把大模型后训练变成直播：从小米 MiMo 2.6 看实时可观测性

from __future__ import annotations

import json
import random
import threading
import time
from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer

STATE = {
    "model": "demo-pro",
    "step": 0,
    "generated_samples": 0,
    "accepted_samples": 0,
    "tokens": 0,
    "estimated_cost_usd": 0.0,
    "updated_at": None,
}
LOCK = threading.Lock()


def training_loop() -> None:
    while True:
        generated = random.randint(32, 64)
        accepted = random.randint(16, generated)
        tokens = random.randint(40_000, 90_000)

        with LOCK:
            STATE["step"] += 1
            STATE["generated_samples"] += generated
            STATE["accepted_samples"] += accepted
            STATE["tokens"] += tokens
            # Replace this estimate with your actual billing or cluster accounting data.
            STATE["estimated_cost_usd"] += tokens / 1_000_000 * 2.5
            STATE["updated_at"] = time.strftime("%Y-%m-%dT%H:%M:%SZ", time.gmtime())

        time.sleep(2)


class Handler(BaseHTTPRequestHandler):
    def do_GET(self) -> None:
        if self.path != "/metrics/latest":
            self.send_error(404)
            return

        with LOCK:
            payload = dict(STATE)
            generated = payload["generated_samples"]
            payload["acceptance_rate"] = (
                payload["accepted_samples"] / generated if generated else 0.0
            )

        body = json.dumps(payload, ensure_ascii=False).encode("utf-8")
        self.send_response(200)
        self.send_header("Content-Type", "application/json; charset=utf-8")
        self.send_header("Content-Length", str(len(body)))
        self.end_headers()
        self.wfile.write(body)


if __name__ == "__main__":
    threading.Thread(target=training_loop, daemon=True).start()
    print("Metrics available at http://127.0.0.1:8000/metrics/latest")
    ThreadingHTTPServer(("127.0.0.1", 8000), Handler).serve_forever()

python telemetry_demo.py
curl -s http://127.0.0.1:8000/metrics/latest | python -m json.tool