When I set per_process_gpu_memory from 0.5 to 1.0, there is not enough memory and it will crashed.
1) So, any ideas or suggestions to make it work?
2) Does convert tensorflow code to tensorRT will improve the performance (not for training, only for prediction)?