Perplexity not changes with tgt_len

[bajajahsaas]

Hi, I am using the eval script by changing the tgt_len parameter. It ideally changes the "number of tokens to predict" and even changes the way data is pre-processed. However, tweaking this parameter during evaluation does not perplexity even slightly. What could be the reason? And how can we test this model's performance for different context windows? Thanks.