Linux little endian and gpt

I’m using a graph with gpt in a Linux little-endian system to process S-1 IW burst data from SLC to terrain-corrected data in the covariance matrix format.

Anyone know why do I get my beam-dimap output data as big endian?

Thanks!