18 lines
1.1 KiB
Bash
18 lines
1.1 KiB
Bash
|
#!/bin/bash
|
|||
|
|
# dengyuyun-joy: quick functional check against the trial API.
|
|||
|
|
U=http://127.0.0.1:18100
|
|||
|
|
curl -s -m 10 $U/v1/models | python3 -c "import json,sys;d=json.load(sys.stdin);print('models:',[(m['id'],m.get('max_model_len')) for m in d['data']])"
|
|||
|
|
for q in "用一句话介绍你自己。" "计算 123456 × 789，只给出结果。" "Write a Python function that returns the n-th Fibonacci number, then explain it in one sentence."; do
|
|||
|
|
python3 - "$q" <<'PY'
|
|||
|
|
import json,sys,time,urllib.request
|
|||
|
|
q=sys.argv[1]
|
|||
|
|
body={"model":"deepseek-v4.1-flash","messages":[{"role":"user","content":q}],"temperature":0,"max_tokens":512}
|
|||
|
|
t=time.time()
|
|||
|
|
r=urllib.request.urlopen(urllib.request.Request("http://127.0.0.1:18100/v1/chat/completions",json.dumps(body).encode(),{"Content-Type":"application/json"}),timeout=600)
|
|||
|
|
d=json.load(r); dt=time.time()-t
|
|||
|
|
m=d["choices"][0]["message"]; u=d.get("usage",{})
|
|||
|
|
print("Q:",q); print("A:",(m.get("content") or "").strip()[:600]); print("reasoning chars:",len(m.get("reasoning_content") or m.get("reasoning") or ""))
|
|||
|
|
print(f"usage={u} wall={dt:.1f}s ~{u.get('completion_tokens',0)/dt:.1f} tok/s"); print("-"*60)
|
|||
|
|
PY
|
|||
|
|
done
|