[bitnami/vllm] feat: Add new container

[zhekazuev]

Name and Version

bitnami/vllm - 0.6.2, 0.6.3 and 0.6.4

Is it possible to add images for latest and previous versions?

Version 0.6.4 has some critical issues related gpu memory usage:

• [Bug]: Breaking Change in gpu_memory_utilization Behavior in vLLM 0.6.4 vllm-project/vllm#10451
• [core] overhaul memory profiling and fix backward compatibility vllm-project/vllm#10511
• [Bugfix] GPU memory profiling should be per LLM instance vllm-project/vllm#10498

And related default LLM that should be saved in docker image, looks like vLLM project uses facebook/opt-125m model.

What is the problem this feature will solve?

• add new container for vllm - it makes possible to deploy LLMs to kubernetes cluster with vLLM engine.
• this image can be used in a potential Helm Chart by Bitnami - [bitnami/vllm] feat: Add chart charts#30649

What is the feature you are proposing to solve the problem?

• add Docker container for vllm

What alternatives have you considered?

• use official image by vllm-project team
• use llama-based official image - added as an example

[carrodher]

Thank you for your suggestion!

We truly appreciate your input. Our team will now conduct an internal review to assess the viability of integrating this new open-source product into our catalog. As we have limited resources, please understand that this evaluation process might take some time. To ensure that your idea receives the attention it deserves, we will apply the 'on-hold' label. This will prevent our automated system from prematurely closing the issue.

We will keep you updated on any developments regarding the addition of this product. Your patience and understanding are greatly appreciated.

[carrodher]

Hi, unfortunately after an internal review, and due to other priorities we have, this solution was not considered to be added to the catalog in the short/mid-term.

We apologize for the inconvenience. We will reconsider it in the future.

Thanks for your suggestion!

[zhekazuev]

Thank you for your response!