Search results

[Feature] Add same-tokenizer off-policy distillation support

enhancementNew feature or requestNew feature or request

Feature

Status: Open.

#2545 In NVIDIA-NeMo/RL;

· RayenTian opened on May 22, 2026

Support GRPO for Mistral 3.5

enhancementNew feature or requestNew feature or request

Feature

Status: Open.

#2542 In NVIDIA-NeMo/RL;

· sharonyu-115 opened on May 21, 2026

Add agent_ref level metrics in training for GRPO and OPD

enhancementNew feature or requestNew feature or request

Feature

Status: Open.

#2537 In NVIDIA-NeMo/RL;

· jkyi-nvidia opened on May 20, 2026

Unify duplicated deep_ep dependency commits in pyproject.toml

DocumentationImprovements or additions to documentationImprovements or additions to documentation

Status: Open.

#2522 In NVIDIA-NeMo/RL;

· terrykong opened on May 18, 2026

[bug] async_grpo with in flight weight updates hangs

bugSomething isn't workingSomething isn't working

community-request

waiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respond

Status: Open.

#2512 In NVIDIA-NeMo/RL;

· kaloyan-inherent opened on May 17, 2026

refactor(algorithms): align max_num_steps / max_num_epochs semantics with NeMo Automodel

community-request

Feature

waiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respond

Status: Open.

#2511 In NVIDIA-NeMo/RL;

· bzantium opened on May 17, 2026

[Feature] Muon optimizer support for the DTensor backend

community-request

Feature

waiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respond

Status: Open.

#2510 In NVIDIA-NeMo/RL;

· bzantium opened on May 16, 2026

[data-plane] cleanup tracking issue

Status: Open.

#2509 In NVIDIA-NeMo/RL;

· ZhiyuLi-Nvidia opened on May 16, 2026

GRPO/Distillation validation collapses multi-dataset metrics into a single accuracy

bugSomething isn't workingSomething isn't working

community-request

Feature

waiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respond

Status: Open.

#2501 In NVIDIA-NeMo/RL;

· bzantium opened on May 15, 2026

SFT on Qwen 1B with PP=2 VPP=2 fails for megatron backend

accuracy

bugSomething isn't workingSomething isn't working

Status: Open.

#2494 In NVIDIA-NeMo/RL;

· guyueh1 opened on May 14, 2026

Slowness in KD with Nemo RL during nemotron training

bugSomething isn't workingSomething isn't working

Speed

Status: Open.

#2490 In NVIDIA-NeMo/RL;

· snowmanwwg opened on May 13, 2026

Support precomputing reference log probabilities for DPO training

community-request

enhancementNew feature or requestNew feature or request

waiting-on-maintainersWaiting on maintainers to respondWaiting on maintainers to respond

Status: Open.

#2484 In NVIDIA-NeMo/RL;

· ogaloglu opened on May 13, 2026