[Python] alternating epochs are resulting into all 0 results , loss and accuracy all zero...

while i am running my training loop i am getting a valid accuracy on one epoch but on every alternate epoch i am getting all 0 results my loss and accuracy all the results are 0

I was trying to train my model on a binary dataset, but alternating epochs are resulting into all zero results . i tried the using normal resnet it gives this kind of result. i had to balance my class weights ,using class weights method ,even when i am commenting it off the issue still persists. i have used data augmentation and i am not using early stopping.

# Fix incorrect steps_per_epoch calculation
steps_per_epoch = (train_generator.samples // train_generator.batch_size) + int(train_generator.samples % train_generator.batch_size > 0)
validation_steps = (val_generator.samples // val_generator.batch_size) + int(val_generator.samples % val_generator.batch_size > 0)

# Compile the model (Lower initial LR for stability)
final_model.compile(
optimizer=Adam(learning_rate=1e-4), # ✅ Reduced LR
loss='binary_crossentropy',
metrics=['accuracy', tf.keras.metrics.AUC(name='auc')]
)
# Train the model
history = final_model.fit(
train_generator,
steps_per_epoch=steps_per_epoch,
validation_data=val_generator,
validation_steps=validation_steps,
epochs=10,
#class_weight=class_weights_dict,
#callbacks=callbacks
)

# Evaluation
print("Validation Results:")
final_model.evaluate(val_generator)

# Get the final training accuracy from the history of fine-tuning
train_accuracy = history.history['accuracy'][-1]

# Evaluate on the test set
test_loss, test_accuracy, test_auc = final_model.evaluate(test_generator)

# Print results
print(f"Final Training Accuracy: {train_accuracy:.4f}")
print(f"Test Loss: {test_loss:.4f}")
print(f"Test Accuracy: {test_accuracy:.4f}")
print(f"Test AUC: {test_auc:.4f}")

Continue reading...