poc improvements

[fabriziosalmi]

import argparse
import logging
import tensorflow as tf
from tensorflow.keras import datasets, layers, models
import matplotlib.pyplot as plt
import datetime

# Initialize Argument Parser
parser = argparse.ArgumentParser(description='TensorFlow PoC Script for MNIST Dataset')
parser.add_argument('--epochs', type=int, default=10, help='Number of epochs for training the model')
parser.add_argument('--batch-size', type=int, default=32, help='Batch size for training')
parser.add_argument('--save-model', action='store_true', help='Flag to save the trained model')
parser.add_argument('--learning-rate', type=float, default=0.001, help='Learning rate for the optimizer')
parser.add_argument('--optimizer', type=str, default='adam', help='Type of optimizer to use')
parser.add_argument('--validation-split', type=float, default=0.2, help='Fraction of the training data to be used as validation data')
parser.add_argument('--log-level', type=str, default='INFO', help='Logging level')

args = parser.parse_args()

# Configure Logging
log_file = f"mnist_poc_{datetime.datetime.now().strftime('%Y%m%d_%H%M%S')}.log"
logging.basicConfig(filename=log_file, level=args.log_level, format='%(asctime)s - %(levelname)s - %(message)s')

def load_and_preprocess_data():
    # Implementation remains the same
    pass  # Replace with actual implementation

def create_model():
    # Implementation remains the same
    pass  # Replace with actual implementation

def compile_and_train_model(model, train_images, train_labels):
    # Use args.learning_rate, args.optimizer, and args.validation_split
    optimizer = args.optimizer
    if optimizer.lower() == 'adam':
        optimizer = tf.keras.optimizers.Adam(learning_rate=args.learning_rate)
    elif optimizer.lower() == 'sgd':
        optimizer = tf.keras.optimizers.SGD(learning_rate=args.learning_rate)
    # Other optimizers can be added similarly

    model.compile(optimizer=optimizer,
                  loss='sparse_categorical_crossentropy',
                  metrics=['accuracy'])

    history = model.fit(train_images, train_labels, epochs=args.epochs, 
                        validation_split=args.validation_split, 
                        batch_size=args.batch_size)

    return history

def main():
    # Implementation remains the same
    pass  # Replace with actual implementation

if __name__ == "__main__":
    main()

[fabriziosalmi]

Convert keras h5 model to Tensorflow Lite:

• Documentation

  
  import tensorflow as tf

Create a model using high-level tf.keras.* APIs

model = tf.keras.models.Sequential([ tf.keras.layers.Dense(units=1, input_shape=[1]), tf.keras.layers.Dense(units=16, activation='relu'), tf.keras.layers.Dense(units=1) ]) model.compile(optimizer='sgd', loss='mean_squared_error') # compile the model model.fit(x=[-1, 0, 1], y=[-3, -1, 1], epochs=5) # train the model

(to generate a SavedModel) tf.saved_model.save(model, "saved_model_keras_dir")

Convert the model.

converter = tf.lite.TFLiteConverter.from_keras_model(model) tflite_model = converter.convert()

Save the model.

with open('model.tflite', 'wb') as f: f.write(tflite_model)