CodeFinetuner: fine-tune a local code autocomplete model on your own codebase

MountainTop321 · reddit · 2026-09-12

Redditor releases CodeFinetuner, an open-source end-to-end pipeline for LoRA fine-tuning small code autocomplete models (e.g. Qwen2.5-Coder-3B) on your own codebase, outputting GGUF for fully local use via llama.vim/llama.vscode. Supports Mac (MPS) and NVIDIA GPUs, with optional Unsloth.

Pipeline: raw code → tree-sitter parsing into Structure-Aware FIM examples → LoRA fine-tuning → eval (CodeBLEU, edit similarity, exact match, perplexity) → GGUF conversion.

The author cautions that test-set gains don't guarantee editor usefulness (autocomplete tools sample differently from the greedy decoding used in evals), so real-world verification is needed — but it also serves as a complete reference LoRA fine-tuning pipeline.

Original post →

More from coding & agent

coding & agent channel →