Issues

• [Feature] Support user-selectable LoRA adapter files at load/runtime

  architectureenhancementNeed Review
  nntrainer/Quick.AI#37 · Filter by author dlwlzzero opened on Jun 5, 2026
  

• [Feat(API)] Accept full accumulated context and compute only the unprocessed delta via recorded context length

  apienhancement
  nntrainer/Quick.AI#34 · Filter by author dlwlzzero opened on Jun 4, 2026
  

• [Feasibility] Unify the nntrainer KV-cache manager with the QNN KV-cache manager

  architectureNeed Review
  nntrainer/Quick.AI#33 · Filter by author dlwlzzero opened on Jun 4, 2026
  

• [Feat] Support concurrent multi-session inference

  apiarchitectureenhancement
  nntrainer/Quick.AI#31 · Filter by author dlwlzzero opened on Jun 4, 2026
  

• [Feat] Expose per-step logit values through the model-run API

  apienhancement
  nntrainer/Quick.AI#30 · Filter by author dlwlzzero opened on Jun 4, 2026
  

• [model] Support SLM

  nntrainer/Quick.AI#26 · Filter by author Seunghui98 opened on May 11, 2026

• [nntrainer/PR] port #3850 from upstream nntrainer

  nntrainer/Quick.AI#19 · Filter by author EunjuYang opened on Apr 29, 2026

• [Issue] 30B MoE with FSU (slimMoE, slimCachedMoE) fails with mmap-read=true option

  nntrainer/Quick.AI#13 · Filter by author EunjuYang opened on Apr 28, 2026

• [model] Support gliner2-multi-v1 model

  nntrainer/Quick.AI#4 · Filter by author Seunghui98 opened on Apr 24, 2026

• [model] Support tiny-BERT

  nntrainer/Quick.AI#3 · Filter by author Seunghui98 opened on Apr 24, 2026