How to include the sudden peaks/bursts in LSTM based time-series model's training

I am using LSTM for time-series prediction whereby I am taking past 50 values as my input. Now, the thing is that it is predicting just OKish, and not doing the exact prediction, especially for the peaks.

Any help about how can I train my model to tackle this problem and take the peaks into account so that I can predict more accurately (if not EXACTLY).

THe model summary and the results are as below: