marzoukbaig14/committed-qwen3-1.7b-lora

Committed — Qwen3-1.7B LoRA adapter

A QLoRA adapter that fine-tunes Qwen/Qwen3-1.7B to write Conventional Commits messages from a git diff: a single-file diff in, one type(scope): description subject line out.

This repo holds the LoRA adapter weights. For local CPU inference most people want the merged, quantized GGUF instead — it's what the serving layer uses. Use this adapter if you want to merge it yourself, train further on top of it, or run it with PEFT on GPU.

Details

• Base: Qwen/Qwen3-1.7B (Apache-2.0)
• Method: QLoRA (PEFT LoRA + TRL SFTTrainer, vanilla transformers)
• Task: single-file git diff → one Conventional Commits subject line
• Trained on: marzoukbaig14/committed-train (~58k filtered CommitChronicle commits, 16 languages)

Usage

The trained behavior depends on the exact prompt rendering used in training (a canonical zero-shot Diff:\n{diff} format with enable_thinking=False) plus the GBNF grammar applied at decode time. Loading the adapter with a bare prompt will not reproduce the evaluated output. To match what was evaluated, run it through the project's engine.py, or use the FastAPI / Gradio Space. See github.com/marzoukbaig14/Committed.

To load the adapter on top of the base for your own use:

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-1.7B")
model = PeftModel.from_pretrained(base, "marzoukbaig14/committed-qwen3-1.7b-lora")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-1.7B")

Results

Evaluated against the un-tuned base on a 442-example test set, scored by an LLM judge on four axes (judge validated against 50 hand-rated examples). Headline numbers reweighted to the true commit-type distribution.

Metric	Base	Fine-tuned
Type accuracy	0.131	0.637
Conjunctive pass-rate	0.181	0.471
Graded mean (0–3)	1.207	2.188
Faithfulness	0.43	0.86

The base model collapsed ~95% of outputs to feat regardless of the diff; fine-tuning fixed that. One axis (specificity) regressed slightly (0.81 → 0.71). The full before/after, the regression analysis, and where the model disagrees with gold labels (sometimes the model is the more defensible call) are in the eval writeup: FINDINGS_v1.md.

Related

• GGUF, for local CPU serving — marzoukbaig14/committed-gguf. The merged, quantized model most people actually want to run.
• Training dataset — marzoukbaig14/committed-train. The ~58k filtered CommitChronicle commits this was trained on, with composition tables and provenance.
• Source, training & eval code — github.com/marzoukbaig14/Committed.
• Eval writeup — FINDINGS_v1.md. How it was evaluated and how it did, honestly.
• Live demo — try it in the browser.

License

Apache-2.0, inherited from the Qwen3-1.7B base.

Citation

Trained with TRL. Dataset derived from CommitChronicle (Eliseeva et al., From Commit Message Generation to History-Aware Commit Message Generation, arXiv:2308.07655).