| |
| |
| |
| |
| |
| |
| |
| |
| """One shared-adapter experiment; private mounted artifacts, no model publishing.""" |
| import argparse |
| import hashlib |
| import json |
| import subprocess |
| import sys |
| import zipfile |
| from pathlib import Path |
|
|
|
|
| def main(): |
| parser = argparse.ArgumentParser(description=__doc__) |
| parser.add_argument('--inputs', type=Path, required=True) |
| parser.add_argument('--output', type=Path, required=True) |
| parser.add_argument('--revision', required=True) |
| parser.add_argument('--repo', default='murabcd/fev-4b') |
| parser.add_argument('--steps', type=int, default=6) |
| args = parser.parse_args() |
| if len(args.revision) != 40 or any(c not in '0123456789abcdef' for c in args.revision): |
| raise ValueError('Pin an immutable 40-character Hub commit') |
| args.output.mkdir(parents=True, exist_ok=False) |
| names = ('source.zip', 'train.jsonl', 'development.jsonl', 'calibration.jsonl', 'test.jsonl') |
| intent = {'repo': args.repo, 'revision': args.revision, 'steps': args.steps, |
| 'inputHashes': {n: hashlib.sha256((args.inputs/n).read_bytes()).hexdigest() for n in names}, |
| 'published': False, 'automaticAdoption': False} |
| (args.output/'intent.json').write_text(json.dumps(intent, indent=2)+'\n') |
| code = Path('/tmp/fev-training-source') |
| with zipfile.ZipFile(args.inputs/'source.zip') as archive: |
| if any(Path(n).is_absolute() or '..' in Path(n).parts or |
| not n.startswith(('fev/', 'fev4b/')) or not n.endswith('.py') for n in archive.namelist()): |
| raise ValueError('Source ZIP must contain only fev/ and fev4b/ Python files') |
| archive.extractall(code) |
| from huggingface_hub import snapshot_download |
| model = snapshot_download(args.repo, revision=args.revision, local_dir='/tmp/fev-training-base') |
| def run(module, *parameters): |
| subprocess.run([sys.executable, '-m', module, '--model-dir', model, *map(str, parameters)], |
| cwd=code, check=True) |
| run('fev4b.train', '--train', args.inputs/'train.jsonl', '--development', args.inputs/'development.jsonl', |
| '--steps', args.steps, '--out', args.output/'adapter') |
| run('fev4b.calibrate', '--adaptation-dir', args.output/'adapter', '--data', args.inputs/'calibration.jsonl', |
| '--out', args.output/'calibrated') |
| run('fev4b.benchmark', '--adaptation-dir', args.output/'calibrated', '--data', args.inputs/'test.jsonl', |
| '--out', args.output/'test') |
| report = json.loads((args.output/'test/report.json').read_text()) |
| if report['overall']['errors']: |
| raise RuntimeError('Incomplete test predictions; do not adopt') |
| (args.output/'complete.json').write_text(json.dumps({'complete': True, 'adopted': False})+'\n') |
|
|
|
|
| if __name__ == '__main__': |
| main() |
|
|
