Search results

Feature Request: Discrete Diffusion Language Models

enhancementNew feature or requestNew feature or request

Status: Open.

#2728 In NVIDIA/Megatron-LM;

· sbhavani opened on Dec 20, 2025

Feature Request: On-the-fly tokenization for dataloader

enhancementNew feature or requestNew feature or request

module: data pipeline

Status: Open.

#2727 In NVIDIA/Megatron-LM;

· sbhavani opened on Dec 20, 2025

Information or check added in get_num_layers_to_build

community-request

questionFurther information is requestedFurther information is requested

Status: Open.

#2718 In NVIDIA/Megatron-LM;

· lixin-sxty opened on Dec 19, 2025

Consider SonicMOE kernel comparison for potentially better IO/activation memory management

community-request

enhancementNew feature or requestNew feature or request

module: moe

Status: Open.

#2709 In NVIDIA/Megatron-LM;

· Skylion007 opened on Dec 18, 2025

[gated attention] Support Gated Attention from NeurIPS 25 best paper

community-request

enhancementNew feature or requestNew feature or request

Status: Open.

#2704 In NVIDIA/Megatron-LM;

· louiswang524 opened on Dec 18, 2025

[QUESTION] How Permute Kernel don't copy the tokens top-k times in deepseek-v3-gb200-optimization.md

community-request

module: moe

questionFurther information is requestedFurther information is requested

Status: Open.

#2692 In NVIDIA/Megatron-LM;

· lk137095576 opened on Dec 17, 2025

[QUESTION] In what cases will tp_comm_overlap accelerate training

community-request

questionFurther information is requestedFurther information is requested

Status: Open.

#2662 In NVIDIA/Megatron-LM;

· Cccei000 opened on Dec 15, 2025

Update docstrings for MCore

docs-onlydocumentation only (docs or docstrings)documentation only (docs or docstrings)

Status: Open.

#2653 In NVIDIA/Megatron-LM;

· Phlip79 opened on Dec 12, 2025

Feature Request: NCCL GIN (GPU-Initiated Networking) Support for AWS EFA

enhancementNew feature or requestNew feature or request

module: moe

Status: Open.

#2647 In NVIDIA/Megatron-LM;

· sbhavani opened on Dec 12, 2025

[BUG] dist_muon checkpoint load error

bugSomething isn't workingSomething isn't working

community-request

module: optimizer

Status: Open.

#2625 In NVIDIA/Megatron-LM;

· lk137095576 opened on Dec 11, 2025

Integrate AI reviewer

enhancementNew feature or requestNew feature or request

Status: Open.

Initiative

#2621 In NVIDIA/Megatron-LM;

· Phlip79 opened on Dec 10, 2025

Overhaul oncall

enhancementNew feature or requestNew feature or request

Status: Open.

Initiative

#2620 In NVIDIA/Megatron-LM;

· Phlip79 opened on Dec 10, 2025