Generates equal keras models with the desired data format
Project description
Keras data format converter
Generates equal keras models with the desired data format
Requirements
tensorflow >= 2.0
API
convert_channels_first_to_last(model: keras.Model, inputs_to_transpose: List[str] = None, verbose: bool = False) -> keras.Model
convert_channels_last_to_first(model: tf.keras.Model, inputs_to_transpose: List[str] = None, verbose: bool = False) \ -> tf.keras.Model
model
: Keras model to convert
inputs_to_transpose
: list of input names that need to be transposed due tothe data foramt changing
verbose
: detailed output
Getting started
from tensorflow import keras
from keras_data_format_converter import convert_channels_last_to_first
# Load Keras model
keras_model = keras.models.load_model("my_image_model")
# Call the converter (image_input is an input that needs to be transposed, can be different for your model)
converted_model = convert_channels_last_to_first(keras_model, ["image_input"])
Supported Layers with Special handling
- Normalization layers
- Permute
- Reshape
- Concatenate
- Dot
- MultiHeadAttention
- TFOpLambda (Inserted by the Functional API construction whenever users call a supported TF symbol on KerasTensors, see here at Tensorflow repo for more info)
Unsupported Layers due to lack of data_format property
- Cropping1D
- Upsampling1D
- Zeropadding1D
- All layers in tensorflow.keras.preprocessing
How to deploy
- Create a new release version on GitHub
- Update parameters in setup.py (usually
version
anddownload_url
) - Run
python setup.py sdist
in root directory - Run
pip install twine
- Run
twine upload dist/*
License
This software is covered by MIT License.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Close
Hashes for keras-data-format-converter-0.1.1.dev1.tar.gz
Algorithm | Hash digest | |
---|---|---|
SHA256 | 4dacd9da9ea38709c65a463d6e196c3a46395dc8a82d90b08a788412e5294a63 |
|
MD5 | b8764e5b5f99afa4a63cf71a4f03715c |
|
BLAKE2b-256 | 8dda03d78db13378ae6f96636a9d40c602cf2723a18a8db04ea0967095abc5e0 |
Close
Hashes for keras_data_format_converter-0.1.1.dev1-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 7de0551705ee5730b8c00cf164122f91e078b7d3631cad2710d7521b40c879fa |
|
MD5 | f6531a79974c78dd80ce76218cda45fd |
|
BLAKE2b-256 | 17131a5d1522e210337d9825bde13a0719e5a5e66beea784a2eb87e66e5131ca |