Any chance to make the model work faster

[UtkuBulkan]

Resemble-enhance is pretty good in terms of enhancing the audio quality.

Is there any chance to make it faster ?

Changing argumemnts nfe or euler approach makes small changes.

For a 22 seconds audio, it takes generally about 14 seconds.

Any suggestions are welcome.

Note : Here are some of the approaches we have employed, but so far no luck in terms of making things faster

resemble-enhance --device cuda --solver euler --nfe 32 --parallel_mode {enhance_in_temp_directory} {enhance_out_temp_directory}"

resemble-enhance --device cuda --nfe 16 --parallel_mode {enhance_in_temp_directory} {enhance_out_temp_directory}"

resemble-enhance --device cuda --nfe 2 --parallel_mode {enhance_in_temp_directory} {enhance_out_temp_directory}"