Fine tuning the model


I am in search of appropriate loss function and corresponding learning rate for my problem.
I am new to deep learning and pytorch, so just want to confirm if my approach is correct ?

Model is trained for each loss function in different attempts with different learning rates, such as 0.01,0.001,0.0001.
I observed that at first epoch only, there is a considerable difference in the metric at different learning rates and similar difference is more or less maintained upto 5 epochs.
May i know if observing just 5 epochs is sensible approach or should I wait for the model to converge to make decision regarding learning rate?