tlt.models.image_classification.tf_image_classification_model.TFImageClassificationModel

class tlt.models.image_classification.tf_image_classification_model.TFImageClassificationModel(model_name: str, model=None, optimizer=None, loss=None, **kwargs)

Class to represent a TF custom pretrained model for image classification

__init__(model_name: str, model=None, optimizer=None, loss=None, **kwargs)

Class constructor

Methods

__init__(model_name[, model, optimizer, loss])	Class constructor
benchmark(dataset[, saved_model_dir, ...])	Use Intel Neural Compressor to benchmark the model with the dataset argument.
cleanup_saved_objects_for_distributed()
evaluate(dataset[, use_test_set, callbacks, ...])	Evaluate the accuracy of the model on a dataset.
export(output_dir)	Exports a trained model as a saved_model.pb file.
export_for_distributed([export_dir, ...])	Exports the model, optimizer, loss, train data and validation data to the export_dir for distributed script to access.
load_from_directory(model_dir)	Loads a saved model from the specified directory
optimize_graph(output_dir[, overwrite_model])	Performs FP32 graph optimization using the Intel Neural Compressor on the model and writes the inference-optimized model to the output_dir.
predict(input_samples[, return_type, ...])	Perform feed-forward inference and predict the classes of the input_samples.
quantize(output_dir, dataset[, config, ...])	Performs post training quantization using the Intel Neural Compressor on the model using the dataset.
set_auto_mixed_precision(...)	Enable auto mixed precision for training.
train(dataset, output_dir[, epochs, ...])	Trains the model using the specified image classification dataset.

Attributes

do_fine_tuning	When True, the weights in all of the model's layers will be trainable.
dropout_layer_rate	The probability of any one node being dropped when a dropout layer is used
framework	Framework with which the model is compatible
image_size	The fixed image size that the pretrained model expects as input, in pixels with equal width and height
learning_rate	Learning rate for the model
model_name	Name of the model
num_classes	The number of output neurons in the model; equal to the number of classes in the dataset
preprocessor	Preprocessor for the model
use_case	Use case (or category) to which the model belongs