Devstral Small 2: serve.sh, baseline.py without thinking args, repair rate, activation error message fix, stop rule in chain (run base 21000)

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014aUaQeLnwbb1zTpN7kHeat
This commit is contained in:
Kral
2026-10-04 08:57:12 +02:00
parent e84ea43a3d
commit 8d1db9c67c
5 changed files with 56 additions and 21 deletions

View File

@@ -13,9 +13,9 @@ ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
sys.path.insert(0, ROOT)
from harness.adt_client import load_env # noqa: E402
from harness.agents import LlmAgent # noqa: E402
from harness.runner import Runner # noqa: E402
from harness.runner import Runner, repair_stats # noqa: E402
MODEL = os.path.expanduser("~/models/Qwen3.8-27B-4bit")
MODEL = os.path.expanduser("~/models/Devstral-Small-2-24B-4bit")
BASE_URL = "http://127.0.0.1:8080/v1"
LOOP_GUARD = 3 # end the run after 3 identical pushes in a row (end_reason "loop"); same for after training
MAX_TOKENS = 16384 # per turn; sent in each request (the server limit stays 32768). Same for before/after.
@@ -32,7 +32,7 @@ def main():
out_path = os.path.join(ROOT, "runs", "stage1", f"{a.label}.json")
os.makedirs(os.path.dirname(out_path), exist_ok=True)
res = json.load(open(out_path)) if os.path.exists(out_path) else {"model": MODEL, "tasks": {}}
res["settings"] = {"max_tokens": MAX_TOKENS, "enable_thinking": False, "temperature": 0.2, "tool_call_budget": 60, "loop_guard": LOOP_GUARD,
res["settings"] = {"max_tokens": MAX_TOKENS, "enable_thinking": None, "temperature": 0.2, "tool_call_budget": 60, "loop_guard": LOOP_GUARD,
"subset": "train/subset.json", "date": "2026-10-04"}
runs_root = os.path.join(ROOT, "runs", "stage1", a.label)
for i, t in enumerate(subset):
@@ -40,8 +40,7 @@ def main():
if (a.only and tid not in a.only) or tid in res["tasks"]:
continue
pool = os.path.join(ROOT, "tasks") if tid.startswith("T") else os.path.join(ROOT, "tasks_gen", "eval")
agent = LlmAgent(MODEL, BASE_URL, max_tokens=MAX_TOKENS, chat_template_kwargs={"enable_thinking": False},
loop_guard=LOOP_GUARD)
agent = LlmAgent(MODEL, BASE_URL, max_tokens=MAX_TOKENS, loop_guard=LOOP_GUARD)
try:
rep, run_dir = Runner(pool, runs_root).run(tid, agent, a.run_base + i)
except Exception as e: # noqa: BLE001 one broken run must not stop the series
@@ -56,6 +55,7 @@ def main():
"tool_calls": rep.get("tool_calls"), "seconds": rep.get("seconds"),
"end_reason": rep.get("end_reason"), "activation_failures": rep.get("activation_failures"),
"activation_error_messages": rep.get("activation_error_messages"),
**repair_stats(os.path.join(run_dir, "trajectory.jsonl")),
"agent_seconds": rep.get("agent_seconds"), "final": (rep.get("final_report") or "")[:300],
"run_dir": os.path.relpath(run_dir, ROOT)}
json.dump(res, open(out_path, "w"), indent=1)