Transfer masking in Keras
Project description
Keras Transfer Masking
Remove and restore masks for layers that do not support masking. Note that the result may be incorrect in most cases.
Install
pip install keras-trans-mask
Usage
Conv1D
does not support masking. By removing the mask you'll get a "nearly correct" output:
from tensorflow import keras
from keras_trans_mask import RemoveMask, RestoreMask
input_layer = keras.layers.Input(shape=(None,))
embed_layer = keras.layers.Embedding(
input_dim=10,
output_dim=15,
mask_zero=True,
)(input_layer)
removed_layer = RemoveMask()(embed_layer) # Remove mask from embeddings
conv_layer = keras.layers.Conv1D(
filters=32,
kernel_size=3,
padding='same',
)(removed_layer)
restored_layer = RestoreMask()([conv_layer, embed_layer]) # Restore mask from embeddings
lstm_layer = keras.layers.LSTM(units=5)(restored_layer)
dense_layer = keras.layers.Dense(units=2, activation='softmax')(lstm_layer)
model = keras.models.Model(inputs=input_layer, outputs=dense_layer)
model.compile(optimizer='adam', loss='sparse_categorical_crossentropy')
model.summary()
Project details
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
File details
Details for the file keras-trans-mask-0.6.0.tar.gz
.
File metadata
- Download URL: keras-trans-mask-0.6.0.tar.gz
- Upload date:
- Size: 2.9 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/1.13.0 pkginfo/1.5.0.1 requests/2.22.0 setuptools/41.0.1 requests-toolbelt/0.9.1 tqdm/4.32.2 CPython/3.7.4
File hashes
Algorithm | Hash digest | |
---|---|---|
SHA256 | 6184bbdb5dc50a9e1561bc331a7807de6404c44741f370451e3dfc3905127d4b |
|
MD5 | 94949e836d8103bf88459b22e57facfc |
|
BLAKE2b-256 | 7402a8d29d17cabdbdb19ae38e7b83c76d32490ac70ae6e1d12fa98c2f511a03 |