GPU support + examples #203
Conversation
* add a GPU Dockerfile * add examples utilizing GPU on ray-on-golem
shadeofblue
force-pushed
the
blue/gpu
branch
from
c53486f to
a7c8bf4
Compare
examples/GPU.Dockerfile
Outdated
| rsync \ | ||
| vim \ | ||
| curl \ | ||
| wget \ |
There was a problem hiding this comment.
Please stick to one: tabulators or spaces
examples/gpu.yaml
Outdated
| node_config: | ||
| # Parameters for golem demands (same for head and workers) | ||
| demand: | ||
| image_tag: "blueshade/raytest:0.9.0-py3.10-ray2.9.2-cuda11.8" |
There was a problem hiding this comment.
I'm unsure if I get the question but anyway, yeah, maybe we can upload the image to the production repo in the registry when we release a new version... otherwise, we never put images from develop in the golem/ray-on-golem repo...
There was a problem hiding this comment.
raytest is a very raw name for an image. Can we change it to something more descriptive and accurate?
There was a problem hiding this comment.
that's not the name of the image, that's the name of the repository... but anyway, yeah, we can
examples/hello_gpu.py
Outdated
| from numba import cuda | ||
|
|
||
|
|
||
| def test(v: np.array): |
There was a problem hiding this comment.
Rename this function or add PyDoc please
|
|
||
|
|
||
| @ray.remote(num_gpus=1) | ||
| def cuda_test(width, height): |
There was a problem hiding this comment.
Rename this function or add PyDoc please
There was a problem hiding this comment.
actually, on second thought, this one seems quite accurate...
vm-nvidiaruntime