Releases: cabrust/chia
Releases · cabrust/chia
v2.5.0
Fixes:
EpochRunner: Fixed wrong number of validation examples and result_dict possibly not initialized
extrapolators: Fix extrapolators sometimes reducing precision
Changes:
test_extrapolators: Test actual output instead of just running experiments
Factory: Allow instantiation by class name only
Treat held-out test sets and validation sets separately
docs: Added JSONDataset documentation
test_epoch_runner_load_save: Test all combinations of held-out test and validation sets
keras_idk_hc: Adapt CHILLAX terminology
Fixed NLTK version because 3.6.6 renames omw corpus to omw-1.4
v2.4.2
Changes:
Updated documentation to include descriptions of architecture and configuration mechanism
Moved repository to cabrust/chia
Fixes:
#58 Fixed wrong version of keras installed by setup.py
#59 Fixed double instantation of KerasFeatureExtractor
v2.4.1
Changes:
Updated TensorFlow version to v2.6
Fixes:
ExperimentContainer: Fixed not calling dataset setup if config key is missing
v2.4.0
Enhancements:
Added methods from the CHILLAX project to unify code base.
Fixes:
Bumped TensorFlow and TensorFlow Addons versions to include mitigations against vulnerabilities.
v2.3.1
Fixes:
keras_labelsharing: Fixed hardcoded kappa value
v2.3.0
Changes:
experiment: Initialize dataset setup from configuration dict if possible
relation: Added reset method
v2.2.0
Changes:
Added CUB200-2011 dataset
Added Semantic Label Sharing (Fergus et al. 2010) implementation
v2.1.0
Changes:
Updated TensorFlow to 2.4
v2.0.3
Enhancements:
Better ioqueue for multi-threaded image input
You can’t perform that action at this time.