Spaces:
Sleeping
Sleeping
File size: 6,252 Bytes
5db30ac | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 | import asyncio
import sys
import unittest
from pathlib import Path
import httpx
ROOT = Path(__file__).resolve().parents[1]
SRC = ROOT / "src"
if str(SRC) not in sys.path:
sys.path.insert(0, str(SRC))
from elbench.providers.openai_compatible import OpenAICompatibleClient # noqa: E402
from elbench.schemas.config import ModelConfig, ProviderConfig # noqa: E402
from elbench.schemas.evaluation import GenerationRequest # noqa: E402
class OpenAICompatibleStreamingTest(unittest.TestCase):
def test_streaming_chat_completion_accumulates_reasoning_and_content(self) -> None:
async def handler(request: httpx.Request) -> httpx.Response:
self.assertEqual(request.url.path, "/v1/chat/completions")
body = b"".join(
[
b'data: {"id":"stream-1","object":"chat.completion.chunk","model":"deepseek-r1",'
b'"choices":[{"index":0,"delta":{"role":"assistant","content":"","reasoning_content":"think "},'
b'"finish_reason":null}]}\n\n',
b'data: {"id":"stream-1","object":"chat.completion.chunk","model":"deepseek-r1",'
b'"choices":[{"index":0,"delta":{"content":"","reasoning_content":"steps"},'
b'"finish_reason":null}]}\n\n',
b'data: {"id":"stream-1","object":"chat.completion.chunk","model":"deepseek-r1",'
b'"choices":[{"index":0,"delta":{"content":"Final answer is \\\\boxed{204}"},'
b'"finish_reason":null}]}\n\n',
b'data: {"id":"stream-1","object":"chat.completion.chunk","model":"deepseek-r1",'
b'"usage":{"prompt_tokens":10,"completion_tokens":20,'
b'"completion_tokens_details":{"reasoning_tokens":12}},'
b'"choices":[{"index":0,"delta":{"content":""},"finish_reason":"stop"}]}\n\n',
b"data: [DONE]\n\n",
]
)
return httpx.Response(200, headers={"Content-Type": "text/event-stream"}, content=body)
client = self._client(handler)
try:
response = asyncio.run(
client.generate(
sample=None,
request=GenerationRequest(
prompt="Solve.",
messages=[{"role": "user", "content": "Solve."}],
max_tokens=8192,
stream=True,
),
)
)
finally:
asyncio.run(client.aclose())
self.assertEqual(response.status_code, 200)
self.assertEqual(response.text, "Final answer is \\boxed{204}")
self.assertEqual(response.usage["completion_tokens_details"]["reasoning_tokens"], 12)
self.assertEqual(
response.raw_payload["choices"][0]["message"]["reasoning_content"],
"think steps",
)
self.assertEqual(response.raw_payload["choices"][0]["finish_reason"], "stop")
self.assertTrue(response.raw_payload["stream"])
def test_streaming_chat_completion_requires_done_marker(self) -> None:
async def handler(request: httpx.Request) -> httpx.Response:
body = (
b'data: {"id":"stream-1","object":"chat.completion.chunk","model":"deepseek-r1",'
b'"choices":[{"index":0,"delta":{"content":"partial"},"finish_reason":null}]}\n\n'
)
return httpx.Response(200, headers={"Content-Type": "text/event-stream"}, content=body)
client = self._client(handler)
try:
with self.assertRaises(RuntimeError):
asyncio.run(
client.generate(
sample=None,
request=GenerationRequest(
prompt="Solve.",
messages=[{"role": "user", "content": "Solve."}],
max_tokens=8192,
stream=True,
),
)
)
finally:
asyncio.run(client.aclose())
def test_streaming_chat_completion_respects_total_timeout(self) -> None:
class SlowSseStream(httpx.AsyncByteStream):
async def __aiter__(self):
yield (
b'data: {"id":"stream-1","object":"chat.completion.chunk","model":"deepseek-r1",'
b'"choices":[{"index":0,"delta":{"reasoning_content":"still thinking"},'
b'"finish_reason":null}]}\n\n'
)
await asyncio.sleep(0.05)
yield b"data: [DONE]\n\n"
async def handler(request: httpx.Request) -> httpx.Response:
return httpx.Response(
200,
headers={"Content-Type": "text/event-stream"},
stream=SlowSseStream(),
)
client = self._client(handler, timeout=0.01)
try:
with self.assertRaises(httpx.TimeoutException):
asyncio.run(
client.generate(
sample=None,
request=GenerationRequest(
prompt="Solve.",
messages=[{"role": "user", "content": "Solve."}],
max_tokens=8192,
stream=True,
),
)
)
finally:
asyncio.run(client.aclose())
def _client(self, handler, *, timeout: float | None = None) -> OpenAICompatibleClient:
client = OpenAICompatibleClient(
ProviderConfig(provider_name="openai_compatible", adapter="openai_compatible"),
ModelConfig(
model_id="deepseek-r1-250528",
provider_name="openai_compatible",
model_name="deepseek-r1-250528",
api_base="https://api.innospark.cn/v1",
api_key_env=None,
timeout=timeout,
),
)
client._client = httpx.AsyncClient(transport=httpx.MockTransport(handler))
return client
if __name__ == "__main__":
unittest.main()
|