Error when running python3 generate_voice_pack.py

[killmanda2]

Hello.
Thanks for your hard work.
Please help me.
When I run python3 generate_voice_pack.p I get two errors:

class "__nv_bfloat162" has already been defined (previous definition at line 4787 of /usr/local/cuda-12.6/include/cuda_bf16.h).

All installations were carried out according to the instructions. Installation on Windows 11.

[cktlco]

Hi, thanks for reaching out, we'll get it working for you. Those messages look like the normal and very verbose warnings/errors from the Deepspeed library.

The process continues after that message, right? I expect messages like that (I can't really suppress them since it's in third-party code), but it should continue past that section after a few minutes of initializing.

Does the entire process stop when you see that error? You can also try --disable_deepspeed to perhaps bypass that section entirely.

[cktlco]

Hi, I tried to reproduce, but on a machine with an Nvidia RTX 3090 I did not see the same errors you show above. If you are trying to run with an Nvidia GPU, maybe its a model not supported by deepspeed, in which case maybe the process will just work for you using --disable_deepspeed as part of python3 generate_voice_pack.py ....

Additionally, everything will still work using by using --cpu_only mode, just running more slowly than if you are using a modern nvidia GPU. Maybe try that as a fallback.

[cktlco]

Hi, were you able to get any further with these tips?