AIronTools
AIronTools (Beta) is a Python library that provides the user with deep learning tools built to work with tensorflow (or pytorch in the future) as a backend.
Key features:
- Model constructor that allows multiple models to be optimized in parallel across multiple GPUs.
- Block constructor to build customised blocks/models.
- Layer constructor to build customised layers.
- Preprocessing utils.
Installation
pip install airontools
Custom Keras subclass to build a variational autoencoder (VAE) with airontools
import tensorflow as tf
from tensorflow.keras.datasets import mnist
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input, Layer, Reshape
from tensorflow.keras.metrics import Mean
from tensorflow.keras.losses import binary_crossentropy
from tensorflow.keras.optimizers import Adam
os.environ['AIRONSUIT_BACKEND'] = 'tensorflow'
from aironsuit.suit import AIronSuit
from airontools.preprocessing import train_val_split
from airontools.model_constructors import layer_constructor
class VAE(Model):
def __init__(self, latent_dim, **kwargs):
super(VAE, self).__init__(**kwargs)
self.total_loss_tracker = Mean(name="total_loss")
self.reconstruction_loss_tracker = Mean(name="reconstruction_loss")
self.kl_loss_tracker = Mean(name="kl_loss")
# Encoder
encoder_inputs = Input(shape=(28, 28, 1))
encoder_conv = layer_constructor(
encoder_inputs, name='encoder_conv', filters=32, kernel_size=3, strides=2, advanced_reg=True)
z_mean = layer_constructor(encoder_conv, name='encoder_mean', units=latent_dim, advanced_reg=True)
z_log_var = layer_constructor(encoder_conv, name='encoder_log_var', units=latent_dim, advanced_reg=True)
z = Sampling()([z_mean, z_log_var])
self.encoder = Model(encoder_inputs, [z_mean, z_log_var, z], name="encoder")
# Decoder
latent_inputs = Input(shape=(latent_dim,))
decoder_outputs = layer_constructor(latent_inputs, name='encoder_dense', units=7 * 7 * 64, advanced_reg=True)
decoder_outputs = Reshape((7, 7, 64))(decoder_outputs)
for i, filters, activation in zip([1, 2], [64, 32], ['relu', 'relu']):
decoder_outputs = layer_constructor(
decoder_outputs,
name='decoder_conv', name_ext=str(i), activation=activation, filters=filters, kernel_size=3,
strides=2, padding='same', conv_transpose=True, advanced_reg=True)
decoder_outputs = layer_constructor(
decoder_outputs,
name='decoder_output', activation='sigmoid', filters=1, kernel_size=3, padding='same',
conv_transpose=True, advanced_reg=True)
self.decoder = Model(latent_inputs, decoder_outputs, name="decoder")
More examples
see usage examples in aironsuit/examples
1 maintainer
Release files for airontools 0.1.8
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| airontools-0.1.8-py3-none-any.whl | Python 3 | none | any | Details |
Release files / airontools-0.1.8-py3-none-any.whl
| Download URL | airontools-0.1.8-py3-none-any.whl |
|---|---|
| Size | 16.2 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
430809414867d0b0cfa1b89b7f3dec573d26121c3b5d628442ff9a290049e060
|
|
BLAKE2b-256 checksum How to use checksums |
1ca1ee4f0a89c7f7eb356f1da893ab5726ba05871394ac223304efb0f1926e5d
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/3.4.1 importlib_metadata/4.0.1 pkginfo/1.7.0 requests/2.26.0 requests-toolbelt/0.9.1 tqdm/4.62.3 CPython/3.8.5
|
