(beam) branch fix-vllm-gemma updated: Use 12.2.2 cuda

[vt...@apache.org]

This is an automated email from the ASF dual-hosted git repository.

vterentev pushed a commit to branch fix-vllm-gemma
in repository https://gitbox.apache.org/repos/asf/beam.git


The following commit(s) were added to refs/heads/fix-vllm-gemma by this push:
     new 48c23ac0b94 Use 12.2.2 cuda
48c23ac0b94 is described below

commit 48c23ac0b94fc2c757069f094446489e18b1e329
Author: Vitaly Terentyev <vi...@akvelon.com>
AuthorDate: Sat Sep 27 11:11:27 2025 +0400

    Use 12.2.2 cuda
---
 sdks/python/apache_beam/ml/inference/test_resources/vllm.dockerfile | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/sdks/python/apache_beam/ml/inference/test_resources/vllm.dockerfile b/sdks/python/apache_beam/ml/inference/test_resources/vllm.dockerfile
index 0fe0809a05e..b8398381051 100644
--- a/sdks/python/apache_beam/ml/inference/test_resources/vllm.dockerfile
+++ b/sdks/python/apache_beam/ml/inference/test_resources/vllm.dockerfile
@@ -17,7 +17,7 @@
 # Used for any vLLM integration test
 # Dockerfile — Beam dev harness + install dev SDK from LOCAL source package
 
-FROM nvidia/cuda:12.4.1-devel-ubuntu22.04
+FROM nvidia/cuda:12.2.2-devel-ubuntu22.04
 
 # 1) Non-interactive + timezone
 ENV DEBIAN_FRONTEND=noninteractive \