Selecting GPU1 alone resulted in server crash！！！

[endluo]

I used GPU0 to load both fasterwhisper and gendermodel ("device_index=[0]", "cuda: 0") simultaneously, and the server worked properly. However, when I used GPU1 to load both fasterwhisper and gendermodel ("device_index=[1]", "cuda: 1") simultaneously, the server crashed and restarted

my fast_whisper Ver is 1.1.1

I have ruled out the problem with GPU1 itself, and even when using pytorch official code to run it all, there were no errors

[endluo]

two GPUs are of the same ，is 4090. CUDA is 12.3

[endluo]

i find the reason，the reason is turbo model，when i choose other model，GPU1 can load whisper model and pytorch model