Is there a way I can load model checkpoint of Pytorch in Tensorflow? Or atleast can I extract the weights from my Pytorch checkpoint and save it in a .h5 file and use it in Keras, using model.load_weights?
You could export your model to onnx and load it into tensorflow.
Exporting the weights works in theory, too, but it is not uncommon that there are subtle differences in how the operators work (e.g. padding for convolutions, …).