[None][fix] Compute the Cosmos3 rotary table with a broadcast multiply, not a K=1 matmul - #19532

[None][fix] Compute the Cosmos3 rotary table with a broadcast multiply, not a K=1 matmul#19532

chang-l merged 8 commits into

NVIDIA:mainNVIDIA/TensorRT-LLM:mainfrom

ishovkun:fix/cosmos3-rope-tf32ishovkun/TensorRT-LLM:fix/cosmos3-rope-tf32Copy head branch name to clipboard