Add reproducible DGX Spark deployment for Qwen3.8 Flash Next NVFP4

2026-09-17 16:45:50 +08:00
commit 5f3030260e
27 changed files with 1705 additions and 0 deletions
@@ -0,0 +1,42 @@
+import json
+import time
+import urllib.request
+from pathlib import Path
+
+key = Path('/run/secrets/qwen_api_key').read_text().strip()
+base = 'http://127.0.0.1:8000'
+headers = {'Authorization': 'Bearer ' + key, 'Content-Type': 'application/json'}
+
+def request(path, payload=None):
+    data = None if payload is None else json.dumps(payload).encode()
+    with urllib.request.urlopen(urllib.request.Request(base + path, data=data, headers=headers), timeout=900) as response:
+        return json.load(response)
+
+print(json.dumps({'models': request('/v1/models')}, ensure_ascii=False), flush=True)
+cases = [
+    ('arithmetic', '计算 17 × 19，只给出数字。', '323'),
+    ('chinese', '请用中文简短说明为什么天空是蓝色的。', None),
+]
+for name, prompt, expected in cases:
+    start = time.monotonic()
+    result = request('/v1/chat/completions', {
+        'model': 'qwen3.8-flash-next',
+        'messages': [{'role': 'user', 'content': prompt}],
+        'temperature': 0, 'max_tokens': 1024, 'reasoning_effort': 'low',
+    })
+    choice = result['choices'][0]
+    content = choice['message'].get('content') or ''
+    record = {'test': name, 'seconds': round(time.monotonic()-start, 2),
+        'content': content, 'finish_reason': choice['finish_reason'], 'usage': result.get('usage')}
+    print(json.dumps(record, ensure_ascii=False), flush=True)
+    assert content.strip(), 'No final answer'
+    if expected:
+        assert expected in content, 'Arithmetic mismatch'
+
+
+p={'model':'qwen3.8-flash-next','messages':[{'role':'user','content':'请调用 get_weather 查询上海的天气，不要自行编造。'}],'tools':[{'type':'function','function':{'name':'get_weather','description':'查询城市天气','parameters':{'type':'object','properties':{'city':{'type':'string'}},'required':['city']}}}],'tool_choice':'auto','temperature':0,'max_tokens':512,'reasoning_effort':'low'}
+t=time.monotonic();r=request('/v1/chat/completions', p);m=r['choices'][0]['message'];print(json.dumps({'test':'tool_call','seconds':time.monotonic()-t,'message':m},ensure_ascii=False),flush=True)
+assert m.get('tool_calls') and m['tool_calls'][0]['function']['name']=='get_weather'
+print('TOOL_TEST_PASS',flush=True)
+
+print("SMOKE_TEST_PASS", flush=True)