sql-agent / src /models

Commit History

  1. Result narration (Qwen-powered) + UI polish for upload area + dark-mode fixes

    Verified signature
    5ae5227
  2. SQL self-correction: 3 retries with error context + GROUP BY rule in prompt

    Verified signature
    e6681c9
  3. SVG: standalone download (XML prolog, white bg, fixed dims) + Plotly first

    Verified signature
    0ac4ef2
  4. Better SQL prompt: column-name mapping rules + few-shot + distinct value hints

    Verified signature
    a7fa6d3

  1. Load LoRA via PeftModel on top of standard base models (fixes r=16 vs r=8 mismatch)

    Verified signature
    1bbdff9
  2. Load LoRA via PeftModel on top of standard base models (fixes r=16 vs r=8 mismatch)

    Verified signature
    eb30a86
  3. Load LoRA via PeftModel on top of standard base models (fixes r=16 vs r=8 mismatch)

    Verified signature
    730b25d
  4. ZeroGPU best practice: load models at module level (cuda), inference only inside @spaces.GPU

    Verified signature
    61aee8d
  5. ZeroGPU best practice: load models at module level (cuda), inference only inside @spaces.GPU

    Verified signature
    a57eca6
  6. ZeroGPU best practice: load models at module level (cuda), inference only inside @spaces.GPU

    Verified signature
    c2ac226
  7. Sequential model load/unload: keep peak VRAM at ~14GB

    Verified signature
    fab3ba1
  8. Initial deploy: Apple/Claude design, DuckDB, 3 trained LoRAs

    Verified signature
    a067ada