davidwdw's picture
Upload B1K model checkpoints, training recipes, and evaluation artifacts (part 2)
0c3a82b verified
Raw History Blame Contribute Delete
1.35 kB
"""Resume only statistics and figures after all sampling is complete."""
import fcntl
import json
import subprocess
import sys
import traceback
from common import ROOT, VARIANTS, conditions, local_only, progress, write_json, sha
if __name__ == '__main__':
local_only()
lock=(ROOT/'campaign.lock').open('a')
fcntl.flock(lock,fcntl.LOCK_EX|fcntl.LOCK_NB)
try:
assert all((ROOT/'samples'/v/f'{c}.npz').exists() for v in VARIANTS for c in conditions())
progress('status.json',state='analyzing',resumed_analysis=True)
write_json(ROOT/'analysis_source_manifest.json',{p.name:sha(p) for p in ROOT.glob('*.py')})
subprocess.run([sys.executable,str(ROOT/'analyze.py')],check=True)
subprocess.run([sys.executable,str(ROOT/'illustrate.py')],check=True)
validation=json.loads((ROOT/'validation.json').read_text())
validation.update(png_figures=len(list((ROOT/'figs').glob('*.png'))),
pdf_figures=len(list((ROOT/'figs').glob('*.pdf'))),report_sha256=sha(ROOT/'report.md'))
write_json(ROOT/'validation.json',validation)
progress('status.json',state='complete',conditions=20,models=3,samples=30720,report=str(ROOT/'report.md'))
except BaseException as e:
progress('status.json',state='failed',error=repr(e));traceback.print_exc();sys.exit(1)