mjdietzx

'''Trains a simple convnet on the MNIST dataset.
Gets to 99.25% test accuracy after 12 epochs
(there is still a lot of margin for parameter tuning).
16 seconds per epoch on a GRID K520 GPU.
'''

from __future__ import print_function
import numpy as np
np.random.seed(1337)  # for reproducibility

mjdietzx

#!/bin/bash

# install CUDA Toolkit v8.0
# instructions from https://developer.nvidia.com/cuda-downloads (linux -> x86_64 -> Ubuntu -> 16.04 -> deb (network))
CUDA_REPO_PKG="cuda-repo-ubuntu1604_8.0.61-1_amd64.deb"
wget http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/${CUDA_REPO_PKG}
sudo dpkg -i ${CUDA_REPO_PKG}
sudo apt-get update
sudo apt-get -y install cuda

mjdietzx

#!/bin/bash

# creates a GitHub release (draft) and adds pre-built artifacts to the release
# after running this script user should manually check the release in GitHub, optionally edit it, and publish it

# args: :version_number (the version number of this release), :body (text describing the contents of the tag)
# example usage: ./gh_release_bamboo.sh "1.0.0" "Release notes: ..."
# => name: nRF5-ble-driver_<platform_name>_1.0.0_compiled-binaries.zip example: nRF5-ble-driver_win-64_2.0.1_compiled-binaries.zip

# to ensure that bash is used: https://answers.atlassian.com/questions/28625/making-a-bamboo-script-execute-using-binbash

mjdietzx

# http://www.nvidia.com/download/driverResults.aspx/117079/en-us
wget http://us.download.nvidia.com/tesla/375.51/nvidia-driver-local-repo-ubuntu1604_375.51-1_amd64.deb
sudo dpkg -i nvidia-driver-local-repo-ubuntu1604_375.51-1_amd64.deb
sudo apt-get update
sudo apt-get -y install cuda-drivers

echo "Reboot required."

mjdietzx

"""
Clean and simple Keras implementation of network architectures described in:
    - (ResNet-50) [Deep Residual Learning for Image Recognition](https://arxiv.org/pdf/1512.03385.pdf).
    - (ResNeXt-50 32x4d) [Aggregated Residual Transformations for Deep Neural Networks](https://arxiv.org/pdf/1611.05431.pdf).
    
Python 3.
"""

from keras import layers
from keras import models

mjdietzx

from keras import layers


def residual_block(y, nb_channels, _strides=(1, 1), _project_shortcut=False):
    shortcut = y

    # down-sampling is performed with a stride of 2
    y = layers.Conv2D(nb_channels, kernel_size=(3, 3), strides=_strides, padding='same')(y)
    y = layers.BatchNormalization()(y)
    y = layers.LeakyReLU()(y)

mjdietzx

from keras import layers
from keras import models
import tensorflow as tf


#
# generator input params
#

rand_dim = (1, 1, 2048)  # dimension of the generator's input tensor (gaussian noise)

mjdietzx

import torch
from torch.autograd import Variable
import torch.nn as nn


class Bottleneck(nn.Module):
    cardinality = 32  # the size of the set of transformations

    def __init__(self, nb_channels_in, nb_channels, nb_channels_out, stride=1):
        super().__init__()

mjdietzx

"""
wGAN implemented on top of tensorflow as described in: [Wasserstein GAN](https://arxiv.org/pdf/1701.07875.pdf)
with improvements as described in: [Improved Training of Wasserstein GANs](https://arxiv.org/pdf/1704.00028.pdf).

"""

import tensorflow as tf


#

mjdietzx

import torch
from torch import autograd
from torch import nn


class CrossEntropyLoss(nn.Module):
    """
    This criterion (`CrossEntropyLoss`) combines `LogSoftMax` and `NLLLoss` in one single class.
    
    NOTE: Computes per-element losses for a mini-batch (instead of the average loss over the entire mini-batch).