[https://nvbugs/6418815][test] Pin fp32-matmul precision in VisualGen LPIPS tests and re-baseline Cosmos3 goldens - #17780

[https://nvbugs/6418815][test] Pin fp32-matmul precision in VisualGen LPIPS tests and re-baseline Cosmos3 goldens#17780

chang-l merged 15 commits into

NVIDIA:mainNVIDIA/TensorRT-LLM:mainfrom

ishovkun:cosmos3-rebaseline-goldensishovkun/TensorRT-LLM:cosmos3-rebaseline-goldensCopy head branch name to clipboard