Code PlaidML-支持AMD显卡的深度学习框架-PlaidML安装教程

前言

自从Ryzen发布以后，AMD的CPU部门慢慢赶上了Intel。但A家的显卡这几年却还是慢悠悠的发展着，老黄在游戏卡领域叱咤风云，还借着深度学习的热潮，又在云计算这种专业领域也分到了不少蛋糕。诸如TensorFlow、Pytorch等深度学习框架均只支持N卡，作为AMD精神股东的我一直对此表示羡慕。今天终于可以不用羡慕了，因为我知道了一款支持AMD显卡的深度学习框架—>PlaidML!
地址为(https://github.com/plaidml/plaidml)

更让人激动的是，这款框架可以作为Keras的后端来使用，也就是说我之前略知一二的Keras知识可以直接用了。

安装

[Windows]
直接上pip即可

1
pip install plaidml-keras

安装好以后，还需要使用plaidml-setup命令进行后续安装工作。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
C:\Users\xxx>plaidml-setup

PlaidML Setup (0.6.3)

Thanks for using PlaidML!

Some Notes:
  * Bugs and other issues: https://github.com/plaidml/plaidml
  * Questions: https://stackoverflow.com/questions/tagged/plaidml
  * Say hello: https://groups.google.com/forum/#!forum/plaidml-dev
  * PlaidML is licensed under the Apache License 2.0


Default Config Devices:
   opencl_amd_ellesmere.0 : Advanced Micro Devices, Inc. Ellesmere (OpenCL)

Experimental Config Devices:
   opencl_amd_ellesmere.0 : Advanced Micro Devices, Inc. Ellesmere (OpenCL)
   llvm_cpu.0 : CPU (LLVM)

Using experimental devices can cause poor performance, crashes, and other nastiness.

Enable experimental device support? (y,n)[n]:y

Multiple devices detected (You can override by setting PLAIDML_DEVICE_IDS).
Please choose a default device:

   1 : opencl_amd_ellesmere.0
   2 : llvm_cpu.0

Default device? (1,2)[1]:1

Selected device:
    opencl_amd_ellesmere.0

Almost done. Multiplying some matrices...
Tile code:
  function (B[X,Z], C[Z,Y]) -> (A) { A[x,y : X,Y] = +(B[x,z] * C[z,y]); }
Whew. That worked.

Save settings to C:\Users\xxx\.plaidml? (y,n)[y]:y
Success!

在执行过程中，会让用户选择默认硬件，选择gpu即可。

使用

手动指定keras的后端为plaidml

在import keras前，设置环境变量KERAS_BACKEND为plaidml.keras.backend

1
2
os.environ["KERAS_BACKEND"] = "plaidml.keras.backend"
import keras

也可以通过修改~/.keras文件夹内keras.json进行一劳永逸式修改

1
2
3
4
5
6
{
    "floatx": "float32",
    "epsilon": 1e-07,
    "backend": "plaidml.keras.backend",
    "image_data_format": "channels_last"
}

测试

那就让被誉为深度学习届的Hello World项目—MNIST手写数字集来做测试吧，哈哈。

测试代码

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
import keras
from keras.datasets import mnist
from keras.models import Sequential
from keras.layers import Dense
from keras.optimizers import SGD

# Load datasets
(x_train, y_train), (x_test, y_test) = mnist.load_data()

x_train = x_train.reshape(60000, 784)
x_test = x_test.reshape(10000, 784)
# Normalization
x_train = x_train/255
x_test = x_test/255

# one hot code
y_train = keras.utils.to_categorical(y_train, 10)
y_test = keras.utils.to_categorical(y_test, 10)

model = Sequential()

model.add(Dense(256, activation='relu',input_shape=(784,)))
model.add(Dense(256, activation='relu'))
model.add(Dense(10, activation='softmax'))
model.summary()
model.compile(optimizer=SGD(),loss="categorical_crossentropy",metrics=['accuracy'])
model.fit(x_train,y_train,batch_size=64,epochs=5,validation_data=(x_test,y_test))

运行结果

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
G:\>python mnist.py
Using plaidml.keras.backend backend.
INFO:plaidml:Opening device "opencl_amd_ellesmere.0"
_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
dense_1 (Dense)              (None, 256)               200960
_________________________________________________________________
dense_2 (Dense)              (None, 256)               65792
_________________________________________________________________
dense_3 (Dense)              (None, 10)                2570
=================================================================
Total params: 269,322
Trainable params: 269,322
Non-trainable params: 0
_________________________________________________________________
Train on 60000 samples, validate on 10000 samples
Epoch 1/5
60000/60000 [==============================] - 20s 341us/step - loss: 0.8436 - acc: 0.7939 - val_loss: 0.3978 - val_acc: 0.8943
Epoch 2/5
60000/60000 [==============================] - 13s 213us/step - loss: 0.3603 - acc: 0.8990 - val_loss: 0.3082 - val_acc: 0.9136
Epoch 3/5
60000/60000 [==============================] - 11s 192us/step - loss: 0.3015 - acc: 0.9138 - val_loss: 0.2727 - val_acc: 0.9235
Epoch 4/5
60000/60000 [==============================] - 10s 166us/step - loss: 0.2677 - acc: 0.9232 - val_loss: 0.2466 - val_acc: 0.9312
Epoch 5/5
60000/60000 [==============================] - 10s 165us/step - loss: 0.2431 - acc: 0.9307 - val_loss: 0.2264 - val_acc: 0.9359

附:

本来我是打算在MacOS上安装的，但是遇到了找不到experimental.json的问题。但是等我解决了找不到experimental.json的问题后，又开始报找不到libplaidml.dylib错误，在网上几乎找不到什么资料。唯一找到的资料居然是重装MacOS，这个代价对我来说太大了，于是就转到了Windows下安装，然后一帆风顺。
参考资料

阅读量: