Understood.
I have tried everything, but is there any way to train using GPU with intel iris xe graphics? I can only train on cpu I can use GPU using directml for voice conversion but not for training.
@Disty
Understood.
I have tried everything, but is there any way to train using GPU with intel iris xe graphics? I can only train on cpu I can use GPU using directml for voice conversion but not for training.
@Disty
If you do it this way, the GPU itself seems to recognize it, but no actual conversion or training can be done!
INFO | configs. Config | Found GPU Intel(R) Graphics [0x9a49].
2023-10-02 23:05:48 | INFO | configs. Config | Use xpu: 0 instead
2023-10-02 23:05:48 | INFO | __main__ | Use Language: En_US
Describe the bug When building for IGPU, On xpu-master (320f4da), the script errors out late in the build. I started from a fresh checkout and am in a conda environment with python 3.10. My command...
@Disty I tried this (Installation Guide (Linux/WSL2) — intel_extension_for_pytorch 2.0.110+xpu documentation), but I was able to build up to step 6, but an error occurred if the necessary requirements were not installed. I tried to install what I needed, but I couldn't do anything because I had no idea what to install or what command to run using only the links in steps 1 and 3. Help me. I've been experimenting with this for about 2 weeks and haven't gotten any results.
@Disty I made a mistake in the link. I tried various things according to this procedure, but I couldn't figure out what to do with only the links in steps 1 and 3.
@Disty
Yup, thanks, adding to render was it. I had added myself to video instead for some reason.
Closing this as the problems were just a result of misconfiguration, but here's a summary for future readers of what I did.
Create a fresh python environment and activate it.
Optionally source /opt/intel/oneapi/setvars.sh and run the collect_env.py script check installation
Clone intel-extension-for-pytorch, checkout xpu-master, and build with bash scripts/compile_bundle.sh /opt/intel/oneapi/compiler/2023.2.0/ /opt/intel/oneapi/mkl/2023.2.0/ Gen12LP, replacing Gen12LP for the desired IGPU configuration
And then remember to source /oipt/intel/oneapi/setvars.sh before using torch.
@Disty I think I'll give up on training on GPU.
By the way, I downloaded the version introduced on this site, and although I was able to start the webui, I was unable to convert the audio. Is it okay to just download this?