dgbpy.dgbkeras
Module Contents
Classes
Send progress message to bokeh |
|
Functions
|
|
|
|
Use this function to set the global policy to mixed precision. |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Transfer learning utility function for fine-tuning a Keras model. |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Attributes
- dgbpy.dgbkeras.hasKeras()
- dgbpy.dgbkeras.withtensorboard
- dgbpy.dgbkeras.platform
- dgbpy.dgbkeras.cudacores = ['1', '2', '4', '8', '16', '32', '48', '64', '96', '128', '144', '192', '256', '288', '384',...
- dgbpy.dgbkeras.getMLPlatform()
- dgbpy.dgbkeras.getUIMLPlatform()
- dgbpy.dgbkeras.defbatchstr = defaultbatchsz
- dgbpy.dgbkeras.keras_infos
- dgbpy.dgbkeras.default_transforms = []
- dgbpy.dgbkeras.keras_dict
- dgbpy.dgbkeras.can_use_gpu()
- dgbpy.dgbkeras.get_cpu_preference()
- dgbpy.dgbkeras.get_keras_infos()
- dgbpy.dgbkeras.set_compute_device(prefercpu)
- dgbpy.dgbkeras.use_mixed_precision()
Use this function to set the global policy to mixed precision.
- dgbpy.dgbkeras.getParams(dodec=keras_dict[dgbkeys.decimkeystr], nbchunk=keras_dict['nbchunk'], epochs=keras_dict['epochs'], batch=keras_dict['batch'], patience=keras_dict['patience'], learnrate=keras_dict['learnrate'], epochdrop=keras_dict['epochdrop'], nntype=keras_dict['type'], prefercpu=keras_dict['prefercpu'], transform=keras_dict['transform'], validation_split=keras_dict['split'], nbfold=keras_dict['nbfold'], scale=keras_dict['scale'], withtensorboard=keras_dict['withtensorboard'], tofp16=keras_dict['tofp16'])
- dgbpy.dgbkeras.adaptive_schedule(initial_lrate=keras_dict['learnrate'], epochs_drop=keras_dict['epochdrop'])
- dgbpy.dgbkeras.get_data_format(model)
- dgbpy.dgbkeras.hasValidCubeletShape(cubeszs)
- dgbpy.dgbkeras.getCubeletShape(model)
- dgbpy.dgbkeras.get_model_shape(shape, nrattribs, attribfirst=True)
- dgbpy.dgbkeras.getModelDims(model_shape, data_format)
- dgbpy.dgbkeras.getModelsByType(learntype, classification, ndim)
- dgbpy.dgbkeras.getModelsByInfo(infos)
- dgbpy.dgbkeras.getDefaultModel(setup, type=keras_dict['type'], learnrate=keras_dict['learnrate'], data_format='channels_first')
- dgbpy.dgbkeras.hasFastprogress()
- class dgbpy.dgbkeras.ProgressBarCallback(config)
Bases:
keras.callbacks.Callback- on_train_begin(self, logs=None)
- on_epoch_begin(self, epoch, logs=None)
- on_epoch_end(self, epoch, logs=None)
- on_test_begin(self, logs=None)
- on_train_batch_end(self, batch, logs=None)
- on_test_batch_end(self, batch, logs=None)
- on_train_end(self, logs=None)
- class dgbpy.dgbkeras.ProgressNoBarCallback(config)
Bases:
keras.callbacks.Callback- on_train_begin(self, logs=None)
- on_epoch_begin(self, epoch, logs=None)
- on_epoch_end(self, epoch, logs=None)
- class dgbpy.dgbkeras.BokehProgressCallback(config)
Bases:
keras.callbacks.CallbackSend progress message to bokeh
- on_train_begin(self, logs=None)
- on_epoch_begin(self, epoch, logs=None)
- on_epoch_end(self, epoch, logs=None)
- on_train_batch_begin(self, batch, logs=None)
- on_test_batch_begin(self, batch, logs=None)
- on_train_batch_end(self, batch, logs=None)
- on_test_batch_end(self, batch, logs=None)
- on_train_begin_chunk(self)
- on_train_begin_fold(self)
- class dgbpy.dgbkeras.LogNrOfSamplesCallback(config)
Bases:
keras.callbacks.Callback- on_train_begin(self, logs=None)
- on_train_begin_fold(self)
- class dgbpy.dgbkeras.TransformCallback(config)
Bases:
keras.callbacks.Callback- on_epoch_begin(self, epoch, logs=None)
- dgbpy.dgbkeras.epoch0endCB(epoch, logs)
- dgbpy.dgbkeras.init_callbacks(monitor, params, logdir, silent, custom_config, cbfn=None)
- dgbpy.dgbkeras.train(model, training, params=keras_dict, trainfile=None, silent=False, cbfn=None, logdir=None, tempnm=None)
- dgbpy.dgbkeras.updateModelShape(infos, model, forinput)
- dgbpy.dgbkeras.save(model, outfnm)
- dgbpy.dgbkeras.load(modelfnm, fortrain, infos=None, pars=keras_dict)
- dgbpy.dgbkeras.transfer(model)
Transfer learning utility function for fine-tuning a Keras model.
This function takes a Keras model and prepares it for transfer learning by selectively setting layers to be trainable. The layers to be made trainable are determined as follows:
All layers before the first Conv1D, Conv2D, or Conv3D layer (or a Sequential containing such layers) are set to trainable.
All layers after the last Conv1D, Conv2D, Conv3D, or Dense layer (or a Sequential containing such layers) are set to trainable.
- All layers between the first and last Conv1D, Conv2D, Conv3D, or Dense layer (or a Sequential
containing such layers) are set to non-trainable.
- dgbpy.dgbkeras.apply(model, samples, isclassification, withpred, withprobs, withconfidence, doprobabilities, dictinpshape=None, scaler=None, batch_size=None)
- dgbpy.dgbkeras.adaptToModel(model, samples, dictinpshape=None, sample_data_format='channels_first')
- dgbpy.dgbkeras.adaptFromModel(model, samples, inp_shape, ret_data_format)
- dgbpy.dgbkeras.plot(model, outfnm, showshapes=True, withlaynames=False, vertical=True)
- dgbpy.dgbkeras.compute_capability_from_device_desc(device_desc)
- dgbpy.dgbkeras.getDevicesInfo(gpusonly=True)
- dgbpy.dgbkeras.is_gpu_ready()
- dgbpy.dgbkeras.is_mixed_precision_compatible(min_version=(7, 0))
- dgbpy.dgbkeras.need_channels_last()
- dgbpy.dgbkeras.get_validation_data(trainseq)