junn_predict.common package¶
Submodules¶
junn_predict.common.cli module¶
junn_predict.common.logging module¶
junn_predict.common.configure_tensorflow module¶
Helper functionality to configure TensorFlow.
- junn_predict.common.configure_tensorflow.configure_tensorflow(seed=None, windows_maximum_gpu_memory=0.75)¶
Configure TensorFlow.
It is important that this function is called BEFORE first TF import.
Reduces TF’s log-level (before) loading.
Sets all devices to dynamic memory allocation (growing instead of complete)
(On Windows) Sets overall maximum TensorFlow memory to windows_maximum_gpu_memory
Removes the tensorflow log adapter from the global logger
Sets Keras to use TensorFlow (and sets USE_TENSORFLOW_KERAS to 1, custom environment variable)
- Parameters
seed – Optional. If set, will be passed to set_seed()
windows_maximum_gpu_memory – Set the maximum GPU memory fraction to use (Windows only).
- Returns
None
- junn_predict.common.configure_tensorflow.get_gpu_memory_usages_megabytes()¶
Get the current GPU memory usages.
- Returns
List of memory usages.
- junn_predict.common.configure_tensorflow.set_seed(seed)¶
Set various RNG seeds, so their behavior becomes reproducible.
Seeds NumPy, Python random, and TensorFlow.
- Parameters
seed – Seed value to use.
- Returns
junn_predict.common.tensorflow_addons module¶
Helper functions to dynamically load TensorFlow Addons.
- junn_predict.common.tensorflow_addons.try_load_tensorflow_addons() → None¶
Load the TensorFlow Addons module, if present.
Will ignore any compatibility warnings, and loads all tfa activations into the normal namespace.