Fix(gemma3_mm): Add robust weight loading for quantized VLM#20066

Flink-ddd wants to merge 1 commit intovllm-project:mainvllm-project/vllm:mainfrom

Flink-ddd:fix-gemma3-final-v2Flink-ddd/vllm:fix-gemma3-final-v2Copy head branch name to clipboard