Search results

qwen3_moe_30b_hellaswag OOMs during model loading on single GPU

bugSomething isn't workingSomething isn't working

Status: Open.

#2041 In NVIDIA-NeMo/Automodel;

· torsli opened on Apr 24, 2026

Support DSV4 flash

enhancementNew feature or requestNew feature or request

Status: Open.

#2034 In NVIDIA-NeMo/Automodel;

· HuiyingLi opened on Apr 24, 2026

Upgrade to DeepEPv2

enhancementNew feature or requestNew feature or request

Status: Open.

#2021 In NVIDIA-NeMo/Automodel;

· akoumpa opened on Apr 23, 2026

·  

26.06

Support NVIDIA-Nemotron-3-Nano-4B-BF16

enhancementNew feature or requestNew feature or request

Status: Open.

#2004 In NVIDIA-NeMo/Automodel;

· pzelasko opened on Apr 22, 2026

GPT-OSS 120B HF load fails with "Missing key in checkpoint state_dict: lm_head.weight"

bugSomething isn't workingSomething isn't working

community-request

Status: Open.

#1952 In NVIDIA-NeMo/Automodel;

· jqwang2373 opened on Apr 21, 2026

Integrate Mega MoE

enhancementNew feature or requestNew feature or request

Status: Open.

#1918 In NVIDIA-NeMo/Automodel;

· akoumpa opened on Apr 20, 2026

use lazy map for dataset preprocessing

enhancementNew feature or requestNew feature or request

good first issueGood for newcomersGood for newcomers

Status: Open.

#1917 In NVIDIA-NeMo/Automodel;

· akoumpa opened on Apr 20, 2026

Does Gemma 4 support Context Parallelism (CP)?

community-request

enhancementNew feature or requestNew feature or request

waiting-on-customerWaiting on the original author to respondWaiting on the original author to respond

Status: Open.

#1912 In NVIDIA-NeMo/Automodel;

· liyandong001 opened on Apr 20, 2026

[Bug] Gemma4 MoE backend missing vision-aware attention mask for use_bidirectional_attention="vision" models

bugSomething isn't workingSomething isn't working

Status: Open.

#1891 In NVIDIA-NeMo/Automodel;

· jQizhang opened on Apr 17, 2026

Gemma4 26B-A4B MoE Expert Weights float32 After FSDP2 Prefetch PR

bugSomething isn't workingSomething isn't working

Status: Open.

#1863 In NVIDIA-NeMo/Automodel;

· sharonyu-115 opened on Apr 15, 2026

bug: MoE expert LoRA checkpoints incompatible with HF PEFT and vLLM (Nemotron Super 120B)

community-request

good first issueGood for newcomersGood for newcomers

Status: Open.

#1814 In NVIDIA-NeMo/Automodel;

· pst2154 opened on Apr 13, 2026

Optimize causal-conv1d/transformer engine installation in CI

Automation

enhancementNew feature or requestNew feature or request

good first issueGood for newcomersGood for newcomers

Status: Open.

#1796 In NVIDIA-NeMo/Automodel;

· akoumpa opened on Apr 13, 2026