# TFTF: TensorFlow TransFormer����
TensorFlow for everybody.
## Quick glance
```python
from tftf.layers import Layer, Dense, Activation
from tftf.models import Model
'''
Build model
'''
model = Model()
model.add(Dense(500, input_dim=784))
model.add(Activation('sigmoid'))
model.add(Dense(10))
model.add(Activation('softmax'))
model.compile()
model.describe()
'''
Train model
'''
model.fit(train_X, train_y)
'''
Test model
'''
print(model.accuracy(test_X, test_y))
```
See [examples](https://github.com/yusugomori/tftf/tree/master/examples) for other implementations.
## Installation
- **Install TFTF from PyPI (recommended):**
```sh
pip install tensorflow
pip install tftf
```
- **Alternatively: install TFTF from the GitHub source:**
First, clone TFTF using `git`:
```sh
git clone https://github.com/yusugomori/tftf.git
```
Then, `cd` to the TFTF folder and run the install command:
```sh
cd tftf
sudo python setup.py install
```
## Importable Layers, APIs
You can import low-level tftf APIs to your own TensorFlow implementations.
```python
from tftf.layers import Dense, Activation, NALU
from tftf import initializers as ini
from tftf import activations as act
from tftf import losses as loss
from tftf import optimizers as opt
from tftf.metrics import accuracy, f1
x = tf.placeholder(tf.float32, shape=[None, 784])
t = tf.placeholder(tf.float32, shape=[None, 10])
# import APIs
W = ini.glorot_normal([784, 200]) # or just write tf.Variable(...)
b = ini.zeros([200])
h = act.tanh(tf.matmul(x, W) + b) # or just write tf.nn.tanh(...)
# import Layers
h = Dense(200)(h)
h = Activation('tanh')(h)
h = NALU(200)(h)
W = ini.glorot_normal([200, 10])
b = ini.zeros([10])
y = act.softmax(tf.matmul(h, W) + b)
cost = loss.categorical_crossentropy(y, t)
train_step = opt.sgd(0.01).minimize(cost)
# Train
# ...
preds = y.eval(session=sess, feed_dict={x: test_X})
acc = accuracy(preds, test_y)
f = f1(preds, test_y)
print('accuracy: {:.3}'.format(acc))
print('f1: {:.3}'.format(f))
```
PyPI 官网下载 | tftf-0.0.20.tar.gz
版权申诉
91 浏览量
2022-02-02
06:49:19
上传
评论
收藏 13KB GZ 举报
tftf-0.0.20.tar.gz （78个子文件） 
tftf-0.0.20 
PKG-INFO 1KB tftf.egg-info PKG-INFO 1KB
requires.txt 35B SOURCES.txt 2KB top_level.txt 5B dependency_links.txt 1B tftf models metrics recall.py 348B f1.py 324B __init__.py 110B precision.py 366B accuracy.py 313B Model.py 9KB callbacks EarlyStopping.py 565B __init__.py 41B optimizers adadelta.py 105B adam.py 122B __init__.py 164B sgd.py 93B rmsprop.py 90B adagrad.py 89B momentum.py 148B losses categorical_crossentropy.py 231B __init__.py 167B mean_squared_error.py 116B binary_crossentropy.py 358B __init__.py 25B metrics __init__.py 39B activations __init__.py 43B optimizers __init__.py 42B losses __init__.py 38B initializers __init__.py 44B __init__.py 92B layers RNN.py 3KB Flatten.py 430B GlobalAveragePooling2D.py 413B activations leaky_relu.py 100B sigmoid.py 70B linear.py 28B softmax.py 70B elu.py 62B relu.py 64B __init__.py 301B tanh.py 64B hard_sigmoid.py 106B prelu.py 238B swish.py 95B selu.py 64B NALU.py 1KB Layer.py 3KB initializers glorot_normal.py 448B zeros.py 170B orthogonal.py 350B glorot_uniform.py 614B __init__.py 194B normal.py 270B ones.py 168B Dense.py 948B BatchNormalization.py 1KB regularizers l1_l2.py 329B l2.py 243B __init__.py 63B l1.py 251B NAC.py 1KB LSTM.py 5KB __init__.py 488B Activation.py 571B Conv2D.py 3KB Dropout.py 494B MaxPooling2D.py 2KB modules ResNet.py 3KB Module.py 31B __init__.py 54B datasets mnist.py 965B Dataset.py 395B __init__.py 30B setup.cfg 38B setup.py 1KB README.md 2KB资源评论
