drscotthawley

#!/usr/bin/env python3

# Do violence to the Logitech C920s webcam settings in realtime
# To annoy your Zoom-mates by constantly changing the image

from pynput.keyboard import Listener
import os, sys
import random
import time

drscotthawley

# In case you didn't think to add k-fold cross-validation until late in your
# ML project,...
# This is built for a situation where datasets are arrays of, say, images. 

def kfold_swap(train_X, train_Y, val_X, val_Y, k):
    """
    Swaps val with a section of train, given a value for k
    "Duct tape" approach used to "retro-fit" k-fold cross-validation while minimally 
        disturbing the rest of the code, while avoiding reloading data from disk and 
        keeping RAM use manageable. (e.g. np.append() is bad b/c it would copy all of train)

drscotthawley

#! /usr/bin/env python
'''
Simulation of a 'magnetic' pendulum. Actually we'll just use
electrostatic charges instead of magnets, but..good enough, right?

Plots an image showing which source the object is closest to after a certain time
(Note: Depending on params like maxiter, the object may still be moving at the
end of the simulation, so the 'ending position' may not be where it comes to rest.)

This code integrates all the (non-interacting) test objects at once, on the GPU.

drscotthawley

#! /usr/bin/env python

# Little script I use to start a new Markdown blog entry.
# Run from the parent directory above the blog directory,
# or supply a full path to run from anywhere.
#
# Usage:  newpost.py <title>
#  It automatically figures out what the current date is to
#  create a new entry, and supplies a default header with a title
#  and a bibliography placement

drscotthawley

diff --git a/src/core/lib/gpr/log_linux.cc b/src/core/lib/gpr/log_linux.cc
index 561276f0c2..1af0935e1f 100644
--- a/src/core/lib/gpr/log_linux.cc
+++ b/src/core/lib/gpr/log_linux.cc
@@ -40,7 +40,7 @@
 #include <time.h>
 #include <unistd.h>
 
-static long gettid(void) { return syscall(__NR_gettid); }
+static long sys_gettid(void) { return syscall(__NR_sys_gettid); }

drscotthawley

#! /usr/bin/env python3
# Test script for Phase 'Unrolling' / Instantaneous frequency
#
# See Jesse Engel's "rainbowgrams" script, https://gist.github.com/jesseengel/e223622e255bd5b8c9130407397a0494
#
# Modifications by Scott H. Hawley, @drscotthawley and Billy Mitchell
# These modified versions seem to be both more accurate (~2000x less reconstruction error)
# and faster (>20%)
#
# note, to really see/hear the difference, change dtypes to np.float16!

drscotthawley

#! /usr/env/python3
#
#  FMA conversion script,  genre classification
#  Author: Scott Hawley
#  License: Do as you like
#
# For FMA dataset https://github.com/mdeff/fma
# to be used with panotti https://github.com/drscotthawley/panotti
#
# This will create a directory called Samples/

drscotthawley

#! /usr/bin/env python3
'''
osc2wek.py
Author: Scott Hawley

This listens for incoming OSC messages and sends them on to Wekinator

Steps to get running (in Terminal):
0. First you need Mercurial "hg".  It might be installed by default.
1. Use hg to clone the grail osc code:

drscotthawley

import numpy as np

def get_1cycle_schedule(lr_max=1e-3, n_data_points=8000, epochs=200, batch_size=40, verbose=0):          
  """
  Creates a look-up table of learning rates for 1cycle schedule with cosine annealing
  See @sgugger's & @jeremyhoward's code in fastai library: https://github.com/fastai/fastai/blob/master/fastai/train.py
  Wrote this to use with my Keras and (non-fastai-)PyTorch codes.
  Note that in Keras, the LearningRateScheduler callback (https://keras.io/callbacks/#learningratescheduler) only operates once per epoch, not per batch
      So see below for Keras callback
      

drscotthawley