Bug Report: TensorFlow 2.x Compatibility Issues

[bysq-2006]

Environment Information

• Python: 3.8.20
• TensorFlow: 2.13.0
• Keras: 2.13.1
• NumPy: 1.24.3
• Librosa: 0.11.0
• Streamlit: 1.40.1
• OS: Windows

Issue 1: predict_classes() method removed in TensorFlow 2.x

File: model.py
Line: 160-162

Before:

predictions = self.model.predict_classes(ps)
class_id = predictions[0]
chord = str(CLASSES[class_id])

After:

predictions = self.model.predict(ps)
class_id = np.argmax(predictions, axis=1)[0]
chord = str(CLASSES[class_id])

Error Message:

AttributeError: 'Sequential' object has no attribute 'predict_classes'

Explanation: The predict_classes() method was deprecated in TensorFlow 2.6 and removed in later versions. The recommended approach is to use predict() combined with np.argmax() to get the predicted class indices.

Issue 2: Syntax error in string concatenation

File: app.py
Line: 29

Before:

format = ''.join[format, 'DB']

After:

format = ''.join([format, 'DB'])

Error Message:

TypeError: 'str' object is not subscriptable

Explanation: The join() method requires parentheses () for function call, not square brackets [] which are used for indexing.

Solution Summary

These changes ensure compatibility with modern TensorFlow 2.x versions while maintaining the same functionality. The modifications are minimal and focused on:

1. API Migration: Updated deprecated TensorFlow methods to current standards
2. Syntax Fix: Corrected Python syntax error in string operations

Both fixes are backward-compatible and follow current best practices for TensorFlow 2.x development.