![]() If you work as a Data Scientist for one of those, you may have faced the need of setting up an Ubuntu server with the CUDA/cuDNN and TensorFlow stack for serving training/prediction jobs and pipelines.Īlso, one of the biggest burdens when dealing with AI is running out of memory while training heavy models (e.g. Usually, companies use cloud-based servers for Machine Learning applications such as Microsoft Azure and Google Cloud Platform, mostly because of their readily available and easy to use services, which are built for a wide range of problems and offer low-code solutions for daily ML operations.Īlthough the majority of companies can deliver value quickly and inexpensively through these tools, there are businesses that rely mainly on their ML capacities as their core product or strategy and most frequently invest lots of money and time on their AI infrastructure and code. Sorry if I’m unable to help much with this.How to setup and optimize CUDA and TensorFlow on Ubuntu 20.04 - 2022 Introduction It also might be easier to test Blender with CUDA rendering if you started with command line rendering rather than GUI render. Such as setup a desktop environment within your Linuxulator? This could be related to the GPU driver issue discussed before, or maybe you need to do a bit of extra setup to get GUI applications working in Linuxulator. You also have errors related to “opening a display” (opening the Blender GUI). Maybe the Nvidia GPU drivers weren’t installed properly? Or you need to disable nouveau manually? Or this is just some issue with Linuxulator? Typically when installing the Nvidia proprietary driver, the loading of the nouveau drivers gets disabled. For example on some Linux distributions I had to install packages like libcuda1 and libnvoptix1 to use CUDA and OptiX on Linux.Īs you pointed out, the error libGL error: failed to load driver: nouveau suggests Blender is trying to load the nouveau driver. This may not be a Blender issue, but a Linuxulator issue.ĭepending on how you installed your GPU drivers on “Linux”, you might not have all the packages required to run CUDA applications. Maybe try verifying that applications that use the GPU work, then CUDA applications, then look into getting Cycles rendering working with CUDA. It’s not guaranteed to work with all applications, and CUDA is likely to be one of the application types it will have issues with. Linuxulator appears to be some kind of compatibility layer. I can’t give you full help with this, but I will share some information based on what I can gather from this post and resources online. Since the nouveau driver does not support CUDA,maybe it should be “unlinked” from Blender and Blender should be “linked” to the nvidia driver,in some way. But it should look like the nVidia driver. It seems to me that Blender looks for the nouveau driver and it can’t find it. The error “CUDA cuInit: Unknown error” disappears,but the other errors are still there. I see this additional error : CUDA cuInit: Unknown errorīut if I remove the lib directory : rm -r /compat/ubuntu2210/usr/share/blender/scripts/addons/cycles/lib blender-3.2.2-linux-圆4/3.2/scripts/addons/cycles/lib /compat/ubuntu2210/usr/share/blender/scripts/addons/cycles/ I’m very curious to understand the reason(s) of the errors I see below : libGL error: glx: failed to create dri2 screen ![]() Why do I use root ? because as a normal user Blender does not start at all. Saved session recovery to '/tmp/quit.blend' var/run/user/1001/gvfs/ non-existent directory LibGL error: failed to load driver: nouveauĬould not get a list of mounted file-systems LibGL error: glx: failed to create dri2 screen Read prefs: /root/.config/blender/3.2/config/userpref.blend Is this supposed to work ? The error that Blender gives when I try to do that are the the following ones : blender What I want to do is to run Blender and I want to render my projects with cycles using the CUDA libraries and my GPU(s) within the linux emulation layer. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |