How long does it take to finish the whole process?

[student-petercai]

I tried to run the example dataset with the command: python run_particlesfm.py --image_dir ./example/snowboard/images --output_dir ./outputs/snowboard/". However, it's already over 30 mins and it's not finished yet. It keeps doing global bundle adjustment. I'm running on a server with 4x 1080 GPUs. Is it normal or there might be something wrong with my setting? Thank you.

[Lifedecoder]

Hi! I met the same problem, I work on one single 3090ti, the mem is increasing very slowly, and after a very long time it haven't finihsed yet, I wonder how long it will take, and is there any possibility to handle multiple inferences.

[fangli333]

Can i know how long will it take?

[michaelz9436]

by the way does the required pytorch and cuda version support 3090? When I work on it, the following error occured: **_NVIDIA GeForce RTX 3090 with CUDA capability sm_86 is not compatible with the current PyTorch installation. The current PyTorch install supports CUDA capabilities sm_37 sm_50 sm_60 sm_61 sm_70 sm_75 compute_37. If you want to use the NVIDIA GeForce RTX 3090 GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/

warnings.warn(incompatible_device_warn.format(device_name, capability, " ".join(arch_list), device_name))_** and then the GPU device is not usable

[guohengkai]

@michaelz9436 You can try PyTorch >= 2.0.0 to see if it works.

[jqliu09]

Same, running on 4070 GPU, and it takes ~ 20 mins to finish the whole optimization procedures for 60+ images, and the train example is even slower.