Installing Nvidia Docker
The NVIDIA A5000 I bought this time Ah... I bought an A5000. I bought it to try natural language processing. Was it 325 or 335...? Anyway, I got it installed in
GGAMI
The NVIDIA A5000 I bought this time Ah... I bought an A5000. I bought it to try natural language processing. Was it 325 or 335...? Anyway, I got it installed in
Ah... I bought an A5000. I bought it to try natural language processing. Was it 325 or 335...?
Anyway, I got it installed in my server PC without any issues! Before this, not knowing any better, I foolishly bought a Radeon card... For machine learning and deep learning, it seems like NVIDIA is the way to go, no matter what (?).
All right... let’s see how to install it.
https://www.server-world.info/en/note?os=CentOS_Stream_8&p=nvidia&f=1
This is the page to check. So clean and clear. Really good.
nvidia-smi
The installation went really well. The A5000 shows up too!!
Now I need to install this in Docker...
https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html
Ah, this is the page to check... There’s nothing difficult about it.
This is all you need. I looked up all sorts of things, but there’s nowhere as clear and well-organized as this.
All right... that’s the installation done...
I’m currently downloading models from huggingface and using them. I’m using FastAPI to communicate with them and get results through an API.
FastAPI is often described as incredibly fast in Python, but what’s even more important is jupyter notebook.
The reason is... loading a model takes 3–4 minutes, and you have to wait that long every time you run Python (after changing the code)...
But if you install and use Jupyter Notebook, the model stays loaded in the process. That means you can immediately see the results of your code changes, which is a big advantage.
I feel a bit foolish for not knowing any better at first and spending 3–4 minutes each time I worked, but if you’re going to write, test, and develop code related to deep learning and machine learning, I absolutely recommend installing jupyter notebook.