apps/FastDeploy
1
0
Fork 0
mirror of https://github.com/PaddlePaddle/FastDeploy.git synced 2025-10-07 17:41:52 +08:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

【Bugfix】修复2.1分支上0.3B模型性能大幅下降 (#3624)

Browse Source 
* 恢复异步方法。
【BugFix】completion接口echo回显支持 (#3245)

* wenxin-tools-511,修复v1/completion无法回显的问题。

* 支持多prompt的回显

* 支持多prompt情况下的流式回显

* 补充了 completion 接口支持 echo 的单元测试

* pre-commit

* 移除了多余的test文件

* 修复了completion接口echo支持的单测方法

* 补充了单元测试文件

* 补充单测

* unittest

* 补充单测

* 修复单测

* 删除不必要的assert.

* 重新提交

* 更新测试方法

* ut

* 验证是否是正确思路单测

* 验证是否是正确思路单测

* 验证是否是正确思路单测3

* 优化单测代码，有针对性地缩小单测范围。

* 优化单测代码2，有针对性地缩小单测范围。

* 优化单测代码3，有针对性地缩小单测范围。

* support 'echo' in chat/completion.

* update

* update

* update

* update

* update

* update

* 补充了关于tokenid的单元测试

* update

* 修正index错误

* 修正index错误

* [Bugfix] Significant performance degradation of 0.3B model on branch 2.1
This commit is contained in:
zhuzixuan
2025-08-27 15:29:01 +08:00
committed by GitHub
parent 96aed92e4a
commit 80db7fce05
2 changed files with 26 additions and 21 deletions
Download Patch File Download Diff File Expand all files Collapse all files

42
test/entrypoints/openai/test_completion_echo.py
View File

@@ -7,13 +7,22 @@ from fastdeploy.entrypoints.openai.serving_completion import (
)
class TestCompletionEcho(unittest.TestCase):
class YourClass:
async def _1(self, a, b, c):
if b["outputs"].get("send_idx", -1) == 0 and a.echo:
if isinstance(a.prompt, list):
text = a.prompt[c]
else:
text = a.prompt
b["outputs"]["text"] = text + (b["outputs"]["text"] or "")
class TestCompletionEcho(unittest.IsolatedAsyncioTestCase):
def setUp(self):
self.mock_engine = MagicMock()
self.completion_handler = None
def test_single_prompt_non_streaming(self):
"""测试单prompt非流式响应"""
self.completion_handler = OpenAIServingCompletion(self.mock_engine, pid=123, ips=None, max_waiting_time=30)
request = CompletionRequest(prompt="test prompt", max_tokens=10, echo=True, logprobs=1)
@@ -42,8 +51,7 @@ class TestCompletionEcho(unittest.TestCase):
self.assertEqual(response.choices[0].text, "test prompt generated text")
def test_echo_back_prompt_and_streaming(self):
"""测试_echo_back_prompt方法和流式响应的prompt拼接逻辑"""
async def test_echo_back_prompt_and_streaming(self):
self.completion_handler = OpenAIServingCompletion(self.mock_engine, pid=123, ips=None, max_waiting_time=30)
request = CompletionRequest(prompt="test prompt", max_tokens=10, stream=True, echo=True)
@@ -57,7 +65,7 @@ class TestCompletionEcho(unittest.TestCase):
mock_echo.side_effect = mock_echo_side_effect
self.completion_handler._echo_back_prompt(request, mock_response, 0)
await self.completion_handler._echo_back_prompt(request, mock_response, 0)
mock_echo.assert_called_once_with(request, mock_response, 0)
@@ -65,7 +73,6 @@ class TestCompletionEcho(unittest.TestCase):
self.assertEqual(request.prompt, "test prompt")
def test_multi_prompt_non_streaming(self):
"""测试多prompt非流式响应"""
self.completion_handler = OpenAIServingCompletion(self.mock_engine, pid=123, ips=None, max_waiting_time=30)
request = CompletionRequest(prompt=["prompt1", "prompt2"], max_tokens=10, echo=True)
@@ -97,7 +104,7 @@ class TestCompletionEcho(unittest.TestCase):
self.assertEqual(response.choices[0].text, "prompt1 response1")
self.assertEqual(response.choices[1].text, "prompt2 response2")
def test_multi_prompt_streaming(self):
async def test_multi_prompt_streaming(self):
self.completion_handler = OpenAIServingCompletion(self.mock_engine, pid=123, ips=None, max_waiting_time=30)
request = CompletionRequest(prompt=["prompt1", "prompt2"], max_tokens=10, stream=True, echo=True)
@@ -114,8 +121,8 @@ class TestCompletionEcho(unittest.TestCase):
mock_echo.side_effect = mock_echo_side_effect
self.completion_handler._echo_back_prompt(request, mock_responses[0], 0)
self.completion_handler._echo_back_prompt(request, mock_responses[1], 1)
await self.completion_handler._echo_back_prompt(request, mock_responses[0], 0)
await self.completion_handler._echo_back_prompt(request, mock_responses[1], 1)
self.assertEqual(mock_echo.call_count, 2)
mock_echo.assert_any_call(request, mock_responses[0], 0)
@@ -125,41 +132,40 @@ class TestCompletionEcho(unittest.TestCase):
self.assertEqual(mock_responses[1]["outputs"]["text"], "prompt2 response2")
self.assertEqual(request.prompt, ["prompt1", "prompt2"])
def test_echo_back_prompt_and_streaming1(self):
async def test_echo_back_prompt_and_streaming1(self):
request = CompletionRequest(echo=True, prompt=["Hello", "World"])
res = {"outputs": {"send_idx": 0, "text": "!"}}
idx = 0
instance = OpenAIServingCompletion(self.mock_engine, pid=123, ips=None, max_waiting_time=30)
instance._echo_back_prompt(request, res, idx)
await instance._echo_back_prompt(request, res, idx)
self.assertEqual(res["outputs"]["text"], "Hello!")
def test_1_prompt_is_string_and_send_idx_is_0(self):
async def test_1_prompt_is_string_and_send_idx_is_0(self):
request = CompletionRequest(echo=True, prompt="Hello")
res = {"outputs": {"send_idx": 0, "text": "!"}}
idx = 0
instance = OpenAIServingCompletion(self.mock_engine, pid=123, ips=None, max_waiting_time=30)
instance._echo_back_prompt(request, res, idx)
await instance._echo_back_prompt(request, res, idx)
self.assertEqual(res["outputs"]["text"], "Hello!")
def test_1_send_idx_is_not_0(self):
async def test_1_send_idx_is_not_0(self):
request = CompletionRequest(echo=True, prompt="Hello")
res = {"outputs": {"send_idx": 1, "text": "!"}}
idx = 0
instance = OpenAIServingCompletion(self.mock_engine, pid=123, ips=None, max_waiting_time=30)
instance._echo_back_prompt(request, res, idx)
await instance._echo_back_prompt(request, res, idx)
self.assertEqual(res["outputs"]["text"], "!")
def test_1_echo_is_false(self):
"""测试echo为False时，_echo_back_prompt不拼接prompt"""
async def test_1_echo_is_false(self):
request = CompletionRequest(echo=False, prompt="Hello")
res = {"outputs": {"send_idx": 0, "text": "!"}}
idx = 0
instance = OpenAIServingCompletion(self.mock_engine, pid=123, ips=None, max_waiting_time=30)
instance._echo_back_prompt(request, res, idx)
await instance._echo_back_prompt(request, res, idx)
self.assertEqual(res["outputs"]["text"], "!")