Search results

Upgrade to transformers v5.5

Status: Open.

#1777 In NVIDIA-NeMo/Automodel;

· athitten opened on Apr 10, 2026

TypeError: Qwen3_5MoeGatedDeltaNet.forward() got an unexpected keyword argument 'cache_position'

bugSomething isn't workingSomething isn't working

Status: Open.

#1770 In NVIDIA-NeMo/Automodel;

· sharonyu-115 opened on Apr 10, 2026

NotImplementedError: aten::equal on meta tensors during multi-GPU model init with transformers >= 5.4.0

bugSomething isn't workingSomething isn't working

Status: Open.

#1765 In NVIDIA-NeMo/Automodel;

· sharonyu-115 opened on Apr 10, 2026

Add convert_single_tensor_to_hf implementation for Gemma4 26B MoE model

bugSomething isn't workingSomething isn't working

Status: Open.

#1761 In NVIDIA-NeMo/Automodel;

· sharonyu-115 opened on Apr 10, 2026

Fix PEFT adapter outputs for v5 support

bugSomething isn't workingSomething isn't working

good first issueGood for newcomersGood for newcomers

MoE

Status: Open.

#1755 In NVIDIA-NeMo/Automodel;

· adil-a opened on Apr 9, 2026

·  

26.06

Guidance on adapting a fine-tuning recipe into a pretraining recipe

community-request

enhancementNew feature or requestNew feature or request

needs-follow-upIssue needs follow-upIssue needs follow-up

Status: Open.

#1746 In NVIDIA-NeMo/Automodel;

· wswaq opened on Apr 9, 2026

Support LLaVA-OneVision-1.5 Model Training

community-request

enhancementNew feature or requestNew feature or request

good first issueGood for newcomersGood for newcomers

Status: Open.

#1744 In NVIDIA-NeMo/Automodel;

· randydl opened on Apr 9, 2026

Add optimized DSA kernels for GLM 5/5.1, DSv3.2

enhancementNew feature or requestNew feature or request

Status: Open.

#1715 In NVIDIA-NeMo/Automodel;

· hemildesai opened on Apr 7, 2026

Validate gemma4 31B and 26B MoE recipes on DGXSpark

Status: Open.

#1713 In NVIDIA-NeMo/Automodel;

· athitten opened on Apr 7, 2026

Support Qwen-Image finetune (T2I)

enhancementNew feature or requestNew feature or request

good first issueGood for newcomersGood for newcomers

Status: Open.

Feature

#1700 In NVIDIA-NeMo/Automodel;

· akoumpa opened on Apr 6, 2026

Remove separate moe_mesh references after https://github.com/NVIDIA-NeMo/Automodel/pull/1684

enhancementNew feature or requestNew feature or request

Status: Open.

#1699 In NVIDIA-NeMo/Automodel;

· hemildesai opened on Apr 6, 2026

Training instability: Periodic loss fluctuations and sinusoidal grad_norm patterns in Llama 3.2 (1B/500M) pre-training

bugSomething isn't workingSomething isn't working

community-request

Status: Open.

#1679 In NVIDIA-NeMo/Automodel;

· ji-chengzhi opened on Apr 4, 2026