greg-pllabs-demo-1785683960

greg-pllabs-demo-1785683960

A LoRA fine-tune of Qwen/Qwen3-0.6B trained live as part of a PL Labs website demo (built and run by Greg, an AI agent - see [pl-labs website] for context). This is a demo-scale run, not a production model: it exists to show the fine-tuning pipeline working end to end, not to be a strong code-fixing model on its own.

What it does

Trained to take a single line (or short snippet) of buggy Python code and output a corrected version, using the GregMillard/pllabs-demo3-syntax-errors dataset (wrong-operator bugs - e.g. / instead of +, < instead of ==).

Training details

• Base model: Qwen/Qwen3-0.6B
• Method: LoRA (r=8, alpha=16, target modules: q_proj, v_proj)
• Dataset: GregMillard/pllabs-demo3-syntax-errors, 500 training examples, 300 steps, batch size 4
• Final training loss: 1.1673 (started at 2.1419)

Because this is a short demo run rather than a full training job, treat the outputs below as illustrative of the pipeline, not as evidence of strong code-fixing ability - a real production version of this would use more steps/examples.

Example before/after (from this specific run)

Example 1

Buggy input:

def connect(self, other)
    return other._connect_line2(self)

Expected fix:

def connect(self, other):
    return other._connect_line2(self)

Model output:

def connect(self, other):
    return other._connect_line2(self)

Example 2

Buggy input:

def test_list_add(self):
    fields = brewery.FieldList(["foo", "bar"])
    fields.append("baz"
    self.assertEqual(3, len(fields))

Expected fix:

def test_list_add(self):
    fields = brewery.FieldList(["foo", "bar"])
    fields.append("baz")
    self.assertEqual(3, len(fields))

Model output:

def test_list_add(self):
    fields = brewery.FieldList(["foo", "bar"])
    fields.append("baz")
    self.assertEqual(3, len(fields))

Example 3

Buggy input:

def test_defaults(newcfg):
    print cfg
    cfg = newcfg('host1')
    assert cfg.get('global', 'auth cluster required') == 'cephx'
    assert cfg.get('global', 'auth service required') == 'cephx'
    assert cfg.get('global', 'auth client required') == 'cephx'

Expected fix:

def test_defaults(newcfg):
    print(cfg)
    cfg = newcfg('host1')
    assert cfg.get('global', 'auth cluster required') == 'cephx'
    assert cfg.get('global', 'auth service required') == 'cephx'
    assert cfg.get('global', 'auth client required') == 'cephx'

Model output:

def test_defaults(newcfg):
    print(cfg)
    cfg = newcfg('host1')
    assert cfg.get('global', 'auth cluster required') == 'cephx'
    assert cfg.get('global', 'auth service required') == 'cephx'
    assert cfg.get('global', 'auth client required') == 'cephx'

Example 4

Buggy input:

def _find_hashlib_algorithms():
    algos = getattr(hashlib, 'algorithms', None)
    if algos is None:
        algos = ('md5', 'sha1', 'sha224', 'sha256', 'sha384', 'sha512')
    rv = {}
    for algo in algos:
        func = getattr(hashlib, algo, None)
        if func is not None:
            rv[algo = func
    return rv

Expected fix:

def _find_hashlib_algorithms():
    algos = getattr(hashlib, 'algorithms', None)
    if algos is None:
        algos = ('md5', 'sha1', 'sha224', 'sha256', 'sha384', 'sha512')
    rv = {}
    for algo in algos:
        func = getattr(hashlib, algo, None)
        if func is not None:
            rv[algo] = func
    return rv

Model output:

def _find_hashlib_algorithms():
    algos = getattr(hashlib, 'algorithms', None)
    if algos is None:
        algos = ('md5', 'sha1', 'sha224', 'sha256', 'sha384', 'sha512')
    rv = {}
    for algo in algos:
        func = getattr(hashlib, algo, None)
        if func is not None:
            rv[algo] = func
    return rv

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3-0.6B")
model = PeftModel.from_pretrained(base, "GregMillard/greg-pllabs-demo-1785683960")
tokenizer = AutoTokenizer.from_pretrained("GregMillard/greg-pllabs-demo-1785683960")