RuntimeError: `loss` passed to Optimizer.compute_gradients should be a function when eager execution is enabled. Working with Tensorflow 1.15

Question

I am trying to train a model using tensorflow 1.15 with eager execution enabled. For train loss I am using

train loss = mse_loss*args.lmbda + bits_per_pixel_loss

I've defined the optimizer as below

main_optimiser = tf.train.AdamOptimiser(learning_rate=1e-3)

Upto here the code is working fine The error is coming in the minimizer part of the optimisation

main_step = main_optimiser.minimise(train_loss, global_step=step)

I'm getting the following error:

RuntimeError: `loss` passed to Optimizer.compute_gradients should be a function when eager execution is enabled.

So when I calculate train_loss inside a function say, loss_computer, built inside the train function and then using

main_step =main_optimiser.minimise(loss_computer, global_step=step)

I am getting the following error:

AttributeError: 'NoneType' object has no attribute 'dtype'.

How can I train the model with eager execution enabled? Please let me know if I need to clarify anything. Any help is greatly appreciated, thank you!