Skip to content

Instantly share code, notes, and snippets.

💭
architecting

Dillon Erb dte

View GitHub Profile
View releaseTest.md

Release Notes

New Thing 1

7/16/2018

About this

New Thing 2

7/14/2018

About this

@dte
dte / loopy_cosine_similarities.py
Created May 9, 2018
Vectorization and Broadcasting with Pytorch
View loopy_cosine_similarities.py
import torch
from torch.nn.functional import cosine_similarity
def embeddings_to_cosine_similarity_matrix(E):
"""
Converts a a tensor of n embeddings to an (n, n) tensor of similarities.
"""
similarities = [[cosine_similarity(a, b, dim=0) for a in E] for b in E]
similarities = list(map(torch.cat, similarities))
return torch.stack(similarities)
View serve.py
import os
from flask import Flask, redirect, url_for, request, render_template, send_from_directory
from werkzeug import secure_filename
import not_hotdog_model
# todo: more pretty interface
# folder to upload pictures
UPLOAD_FOLDER = 'uploads/'
# what files can upload
View Adversarial2.py
torch.manual_seed(10)
Q, P = Q_net() = Q_net(), P_net(0) # Encoder/Decoder
D_gauss = D_net_gauss() # Discriminator adversarial
if torch.cuda.is_available():
Q = Q.cuda()
P = P.cuda()
D_cat = D_gauss.cuda()
D_gauss = D_net_gauss().cuda()
# Set learning rates
gen_lr, reg_lr = 0.0006, 0.0008
View Adversarial.py
#Encoder
class Q_net(nn.Module):
def __init__(self):
super(Q_net, self).__init__()
self.lin1 = nn.Linear(X_dim, N)
self.lin2 = nn.Linear(N, N)
self.lin3gauss = nn.Linear(N, z_dim)
def forward(self, x):
x = F.droppout(self.lin1(x), p=0.25, training=self.training)
x = F.relu(x)
@dte
dte / get_available_gpus.py
Created Nov 16, 2017 — forked from jovianlin/get_available_gpus.py
Get List of Devices in TensorFlow
View get_available_gpus.py
from tensorflow.python.client import device_lib
def get_available_gpus():
local_device_protos = device_lib.list_local_devices()
return [x.name for x in local_device_protos if x.device_type == 'GPU']
get_available_gpus()
@dte
dte / tensorflow_opencv_ubuntu_deps.sh.txt
Created Jul 25, 2017 — forked from CapCap/tensorflow_opencv_ubuntu_deps.sh.txt
Paperspace tensorflow+opencv setup for both python2 and python3 on ubuntu 16
View tensorflow_opencv_ubuntu_deps.sh.txt
#!/bin/bash
# Don't require you to constantly enter password for sudo:
sudo visudo
# In the bottom of the file, paste the following (without the `#`):
# paperspace ALL=(ALL) NOPASSWD: ALL
# Then press `ctl+o` then `enter` to save your changes, and `ctr+x` to exit nano
# Allow connection from your IP to any port- default seems to be just 22 (ssh)
@dte
dte / install-CUDA-docker-nvidia-docker.sh
Created Jul 19, 2017
Install CUDA, Docker, and Nvidia Docker on a new Paperspace GPU machine
View install-CUDA-docker-nvidia-docker.sh
#!/bin/bash
# 1. Install CUDA
echo "Installing CUDA..."
# Only install if CUDA is not already installed.
if ! dpkg-query -W cuda; then
# The 16.04 installer works with 16.10.
curl -O http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/cuda-repo-ubuntu1604_8.0.61-1_amd64.deb
dpkg -i ./cuda-repo-ubuntu1604_8.0.61-1_amd64.deb
apt-get update
apt-get install cuda -y
@dte
dte / install-cuda.sh
Created Jul 19, 2017
Install CUDA 8 on Ubuntu 16.04 / 16.10
View install-cuda.sh
#!/bin/bash
echo "Installing CUDA..."
# Only install if CUDA is not already installed.
if ! dpkg-query -W cuda; then
# The 16.04 installer works with 16.10.
curl -O http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/cuda-repo-ubuntu1604_8.0.61-1_amd64.deb
dpkg -i ./cuda-repo-ubuntu1604_8.0.61-1_amd64.deb
apt-get update
apt-get install cuda -y
fi
View html5-video-streamer.js
/*
* Inspired by: http://stackoverflow.com/questions/4360060/video-streaming-with-html-5-via-node-js
*/
var http = require('http'),
fs = require('fs'),
util = require('util');
http.createServer(function (req, res) {
var path = 'video.mp4';
You can’t perform that action at this time.