深度学习框架 Digits 3.0 安装运行
来源:互联网 发布:中小学生出作业软件 编辑:程序博客网 时间:2024/06/05 19:01
NVIDIA 不愧是推动Deep learning 的中坚力量,之前运行2.0版本正得心应手时,就推出了3.0 版本。Github地址:https://github.com/NVIDIA/DIGITS,3.0版本的安装使用更为简洁,极易上手。
更新内容:
The new DIGITS 3 release improves training productivity with enhanced workflows.
- Train neural network models with Torch support (preview), including pre-built AlexNet and GoogleNet models that make it easy to get started.
- Quickly identify the best model through focused design iteration using the new results browser.
- Easily manage multiple training jobs to optimize use of system resources.
- Explore image datasets using the new dataset browser contributed to the DIGITS open source project by Deepomatic.
操作系统: Ubuntu 14.04 64bit.
首先需要配置Caffe 没有配置的同学自行官网。
命令行:
CUDA_REPO_PKG=cuda-repo-ubuntu1404_7.5-18_amd64.deb &&
wget http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1404/x86_64/$CUDA_REPO_PKG &&
sudo dpkg -i $CUDA_REPO_PKG
ML_REPO_PKG=nvidia-machine-learning-repo_4.0-2_amd64.deb &&
wget http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1404/x86_64/$ML_REPO_PKG &&
sudo dpkg -i $ML_REPO_PKG
wget http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1404/x86_64/$CUDA_REPO_PKG &&
sudo dpkg -i $CUDA_REPO_PKG
ML_REPO_PKG=nvidia-machine-learning-repo_4.0-2_amd64.deb &&
wget http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1404/x86_64/$ML_REPO_PKG &&
sudo dpkg -i $ML_REPO_PKG
待添加完后同样命令行:
apt-get update
apt-get install digits
apt-get install digits
泡杯Caffe,然后等待安装完成。
安装完成后,在浏览器中输入:
http://localhost/
就可以见到亲切的DIGITS界面了,然后进行train吧。
PS:DIGITS开机自启动,同在一个局域网的机器也可以远程操作。同样在浏览器中输入 运行DIGITS的主机IP地址,同样进入。Amazing~
def vis_square(images, padsize=1, normalize=False, colormap='jet', ): """ Visualize each image in a grid of size approx sqrt(n) by sqrt(n) Returns a np.array image (Based on Caffe's filter_visualization notebook) Arguments: images -- an array of shape (N, H, W) or (N, H, W, C) if C is not set, a heatmap is computed for the result Keyword arguments: padsize -- how many pixels go inbetween the tiles normalize -- if true, scales (min, max) across all images out to (0, 1) colormap -- a string representing one of the suppoted colormaps """ assert 3 <= images.ndim <= 4, 'images.ndim must be 3 or 4' # convert to float since we're going to do some math images = images.astype('float32') if normalize: images -= images.min() if images.max() > 0: images /= images.max() images *= 255 if images.ndim == 3: # they're grayscale - convert to a colormap redmap, greenmap, bluemap = get_color_map(colormap) red = np.interp(images*(len(redmap)-1)/255.0, xrange(len(redmap)), redmap) green = np.interp(images*(len(greenmap)-1)/255.0, xrange(len(greenmap)), greenmap) blue = np.interp(images*(len(bluemap)-1)/255.0, xrange(len(bluemap)), bluemap) # Slap the channels back together images = np.concatenate( (red[...,np.newaxis], green[...,np.newaxis], blue[...,np.newaxis]), axis=3 ) images = np.minimum(images,255) images = np.maximum(images,0) # convert back to uint8 images = images.astype('uint8') # Compute the output image matrix dimensions n = int(np.ceil(np.sqrt(images.shape[0]))) ny = n nx = n length = images.shape[0] if n*(n-1) >= length: nx = n-1 # Add padding between the images padding = ((0, nx*ny - length), (0, padsize), (0, padsize)) + ((0, 0),) * (images.ndim - 3) padded = np.pad(images, padding, mode='constant', constant_values=255) # Tile the images beside each other tiles = padded.reshape( (ny, nx) + padded.shape[1:]).transpose( (0,2,1,3) + tuple(range(4, padded.ndim + 1))) tiles = tiles.reshape((ny * tiles.shape[1], nx * tiles.shape[3]) + tiles.shape[4:]) return tilesdef get_color_map(name): """ Return a colormap as (redmap, greenmap, bluemap) Arguments: name -- the name of the colormap. If unrecognized, will default to 'jet'. """ redmap = [0] greenmap = [0] bluemap = [0] if name == 'white': # essentially a noop redmap = [0,1] greenmap = [0,1] bluemap = [0,1] elif name == 'simple': redmap = [0,1,1,1] greenmap = [0,0,1,1] bluemap = [0,0,0,1] elif name == 'hot': redmap = [0, 0.03968253968253968, 0.07936507936507936, 0.119047619047619, 0.1587301587301587, 0.1984126984126984, 0.2380952380952381, 0.2777777777777778, 0.3174603174603174, 0.3571428571428571, 0.3968253968253968, 0.4365079365079365, 0.4761904761904762, 0.5158730158730158, 0.5555555555555556, 0.5952380952380952, 0.6349206349206349, 0.6746031746031745, 0.7142857142857142, 0.753968253968254, 0.7936507936507936, 0.8333333333333333, 0.873015873015873, 0.9126984126984127, 0.9523809523809523, 0.992063492063492, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1] greenmap = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.03174603174603163, 0.0714285714285714, 0.1111111111111112, 0.1507936507936507, 0.1904761904761905, 0.23015873015873, 0.2698412698412698, 0.3095238095238093, 0.3492063492063491, 0.3888888888888888, 0.4285714285714284, 0.4682539682539679, 0.5079365079365079, 0.5476190476190477, 0.5873015873015872, 0.6269841269841268, 0.6666666666666665, 0.7063492063492065, 0.746031746031746, 0.7857142857142856, 0.8253968253968254, 0.8650793650793651, 0.9047619047619047, 0.9444444444444442, 0.984126984126984, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1] bluemap = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.04761904761904745, 0.1269841269841265, 0.2063492063492056, 0.2857142857142856, 0.3650793650793656, 0.4444444444444446, 0.5238095238095237, 0.6031746031746028, 0.6825396825396828, 0.7619047619047619, 0.8412698412698409, 0.92063492063492, 1] elif name == 'rainbow': redmap = [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0.9365079365079367, 0.8571428571428572, 0.7777777777777777, 0.6984126984126986, 0.6190476190476191, 0.53968253968254, 0.4603174603174605, 0.3809523809523814, 0.3015873015873018, 0.2222222222222223, 0.1428571428571432, 0.06349206349206415, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.03174603174603208, 0.08465608465608465, 0.1375661375661377, 0.1904761904761907, 0.2433862433862437, 0.2962962962962963, 0.3492063492063493, 0.4021164021164023, 0.4550264550264553, 0.5079365079365079, 0.5608465608465609, 0.6137566137566139, 0.666666666666667] greenmap = [0, 0.03968253968253968, 0.07936507936507936, 0.119047619047619, 0.1587301587301587, 0.1984126984126984, 0.2380952380952381, 0.2777777777777778, 0.3174603174603174, 0.3571428571428571, 0.3968253968253968, 0.4365079365079365, 0.4761904761904762, 0.5158730158730158, 0.5555555555555556, 0.5952380952380952, 0.6349206349206349, 0.6746031746031745, 0.7142857142857142, 0.753968253968254, 0.7936507936507936, 0.8333333333333333, 0.873015873015873, 0.9126984126984127, 0.9523809523809523, 0.992063492063492, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0.9841269841269842, 0.9047619047619047, 0.8253968253968256, 0.7460317460317465, 0.666666666666667, 0.587301587301587, 0.5079365079365079, 0.4285714285714288, 0.3492063492063493, 0.2698412698412698, 0.1904761904761907, 0.1111111111111116, 0.03174603174603208, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0] bluemap = [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0.01587301587301582, 0.09523809523809534, 0.1746031746031744, 0.2539682539682535, 0.333333333333333, 0.412698412698413, 0.4920634920634921, 0.5714285714285712, 0.6507936507936507, 0.7301587301587302, 0.8095238095238093, 0.8888888888888884, 0.9682539682539679, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1] elif name == 'winter': greenmap = [0, 1] bluemap = [1, 0.5] else: if name != 'jet': print 'Warning: colormap "%s" not supported. Using jet instead.' % name redmap = [0,0,0,0,0.5,1,1,1,0.5] greenmap = [0,0,0.5,1,1,1,0.5,0,0] bluemap = [0.5,1,1,1,0.5,0,0,0,0] return 255.0 * np.array(redmap), 255.0 * np.array(greenmap), 255.0 * np.array(bluemap)
0 0
- 深度学习框架 Digits 3.0 安装运行
- 深度学习框架 Digits 3.0 安装运行
- Digits用于深度学习
- 学习笔记︱Nvidia DIGITS网页版深度学习框架——深度学习版SPSS
- Nvidia DIGITS网页版深度学习框架——深度学习版SPSS
- 深度学习框架Keras安装
- 深度学习框架Keras安装
- 安装torch7 深度学习框架
- python安装深度学习框架
- 安装 Caffe , Berkely的深度学习框架
- 安装 Caffe , Berkely的深度学习框架
- 深度学习框架Keras的安装
- 深度学习框架mxnet安装与使用
- 深度学习框架Keras的安装
- Mac下安装caffe深度学习框架
- 深度学习框架Keras安装指南
- 深度学习框架之keras安装记录
- TX2-安装tensorflow深度学习框架
- 数据结构-线性表的数组实现-Java
- GNU C 9条扩展语法
- Ubuntu 15 安装搜狗输入法
- owap-zap之截断代理
- Codeforces 120I
- 深度学习框架 Digits 3.0 安装运行
- 第三天!神奇的物理引擎!
- easyui的easyui-textbox、easyui-combobox 类
- C#中wpf使用Devexpress画图:柱状图,饼状图,散点图,曲线图
- R call fortran
- java归并排序算法报错
- Linux 输入子系统编码:event-codes.txt
- Material Design学习之Toolbar
- Docker完全解析 三、Docker容器