Buckets:
| #!/usr/bin/env python3 | |
| """Serve a local checkpoint and reader demo; weights load once, no provider calls.""" | |
| import argparse | |
| from http.server import BaseHTTPRequestHandler, HTTPServer | |
| import json | |
| import mimetypes | |
| from pathlib import Path | |
| import time | |
| from urllib.parse import unquote, urlparse | |
| def server_class(engine, web_root): | |
| class Handler(BaseHTTPRequestHandler): | |
| def send(self, code, content, mime='application/json; charset=utf-8'): | |
| self.send_response(code);self.send_header('Content-Type',mime);self.send_header('Content-Length',str(len(content))) | |
| self.send_header('Cache-Control','no-store');self.end_headers();self.wfile.write(content) | |
| def send_json(self, code, data): | |
| self.send(code,json.dumps(data,ensure_ascii=False,allow_nan=False).encode()) | |
| def do_GET(self): | |
| route=urlparse(self.path).path | |
| if route=='/api/health': | |
| self.send_json(200,{'ready':True,'model_loaded_once':True,'provider_calls':0});return | |
| relative=unquote(route).lstrip('/') or 'index.html' | |
| target=(web_root/relative).resolve() | |
| if not target.is_relative_to(web_root) or not target.is_file(): | |
| self.send_json(404,{'error':'File not found'});return | |
| self.send(200,target.read_bytes(),mimetypes.guess_type(str(target))[0] or 'application/octet-stream') | |
| def do_POST(self): | |
| if urlparse(self.path).path!='/api/evaluate': | |
| self.send_json(404,{'error':'Unknown endpoint'});return | |
| try: | |
| length=int(self.headers.get('Content-Length','0')) | |
| if not 0 < length <= 2_000_000:raise ValueError('Request must contain 1..2000000 bytes') | |
| origin=self.headers.get('Origin') | |
| if origin and urlparse(origin).netloc != self.headers.get('Host'):raise ValueError('Cross-origin requests are disabled') | |
| from predict_toy_decisions import unique_object,reject_nonfinite,validate_request | |
| payload=json.loads(self.rfile.read(length),object_pairs_hook=unique_object,parse_constant=reject_nonfinite) | |
| states=validate_request(payload) | |
| questions=[q for s in states for q in s['questions'].values()] | |
| paths=sum(1 if q['type']=='boolean' else len(q['criteria']) for q in questions) | |
| if len(states)>32 or len(questions)>96 or paths>256:raise ValueError('Local demo limit:32 states,96 questions,256 candidate paths per request') | |
| before=time.perf_counter();result=engine.predict(payload) | |
| result['execution']['server_evaluation_seconds']=time.perf_counter()-before | |
| self.send_json(200,result) | |
| except (ValueError,TypeError,KeyError) as exc: | |
| self.send_json(400,{'error':str(exc)}) | |
| except Exception: | |
| self.send_json(500,{'error':'Local model inference failed; inspect the server process. No teacher fallback was used.'}) | |
| raise | |
| def log_message(self, fmt, *args): | |
| # HTTP method/path/status only; request states and credentials are not logged. | |
| print(fmt % args,flush=True) | |
| return Handler | |
| if __name__=='__main__': | |
| p=argparse.ArgumentParser(description=__doc__);p.add_argument('--checkpoint-dir',required=True) | |
| p.add_argument('--web-root',default='web');p.add_argument('--host',default='127.0.0.1');p.add_argument('--port',type=int,default=8765) | |
| p.add_argument('--precision',choices=['fp32','bf16'],default='bf16');p.add_argument('--disable-native-triton',action='store_true') | |
| a=p.parse_args() | |
| from predict_toy_decisions import DecisionPredictor | |
| engine=DecisionPredictor(a.checkpoint_dir,precision=a.precision,disable_native_triton=a.disable_native_triton) | |
| root=Path(a.web_root).resolve() | |
| if not (root/'index.html').is_file():raise ValueError('web-root must contain index.html') | |
| server=HTTPServer((a.host,a.port),server_class(engine,root)) | |
| print(json.dumps({'url':f'http://{a.host}:{a.port}','ready':True,'provider_calls':0}),flush=True) | |
| try:server.serve_forever() | |
| except KeyboardInterrupt:pass | |
| finally:server.server_close() | |
Xet Storage Details
- Size:
- 4.23 kB
- Xet hash:
- 40fe4549fd09692552e3429a0a86033495731d7c833856405b91f7b9253e1c68
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.