Closed abasle closed 2 months ago
Hey Arnaud,
The pytest suite for the package is meant to be run on a CPU (and we just added some text to the docs to clarify this!), and you may indeed get failing tests on a GPU for test_reconstruct
which is testing here if reconstruction outputs match exactly. You can try again on a CPU to double-check — the CUDA warning is just nagging you that you could be using v12 on your cluster but it shouldn't affect DRGN-AI otherwise!
-Mike
Hi Michal,
Thanks! We will tyr DrgnAI a some challenging data with a lot of heterogeneity.
Cheers, Arnaud
Hello,
Our GPU cluster and installed software uses cuda 11.8.
I'm getting from drgnai pytest: [...] FAILED testing/test_reconstruct.py::test_small_dataset_and_load[toy] - assert 15.355 == 14.415 FAILED testing/test_reconstruct.py::TestSetupIntegration::test_configuration_keys[hand_no-ind] - AssertionError: /conda/envs/drgnai-env/lib/python3.9/site-packages/torch/cuda/init.py:128: UserWarning: CUDA initialization: The NVIDIA driver on your system is too old (found version 11080). Please update your GPU driver by downloading and installing a new version from the URL: h... [...]
Which version of cuda is required for DrgnAI? Cheers, Arnaud