Closed Bean95zx closed 1 year ago
i found the solution. add the parameters --machine.num-gpus 2.
Here are the commands I trained on two GPUs.
ns-train in2n --data /sdisk/shome/speed/zx/nerdStudio/nerfstudio/data/nerfstudio/poster/ --load-dir /sdisk/shome/speed/zx/Instruct-Nerf2Nerf/instruct-nerf2nerf/outputs/poster/nerfacto/2023-06-02_103140/nerfstudio_models/ --pipeline.prompt "make it cartoon" --pipeline.guidance-scale 7.5 --pipeline.image-guidance-scale 1.5 --machine.num-gpus 2 --pipeline.ip2p-device cuda:1
i want to use two gpus to train. my command is :
ns-train in2n-tiny --data /sdisk/shome/speed/zx/nerdStudio/nerfstudio/data/nerfstudio/poster/ --load-dir /sdisk/shome/speed/zx/Instruct-Nerf2Nerf/instruct-nerf2nerf/outputs/poster/nerfacto/2023-06-02_103140/nerfstudio_models/ --pipeline.prompt "cartoon" --pipeline.guidance-scale 7.5 --pipeline.image-guidance-scale 1.5 --pipeline.ip2p-device cuda:0,1
but it show RuntimeError.
RuntimeError: Invalid device string: 'cuda:0,1'
i want to know how to set the --pipeline.ip2p-device