無意中發現 MicroSoft 也加入了 openSource 的戰局,提出以下的 ToolKit 想與 Theano 及 TensorFlow 抗衡。試玩了一下,好像還不錯,至少幫我成功用上了 GPU。
而且還順便玩到了 CIFAR-10 影像辨認資料庫,以及 sequence-to-sequence translation。
這2項正是一直想拿來當作 進階版的 PR 教程,放在這裡也順便當作本學期 PR 課程的 happy ending。
The Microsoft Cognitive Toolkit (CNTK)
Microsoft's AI can now understand speech better than humans
https://github.com/Microsoft/CNTK
安裝檔在此: CNTK version 2.0 Beta 6 (Windows+Linux)
直接來到 http://localhost:8888/tree/Tutorials
【溫故】:
CNTK 103 Part A: MNIST Data Loader
CNTK 103: Part B - Feed Forward Network with MNIST
【知新】:
CNTK 201A Part A: CIFAR-10 Data Loader
CNTK 201B: Hands On Labs Image Recognition
https://www.cs.toronto.edu/~kriz/cifar.html
The CIFAR-10 dataset
The CIFAR-10 dataset consistsof 60000 32x32 colour images
in 10 classes, with 6000 images per class.
There are 50000 training images and 10000 test images.
【趕上】:
CNTK 204: Sequence to Sequence Networks with Text Data
The applications of sequence-to-sequence networks are nearly limitless.
It is a natural fit for
machine translation (e.g. English input sequences, French output sequences);
automatic text summarization (e.g. full document input sequence, summary output sequence);
word to pronunciation models (e.g. character [grapheme] input sequence, pronunciation [phoneme] output sequence);
and even parse tree generation (e.g. regular text input, flat parse tree output).
【練習】
https://www.dropbox.com/sh/jtdkwldyergd6au/AACW5jYsBP88fSNZjcfyNIj4a?dl=0
CNTK_103A_MNIST_DataLoader.ipynb
CNTK_103B_MNIST_FeedForwardNetwork.ipynb
CNTK_201A_CIFAR-10_DataLoader.ipynb
CNTK_201B_CIFAR-10_ImageHandsOn.ipynb
CNTK_204_Sequence_To_Sequence.ipynb
【增廣見聞】
收集一些有名的影像辨認資料庫,到目前(2016年底),最好的辨識錯誤率。
http://rodrigob.github.io/are_we_there_yet/build/#datasets
#-------------------------------------------------------------
然後,就在睡前,發現: Google 的 TensorFlow,居然也支援 Windows 了,
在 Python 3.5 之下,把它安裝起來,把 Keras 的 Backend 改回 TensorFlow,也跑了 CIFAR-10 資料庫。
看到這些科技大頭彼此【搶地盤】,頗為有趣!
TensorFlow supports only 64-bit Python 3.5 on Windows.
https://www.tensorflow.org/get_started/os_setup
【安裝 tensorFlow, Gpu 版本】
1. 官網: https://www.tensorflow.org/get_started/os_setup#anaconda-installation
To install the GPU version of TensorFlow, enter the following command at a command prompt:
【練習】
https://www.dropbox.com/sh/jtdkwldyergd6au/AACW5jYsBP88fSNZjcfyNIj4a?dl=0
CNTK_103A_MNIST_DataLoader.ipynb
CNTK_103B_MNIST_FeedForwardNetwork.ipynb
CNTK_201A_CIFAR-10_DataLoader.ipynb
CNTK_201B_CIFAR-10_ImageHandsOn.ipynb
CNTK_204_Sequence_To_Sequence.ipynb
【增廣見聞】
收集一些有名的影像辨認資料庫,到目前(2016年底),最好的辨識錯誤率。
http://rodrigob.github.io/are_we_there_yet/build/#datasets
Datasets who is the best at X ?
#-------------------------------------------------------------
然後,就在睡前,發現: Google 的 TensorFlow,居然也支援 Windows 了,
在 Python 3.5 之下,把它安裝起來,把 Keras 的 Backend 改回 TensorFlow,也跑了 CIFAR-10 資料庫。
看到這些科技大頭彼此【搶地盤】,頗為有趣!
TensorFlow supports only 64-bit Python 3.5 on Windows.
https://www.tensorflow.org/get_started/os_setup
【安裝 tensorFlow, Gpu 版本】
1. 官網: https://www.tensorflow.org/get_started/os_setup#anaconda-installation
To install the GPU version of TensorFlow, enter the following command at a command prompt:
C:\> pip install --upgrade https://storage.googleapis.com/tensorflow/windows/gpu/tensorflow_gpu-0.12.0-cp35-cp35m-win_amd64.whl
可能會出現 bugs
此處有解決: http://peacesky.cn/post/%E8%A7%A3%E5%86%B3%E5%AE%89%E8%A3%85Tensorflow%E6%97%B6%E7%9A%84setup-tool%E9%94%99%E8%AF%AF
重新 執行 官網的建議
pip install --ignore-installed setuptools
C:\> pip install --upgrade https:....
成功!!
沒有留言:
張貼留言