Skip to content

Labels

Labels

63 labels

  • models

    models
    Model management issues

  • performance

    performance
    Performance improvement

  • phase:p0

    phase:p0
    P0: Correctness bugs — fix before anything else

  • phase:p1

    phase:p1
    P1: Foundation refactor — prerequisite for new adapter methods

  • phase:p2

    phase:p2
    P2: Infrastructure, reliability & observability

  • phase:p3

    phase:p3
    P3: New adapter methods — LoRA/QLoRA/AdaLoRA/IA³/Prefix/Prompt strategy registry

  • phase:p3b

    phase:p3b
    P3b: Code quality — componentize finetune.py, break up god-node state

  • phase:p4

    phase:p4
    P4: UX quality — intent filtering, simple/advanced mode, rank warnings

  • phase:p5

    phase:p5
    P5: Wizard paths for DPO and Knowledge Distillation

  • phase:p6

    phase:p6
    P6: VLM fine-tuning — LLaVA, Qwen2-VL LoRA support

  • phase:p7

    phase:p7
    P7: Adapter composition — PeftMixedModel stacking (researcher feature)

  • phase:p8

    phase:p8
    P8: Repo cleanup — remove stale docs, untrack cache dirs (run last)

  • priority:high

    priority:high
    Required for the first complete product workflow

  • priority:low

    priority:low
    Useful after the main platform is stable

  • priority:medium

    priority:medium
    Important but can follow the core workflow

  • question

    question
    Further information is requested

  • refactor

    refactor
    Code restructuring and cleanup

  • size/L

    size/L

  • size/M

    size/M

  • size/S

    size/S

  • size/XL

    size/XL

  • size/XS

    size/XS

  • stale

    stale

  • testing

    testing

  • training

    training
    Training pipeline issues

  • type:docs

    type:docs
    Documentation work

  • type:enhancement

    type:enhancement
    Improvement to an existing workflow

  • type:feature

    type:feature
    New product capability

  • type:research

    type:research
    Investigation or technical spike

  • ui

    ui
    UI/UX related