I built it with AMD6700XT video card
FranklinBBai opened this issue · 2 comments
python networks/merge_lora.py --sd_model F:\Programming language\Stable-Diffusion-ONNX-FP16\chilloutmix_Ni.safetensors --save_to F:\Programming language\Stable-Diffusion-ONNX-FP16\sd-model\chilloutmix-lora-mix VirtualGirl-Rin.safetensors --models F:\Programming language\Stable-Diffusion-ONNX-FP16\Lora\VirtualGirl-Rin.safetensors\ --ratios 0.5
2023-04-03 23:26:07.346759: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'cudart64_110.dll'; dlerror: cudart64_110.dll not found
2023-04-03 23:26:07.346864: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
usage: merge_lora.py [-h] [--v2] [--save_precision {None,float,fp16,bf16}] [--precision {float,fp16,bf16}]
[--sd_model SD_MODEL] [--save_to SAVE_TO] [--models [MODELS ...]] [--ratios [RATIOS ...]]
merge_lora.py: error: unrecognized arguments: language\Stable-Diffusion-ONNX-FP16\chilloutmix_Ni.safetensors language\Stable-Diffusion-ONNX-FP16\sd-model\chilloutmix-lora-mix VirtualGirl-Rin.safetensors
The tool you are using is trying to use CUDA which only works on NVidia cards. You'll have to tell it to use CPU. I might include LORA merges as part of model conversion to simplify things a bit. Some other options exists, I'd have to check what works best.