Adding runtimecatalog on the CML service

[#1]

I'd like to add runtimecatalog on the CML service. I'm currently using 12.8cuda version and Nvidia H100Gpu, and I'd like to use 3.12 jupyterlab. Is this version supported? When I looked it up on github, there didn't seem to be anything. Or can I customize it and use it after image build? Please advise
Thank you.

[#2]

Hello @jI-mi, 

Thanks for being part of our community. 

Can you please explain your question a little bit more? 
Also, can you please tell us your CML version? 

I was quickly checking on CML 1.5.5 and the CUDA 12.8 is not listed as supported, so this may also give you some hints: 
https://docs.cloudera.com/machine-learning/cloud/runtimes/topics/ml-runtimes-nvidia-gpu.html 

[#3]

First of all, thank you for your answer. The CML version we are using is 2.0.42-b80. I'm inquiring because I was asked by the Data Science team that they want to develop it in the 3.12jupyterlab and cuda12.8 environment

[#4 · Accepted solution]

Hello @jI-mi, 

That CML version is kind of old. If I'm not wrong, that version comes with DS 1.5.2. 
There the only supported versions for CUDA where 11.4 and 11.1, so you will not be able to do that. 
https://docs-archive.cloudera.com/machine-learning/1.5.2/runtimes/topics/ml-runtimes-nvidia-gpu.html

And actually, even the most recent versions only support CUDA 12.5.1

Due to Tensorflow requirements, looks like CUDA 12.8 is not planned to be supported. 
You can check with your Cloudera Account Team to request this feature providing your business needs.  

[#5]

@jI-mi, Did the response help answer your question? If so, please consider marking it as the solution so others can easily find it later.