ruoxi_sun
58657935fc
Vendor LLMmap / llm-verify / llm-fingerprint-detector under bash/fingerprint/tools so the three fingerprint benchmarks run with only /data1/eval mounted (no /data1/xii dependency): - run.py DEFAULT_TOOLS_ROOT prefers builtin tools/, falls back to /data1/xii - exclude .git / node_modules / template backups - detector dist/ (pre-built) retained; node_modules not needed at runtime
32 lines
861 B
Python
32 lines
861 B
Python
"""Quick script to view benchmark results."""
|
|
|
|
import httpx
|
|
import json
|
|
import sys
|
|
|
|
RUN_ID = sys.argv[1] if len(sys.argv) > 1 else "5f633956-040d-40e6-bd17-3b47be57f9a8"
|
|
BASE = "http://127.0.0.1:8001"
|
|
|
|
r = httpx.get(f"{BASE}/api/v1/results/{RUN_ID}")
|
|
results = r.json()
|
|
|
|
for i, res in enumerate(results, 1):
|
|
sep = "=" * 80
|
|
prompt = res["prompt_text"][:120]
|
|
response = res["response_text"][:600]
|
|
latency = res["latency_ms"] or 0
|
|
in_tok = res["prompt_tokens"]
|
|
out_tok = res["completion_tokens"]
|
|
err = res["error_message"]
|
|
|
|
print(f"\n{sep}")
|
|
print(f"PROBE #{i}")
|
|
print(f"PROMPT: {prompt}...")
|
|
print(f"RESPONSE:\n{response}")
|
|
print(f"\nLATENCY: {latency:.0f}ms | IN: {in_tok} | OUT: {out_tok}")
|
|
if err:
|
|
print(f"ERROR: {err[:200]}")
|
|
|
|
print(f"\n{'=' * 80}")
|
|
print(f"TOTAL: {len(results)} probes completed")
|