hiropppe

FROM ubuntu:16.04

RUN mkdir /root/_INSTALL

WORKDIR /root/_INSTALL

RUN sed -i.bak -e "s%http://archive.ubuntu.com/ubuntu/%http://ftp.jaist.ac.jp/pub/Linux/ubuntu/%g" /etc/apt/sources.list

ENV TZ Asia/Tokyo
RUN apt-get update \

hiropppe

[2018.6.11]
3  sudo apt update
4  sudo apt upgrade
6  sudo apt install build-essential
11  wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/cuda-repo-ubuntu1604_9.2.88-1_amd64.deb
12  sudo dpkg -i cuda-repo-ubuntu1604_9.2.88-1_amd64.deb
13  sudo apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/7fa2af80.pub
14  sudo apt-get update
15  sudo apt-get install cuda
16  nvidia-smi

hiropppe

# -*- coding:utf-8 -*-

import codecs
import json
import os
import sys


if __name__ == '__main__':
    def walk_json(path):

hiropppe

# cython: boundscheck = False
# cython: wraparound = False
# cython: cdivision = True

import numpy as np

cimport numpy as np

from scipy.linalg cimport cython_blas

hiropppe

from cython.parallel import prange

from libc.stdio cimport printf
from libc.stdlib cimport abort, malloc, free

cimport openmp


def get_int_flag():
    cdef int flag = 0

hiropppe

'''
Synchronous Distributed Tensorflow 0.12.0 example of using data parallelism and share model parameters.
Trains a simple sigmoid neural network on mnist for 20 epochs on three machines using one parameter server.

Change the hardcoded host urls below with your own hosts.
Run like this:

pc-01$ python synchronous_distributed_mnist_training.py --job_name="ps" --task_index=0
pc-02$ python synchronous_distributed_mnist_training.py --job_name="worker" --task_index=0
pc-03$ python synchronous_distributed_mnist_training.py --job_name="worker" --task_index=1

hiropppe

'''
Asynchronous Distributed Tensorflow 0.12.0 example of using data parallelism and share model parameters.
Trains a simple sigmoid neural network on mnist for 20 epochs on three machines using one parameter server.

Change the hardcoded host urls below with your own hosts.
Run like this:

pc-01$ python asynchronous_distributed_mnist_training.py --job_name="ps" --task_index=0
pc-02$ python asynchronous_distributed_mnist_training.py --job_name="worker" --task_index=0
pc-03$ python asynchronous_distributed_mnist_training.py --job_name="worker" --task_index=1

hiropppe

policy gradient methods for reinforcement learning

references

1. Deep Reinforcement Learning: Pong from Pixels
2. 深層強化学習：ピクセルから『ポン』 – 前編
3. 深層強化学習：ピクセルから『ポン』 – 後編
4. Simple Reinforcement Learning with Tensorflow: Part 2 - Policy-based Agents
5. Simple reinforcement learning methods to learn CartPole
6. Pong by karpathy
7. CartPole by christopherhesse

hiropppe

import tables as tb

from scipy.sparse import lil_matrix

h5 = tb.open_file('sparse.h5', 'a')

data_group = h5.create_group(h5.root, 'data_group')
indices_group = h5.create_group(h5.root, 'indices_group')
indptr_group = h5.create_group(h5.root, 'indptr_group')

hiropppe

$ mkdir -p ~/.vim/bundle
$ git clone https://github.com/Shougo/neobundle.vim ~/.vim/bundle/neobundle.vim


"------------------------------------
" neocomplete.vim
"------------------------------------
"Note: This option must set it in .vimrc(_vimrc).  NOT IN .gvimrc(_gvimrc)!
" Disable AutoComplPop.
let g:acp_enableAtStartup = 0