Search results

[bug] async_grpo with in flight weight updates hangs

bugSomething isn't workingSomething isn't working

community-request

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

Status: Open.

#2511 In NVIDIA-NeMo/RL;

· bzantium opened on May 17, 2026

[Feature] Muon optimizer support for the DTensor backend

community-request

Feature

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

feat: support context parallelism for reward model training on DTensor backend

enhancementNew feature or requestNew feature or request

from-slack

Status: Open.

#2482 In NVIDIA-NeMo/RL;

· terrykong opened on May 13, 2026

Flag name should_log_nemo_gym_responses is confusing

bugSomething isn't workingSomething isn't working

Status: Open.

#2472 In NVIDIA-NeMo/RL;

· 1ytic opened on May 12, 2026

[megatron-policy] refit fails on Qwen3.5-MoE A17B with PP > 1: "Object present on multiple PP ranks" due to MTP shared embedding

bugSomething isn't workingSomething isn't working

community-request

DocumentationImprovements or additions to documentationImprovements or additions to documentation

Memory

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

Status: Open.

#2471 In NVIDIA-NeMo/RL;

· linmuchuiyang opened on May 12, 2026

GB200 GRPO training hangs around step 3-4 during DTensor policy offload/refit

bugSomething isn't workingSomething isn't working

Speed

Status: Open.

#2462 In NVIDIA-NeMo/RL;

· jQizhang opened on May 11, 2026