I born and live in Bangladesh.
كما أنني مسلم وأؤمن بالله.

I mostly follow tensorflow and keras as these are my primary ML library. Some of our answers that you may find useful are listed below.

• KERAS output
• Design an optimal CNN?
• TF.Keras model.predict is slower than straight Numpy?
• Multi-input Multi-output Model with Keras Functional API
• TensorFlow 2 - Quantization Aware Training (QAT) with tf.GradientTape
• Tensorflow 2 - tf.slice and its NumPy slice syntax incompatible behavior

and sometimes pytorch.

• Tape-based AUTOGRAD in Pytorch and GradientTape in TensorFlow 2?
• Label Smoothing in PyTorch