AI workstation setup?

[Syirrus (John)]

Hello all!

Is there a up-to-gate guide out there that will help noobs like me setup there own ML workstation?
I have an Nvidia card and windows 10 (I can use Linux too).

Syirrus

[AyersAuthentic (Kevin Ayers)]

An answer to this question would really help me too. I’ve been finding a mix of setup instructions that seem to be wrong and outdated.

[florianl (Florian)]

Tim Dettmers has a quite comprehensive guide:

Tim Dettmers

Also there’s a lot of information in this thread:

Recommendations on new 2 x RTX 3090 setup

Hi, I’m selling my old GTX 1080 and upgrading my deep learning server with a new RTX 3090. I’m also contemplating adding one more RTX 3090 later next year. I’ve read from multiple sources blower-style cooling is recommended when having two or more GPUs. There are not many blower-style options confirmed yet. As a matter of fact, the only one I could find is the Founders Edition. Most (all?) AIB cards seem to opt for either double/triple fan or water-cooling. My only concern about the Founders …

I built a workstation (Intel i9, 64 GB RAM and a RTX 3090) recently - and I have to admit it was not as easy as I thought But all is working now and makes things way more fun :).

[Syirrus (John)]

After looking around, I finally found some information that was pretty straight forward for setting things up. STEPS:

1. Install Ubuntu 20.04 LTS and all updates.
2. Install latest Nvidia drivers.
3. Install latest Anaconda.

I used this guide to accomplish the rest and surprisingly it worked.

Platform: Local Server - Ubuntu

My apologies in advance for the crudeness of this post, but I wanted to provide some setup instructions to those who intend on running the fastai code without using the cloud options. I know Jeremy has recommended that new users focus on learning the deep learning methodologies over troubleshooting a local installation, but the instructions below should be easy enough to follow. We are coders, right? They will show how to get fastai versions 1 and 2 up and running independently with access to b…

My setup is a SFFpc with a 3090. It runs really well so far.

[AyersAuthentic (Kevin Ayers)]

Hey Syiruss, thanks for the response. What version of cuda/cuDNN did you use. That’s the part that’s tripping me up.

[Syirrus (John)]

No worries man.

Unlike with tensorflow, I didn’t install CUDA or CDNN, just the latest nvidia drivers for Ubuntu via package manager. I think when you install fast.ai and all of it’s dependencies it will grab the correct version of CUDA and CDNN.

I ran the introduction model notebook and monitored my GPU and it ran on my GPU. The GPU pulled 350 watts and used 12 gigs of GPU memory when it was training the models in the intro jupyter notebook.

So, from this perspective it looks like it is working, unless I missed something.

[florianl (Florian)]

i had trouble with the driver packages installed with apt so I downloaded the Nvidia toolkit here:

CUDA Toolkit 11.2 Downloads

im running 11.1 / 455 currently (cuda_11.1.1_455.32.00_linux.run).

To install Pytorch:

pip install torch==1.7.0+cu110 torchvision==0.8.1+cu110 torchaudio===0.7.0 -f https://download.pytorch.org/whl/torch_stable.html

[balnazzar (Andrea de Luca)]

@Syirrus

My setup is a SFFpc with a 3090. It runs really well so far.

SFF? Interesting!

May you provide additional details about your setup and maybe some pics? Thanks!

[parveenkr (Parveen)]

Hello Syirrus!

To set up your ML workstation, first install the latest NVIDIA drivers, CUDA Toolkit, and cuDNN for optimal GPU performance. Then, use Anaconda to create a Python environment for managing ML libraries like TensorFlow and PyTorch. Optionally, install Jupyter Notebook for easier project organization. This setup works well on both Windows and Linux—good luck!

Here are some AI Workstations references from ProX PC www.proxpc.com [ Page All AI Workstations