co-ops.nos.noaa.gov/thredds/catalog/NOAA/*')
output = '/localdata/jhansen/thredds/ncfiles'
fileServer = 'https://opendap.co-ops.nos.noaa.gov/thredds/fileServer/'
for station in fpaths:
    # only get stations which have fairly open oceans

johannah

name: py-ros
channels:
- defaults
dependencies:
- backports=1.0=py27h63c9359_1
- backports.shutil_get_terminal_size=1.0.0=py27h5bc021e_2
- ca-certificates=2017.08.26=h1d4fec5_0
- certifi=2017.7.27.1=py27h9ceb091_0
- decorator=4.1.2=py27h1544723_0
- enum34=1.1.6=py27h99a27e9_1

johannah

import matplotlib.pyplot as plt
import numpy as np

# example taken from Jake Vanderplas' talk "Statistics for Hackers"

def is_statistically_significant_shuffle(s1,s2,trials=10000,do_plot=False,null_hyp=.05):
    # find difference of means to test null hyp
    odiff = np.mean(s1)-np.mean(s2)
    # join the lists together
    a = np.array(s1+s2)

johannah

Dataset Compass

An entry level description of various datasets and their accessibility.

Very easy to access:

Many examples, but kind of boring

Datasets housed within sklearn - http://scikit-learn.org/stable/datasets/index.html
(examples - mnist number images, boston housing prices, iris types, Titantic survivors)

Moderately easy to access:

Interesting, likely with example code somewhere on the internet

johannah

import numpy
import time
import matplotlib.pyplot as plt
import numpy as np
from copy import deepcopy
import os, sys
from scipy.misc import face
import GPy
from skimage.transform import resize
from skimage.filters import gaussian_filter