Hoemas Ruud tomas-wood

Make sure you have nvidia-docker installed (I've got this covered on 35 for all users, I believe)
Run nvidia-docker run -it odellus/(sg_srvc:version1|detectron_srvc:version5) and it will take you inside the docker container.
When you get finished with the work you're doing, don't just exit, open a new ssh connection and commit the changes you've made using docker commit (here is how to do it) by entering

nvidia-docker ps (get the container id this way or just read it off the beginning of bash prompt in other terminal)
nvidia-docker commit <container id> odellus/(sg_srvc:version1|detectron_srvc:verstion5)
nvidia-docker push

It will probably ask you for my password, which I can send you through eSpace messenger.

GraphML Links!!!

Articles:
https://medium.com/octavian-ai/an-introduction-to-machine-learning-on-graph-databases-24ee502fd12e

Code:
https://github.com/Octavian-ai/article-0

Partitioning Graph for Mini-Batch Approximate Graph Convolutional Networks (GCNs)

So we want to be able to create partitions of the graph which we can perform graph convolutions on without having to load the entire dataset into memory.

Terms: A, the adjacency matrix of the graph (normalized and with identity matrix added to it, but still A). With N nodes in the graph, A is N X N.
X, a dense matrix of features such that each feature is of dimension F, then for all N nodes we get a dense matrix of size N X F

In the GCN paper and code, the entire dataset and the entire adjacency matrix are loaded into memory. The weight updates are done by

	+-----------------------------------------------------------------------------+
	\| Processes: GPU Memory \|
	\| GPU PID Type Process name Usage \|
	\|=============================================================================\|
	\| 0 5760 C /usr/bin/python2.7 11739MiB \|
	\| 1 5760 C /usr/bin/python2.7 11589MiB \|
	\| 2 5760 C /usr/bin/python2.7 11589MiB \|
	\| 3 5760 C /usr/bin/python2.7 11589MiB \|
	\| 4 5760 C /usr/bin/python2.7 11589MiB \|
	\| 5 5760 C /usr/bin/python2.7 11589MiB \|

	#!/usr/bin/env python2
	# -- coding: utf-8 --

	################################################################
	#
	# This script takes an image as input and outputs the
	# region proposals
	#
	# Author: Aniket Adnaik ([email protected])
	################################################################

	from __future__ import absolute_import
	from __future__ import division
	from __future__ import print_function
	from __future__ import unicode_literals

	from collections import defaultdict
	import argparse
	import cv2 # NOQA (Must import before importing caffe2 due to bug in cv2)
	import glob
	import logging

	import urllib3
	import pickle
	import json

	http = urllib3.PoolManager()

	url = "http://localhost:5000/detectron"

	img_url="https://cdn7.dissolve.com/p/D1028_98_301/D1028_98_301_1200.jpg"

	Wed Sep 19 11:48:26 2018
	+-----------------------------------------------------------------------------+
	\| NVIDIA-SMI 396.37 Driver Version: 396.37 \|
	\|-------------------------------+----------------------+----------------------+
	\| GPU Name Persistence-M\| Bus-Id Disp.A \| Volatile Uncorr. ECC \|
	\| Fan Temp Perf Pwr:Usage/Cap\| Memory-Usage \| GPU-Util Compute M. \|
	\|===============================+======================+======================\|
	\| 0 TITAN X (Pascal) Off \| 00000000:04:00.0 Off \| N/A \|
	\| 23% 40C P2 50W / 250W \| 893MiB / 12196MiB \| 0% Default \|
	+-------------------------------+----------------------+----------------------+

	curl -H "Content-Type:application/json" -X POST --data '{"command":"graph=EywaGraph.open(\"ges_abfe\");g=graph.traversal();g.V().limit(10)"}' http://49.4.88.7/graph/ges/v1.0/1111/1/graphs/222/execute-gremlin-query
	{"errorMessage":"Resource not allowed!","errorCode":"GES.8006"}

	def fetch_data():
	urls = [
	"https://visualgenome.org/static/data/dataset/objects.json.zip",
	"https://visualgenome.org/static/data/dataset/relationships.json.zip",
	"https://visualgenome.org/static/data/dataset/object_alias.txt",
	"https://visualgenome.org/static/data/dataset/relationship_alias.txt"
	]
	if not os.path.exists('data'):
	os.mkdir('data')
	else: