From f237011618f2e2c0f08d4731ad06a6fd893af183 Mon Sep 17 00:00:00 2001 From: Eric Curtin Date: Tue, 25 Feb 2025 00:37:54 +0000 Subject: [PATCH] Use vllm-openai upstream image The one we are currently using is old and doesn't have .gguf compatibility. Signed-off-by: Eric Curtin --- ramalama/model.py | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) diff --git a/ramalama/model.py b/ramalama/model.py index 1463268e..23952b24 100644 --- a/ramalama/model.py +++ b/ramalama/model.py @@ -148,10 +148,7 @@ def _image(self, args): gpu_type, _ = next(iter(env_vars.items())) if args.runtime == "vllm": - if gpu_type == "HIP_VISIBLE_DEVICES": - return "quay.io/modh/vllm:rhoai-2.18-rocm" - - return "quay.io/modh/vllm:rhoai-2.18-cuda" + return "docker.io/vllm/vllm-openai" split = version().split(".") vers = ".".join(split[:2])