hkxIron

""" Trains an agent with (stochastic) Policy Gradients on Pong. Uses OpenAI Gym. """
import numpy as np
import cPickle as pickle
import gym

# hyperparameters
H = 200 # number of hidden layer neurons
batch_size = 10 # every how many episodes to do a param update?
learning_rate = 1e-4
gamma = 0.99 # discount factor for reward

hkxIron

package bbejeck.grouping

import org.apache.log4j.{Level, Logger}
import org.apache.spark.{SparkConf, SparkContext}

import scala.collection.mutable

/**
 * Created by bbejeck on 8/6/15.
 * Example usage of combineByKey

hkxIron

package bbejeck.grouping

import org.apache.log4j.{Level, Logger}
import org.apache.spark.{SparkConf, SparkContext}

import scala.collection.mutable

/**
 * Created by bbejeck on 7/31/15.
 *

hkxIron

# XLA compilation controlled by "compile_ops" option
# compile_ops=False: 4.39 sec
# compile_ops=True: 0.90 sec


import os
os.environ['CUDA_VISIBLE_DEVICES']=''
import tensorflow as tf

hkxIron

import org.apache.spark.{AccumulableParam, SparkConf}
import org.apache.spark.serializer.JavaSerializer
import scala.collection.mutable.{ HashMap => MutableHashMap }

/*
 * Allows a mutable HashMap[String, Int] to be used as an accumulator in Spark.
 * Whenever we try to put (k, v2) into an accumulator that already contains (k, v1), the result
 * will be a HashMap containing (k, v1 + v2).
 *
 * Would have been nice to extend GrowableAccumulableParam instead of redefining everything, but it's

hkxIron

// This program converts a set of vector<float>'s to a lmdb/leveldb by storing
// them as Datum proto buffers.
// Usage:
//   convert_vector [FLAGS] LISTFILE DB_NAME
//
// where LISTFILE should be a list of files as well as the accompanying vector
// of floats, in the format as:
//   subfolder1/file1.JPEG 0.2 0.3 0.1 0.25 0.15
//   ....