nemo-rl

feat: additional kl metrics (#1420)

fix: Adding mean total tokens per sample to the output log (#1406)

fix: Fix grad norm metric in mcore path (#1426)

chore: major version bump (torch 2.8, vllm 0.11, ray 2.49) & SP fixes (#1334)

fix: append to hf_overrides rather than overwriting (#1413)