Open JohnAMacDonald opened 10 months ago
It does exactly this to me as well, GPU used only when extracting pitch with rmvpe_gpu, though I am on an Intel Arc card. Maybe there's something wrong with the initial AMD/Intel release?
And for my part, I don't even see the little window in teh webui with "Enter GPU Index(es),separated by (Example: 0-1-2 to select card 1, 2 and 3" and "GPU Information", it's just blank :
This is what my "Train" tab looks like directly after booting up with an AMD GPU on Windows 11 Pro. Posting in case it can be helpful on fixing the issue :)
Yeah you've got the same problem as me, look at this screenshot I got from a guy that uses an NVIDIA GPU:
The step 2b says that we must specify the GPU index, but currently we cannot
Same error on AMD RX590 GME No GPU information
Also NO GPU DETECTED
I'm experiencing the exact same problem. The UI element for selecting a GPU under step 2b is missing and when using the "one-click-training" button or the "train model" button the command line states no gpu detected. Glad I'm not the only one with this problem. Hope it gets fixed.
CPU: 5 5600 GPU: 6800xt OS: windows 11
7900 XTX on W10 also not detected. (Not 6700k intel gpu for the fact)
It's interesting though that for one step it did use GPU. It went to 80-90% use, I heard the coil whine. A lot of stuff has been done fast, and then boop: no gpu detected and the rest has fallen to cpu...
load model(s) from hubert_base.pt
load model(s) from hubert_base.pt
move model to privateuseone:0
all-feature-108
now-108,all-0,11_1.wav,(120, 768)
move model to privateuseone:0
move model to privateuseone:0
all-feature-109
all-feature-109
now-109,all-0,0_1.wav,(53, 768)
now-109,all-0,10_1.wav,(78, 768)
now-108,all-10,17_123.wav,(149, 768)
now-109,all-10,17_121.wav,(149, 768)
now-109,all-10,17_122.wav,(149, 768)
now-108,all-20,17_158.wav,(149, 768)
now-109,all-20,17_155.wav,(149, 768)
now-109,all-20,17_156.wav,(149, 768)
now-108,all-30,17_189.wav,(149, 768)
now-109,all-30,17_187.wav,(149, 768)
now-109,all-30,17_188.wav,(149, 768)
now-108,all-40,17_221.wav,(149, 768)
now-109,all-40,17_219.wav,(149, 768)
now-109,all-40,17_220.wav,(149, 768)
now-108,all-50,17_252.wav,(149, 768)
now-109,all-50,17_250.wav,(149, 768)
now-109,all-50,17_251.wav,(149, 768)
now-108,all-60,17_283.wav,(149, 768)
now-109,all-60,17_281.wav,(149, 768)
now-109,all-60,17_282.wav,(149, 768)
now-108,all-70,17_314.wav,(149, 768)
now-109,all-70,17_311.wav,(149, 768)
now-109,all-70,17_313.wav,(149, 768)
now-108,all-80,17_346.wav,(149, 768)
now-109,all-80,17_345.wav,(149, 768)
now-109,all-80,17_344.wav,(123, 768)
now-108,all-90,17_52.wav,(149, 768)
now-109,all-90,17_51.wav,(149, 768)
now-109,all-90,17_50.wav,(149, 768)
now-108,all-100,17_82.wav,(149, 768)
now-109,all-100,17_81.wav,(149, 768)
now-109,all-100,17_80.wav,(149, 768)
all-feature-done
all-feature-done
all-feature-done
NO GPU DETECTED: falling back to CPU - this may take a while
I have the same problem with RX 6800 XT and the most strange thing is that it utilizes the CPU only by 65-66%
AMD GPU trained is not supported now. We are trying to support it in the future versions.
Sad to hear, must have misinterpreted that AMD training was supported.
Do you guys have a rough ETA of when AMD training might be implemented? Super excited to start training models! :)
Thanks in advance!
@RVC-Boss what about Intel GPUs?
Hello!
I just downloaded RVC0813AMD_Intel.7z from HuggingFace, great work on the release by the way!
I have an AMD RX 6750 XT 12 GB GPU and when training everything works and it uses my AMD GPU as in Task Manager it shows my GPU usage going to ~80% when extracting pitch with rmvpe_gpu, but when I get to the button "Train Model" and hit it the CMD window say's "NO GPU DETECTED: falling back to CPU - this may take a while". For the GPU Index I put "0" without quotes, I've also put "0-1-2" without quotes, but it still say's the same error.
Any tips/help? Thanks!