You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
* Switching trainer to use layers/optimizers
* Updated documentation for optimizer argument and added support for instance of sub-class of optimizer if user wants to provide optimizer with custom parameters.
* Added support for passing learning_rate Tensor
* Make trainer work with layers/optimize_loss
* Move config_addong.py to estimators and rename ConfigAddon to RunConfig
* Added size=small to all tests
* Moved tf_master and tf_random_seed into RunConfig
* Removing need in Trainer class - replacing it with just a function and creating optimizer in the setup_training function.
* Moved saver params into RunConfig
* Moved creation of placeholders inside data feeder. Can be used for multi inputs/outputs.
* switch to use platform.default._gfile for file operations
* Adding a build rule for all skflow examples
* correcting build rule and import in neural translation example
* Replacing one_hot_matrix by tf.one_hot (speed + GPU)
* Fixing missing test and left over refactoring in montior.py
* Fixed iris custom config example. Also removed duplication of names in trainer.
* rename iris_config_addon to iris_run_config
* Fix keep_prob->dropout arg rename in dnn
* Removing .orig files
* Bump test_early_stopping test to medium
0 commit comments