[CI] 【Hackathon 9th Sprint No.24】NO.24 功能模块单测补充 #5055

xunyoyo · 2025-11-15T10:16:30Z

Motivation

NO.24 功能模块 fastdeploy/model_executor/models/tp_utils.py 单测补充

Modifications

add tests/model_executor/models/test_tp_utils.py

Usage or Command

tests/model_executor/test_tp_utils.py:

python -m coverage run -m unittest tests.model_executor.test_tp_utils \
&& python -m coverage report -m --include='fastdeploy/model_executor/models/tp_utils.py'

Accuracy Tests

旧覆盖

File	Stmts	Miss	Branch	BrPart	Cover(%)
fastdeploy/model_executor/models/tp_utils.py	201	170	106	0	10

tests/model_executor/models/test_tp_utils.py:

Name                                           Stmts   Miss  Cover   Missing
----------------------------------------------------------------------------
fastdeploy/model_executor/models/tp_utils.py     201     27    87%   130, 143, 160, 177, 191, 228, 248, 256-259, 278, 283, 297-3
16, 353-356, 376-382, 387, 440
----------------------------------------------------------------------------
TOTAL                                            201     27    87%

覆盖 ++143 行

Checklist

Add at least a tag in the PR title.
- Tag list: [[FDConfig],[APIServer],[Engine], [Scheduler], [PD Disaggregation], [Executor], [Graph Optimization], [Speculative Decoding], [RL], [Models], [Quantization], [Loader], [OP], [KVCache], [DataProcessor], [BugFix], [Docs], [CI], [Optimization], [Feature], [Benchmark], [Others], [XPU], [HPU], [GCU], [DCU], [Iluvatar], [Metax]]
- You can add new tags based on the PR content, but the semantics must be clear.
Format your code, run pre-commit before commit.
Add unit tests. Please write the reason in this PR if no unit tests.
Provide accuracy results.
If the current PR is submitting to the release branch, make sure the PR has been submitted to the develop branch, then cherry-pick it to the release branch with the [Cherry-Pick] PR tag.

paddle-bot · 2025-11-15T10:16:37Z

Thanks for your contribution!

Copilot

Pull Request Overview

This PR adds comprehensive unit tests for the fastdeploy/model_executor/models/tp_utils.py module as part of Hackathon 9th Sprint No.24, achieving 87% test coverage. The test suite uses sophisticated stubbing techniques to test the tensor parallel utility functions in isolation without requiring the full PaddlePaddle and PaddleFormers dependencies.

Key changes:

Implements extensive mocking infrastructure for paddle, paddleformers, and fastdeploy dependencies
Tests core tensor parallel prerequisite checking, helper functions, key expansion logic, and GQA tensor operations
Achieves 87% code coverage with 201 statements tested (27 lines remain uncovered)

Copilot · 2025-11-15T10:19:12Z

tests/model_executor/test_tp_utils.py

+"""Unit tests for tensor parallel utility helpers."""
+
+from __future__ import annotations


The test file is missing the standard Apache 2.0 copyright header that is present in other test files in this repository (e.g., test_logits_processor.py and test_forward_meta_str.py). Please add the copyright header at the beginning of the file following the project's standard format.

Copilot · 2025-11-15T10:19:13Z

tests/model_executor/test_tp_utils.py

+    def _split_or_merge_func(is_split, tensor_parallel_degree, tensor_parallel_rank, **_kwargs):
+        axis = -1
+
+        def _fn(weight, *, is_column=True, is_naive_2fuse=False):  # pylint: disable=unused-argument


[nitpick] The pylint disable comment uses an inline format. For consistency with project standards and better maintainability, consider using the more specific format or removing unused parameters if they're truly not needed in the stub implementation.

Suggested change

def _fn(weight, *, is_column=True, is_naive_2fuse=False): # pylint: disable=unused-argument

def _fn(weight, *, is_column=True):

Copilot · 2025-11-15T10:19:13Z

tests/model_executor/test_tp_utils.py

+        np.testing.assert_array_equal(merged, np.array([0, 1, 2, 3], dtype=np.float32))
+
+
+if __name__ == "__main__":  # pragma: no cover - entry point for python -m unittest


[nitpick] The pragma comment format is inconsistent. Most test files in this codebase don't use pragma comments for the main block. Consider removing the pragma comment or ensuring it follows the project's coverage reporting conventions.

Suggested change

if __name__ == "__main__": # pragma: no cover - entry point for python -m unittest

if __name__ == "__main__":

xunyoyo · 2025-11-17T12:07:59Z

@luotao1 @YuanRisheng
同样在 pr action 检查过程中 api 调用出现了问题，导致报错停止。

Running pytest file: tests/e2e/test_ernie_03b_pd_router_v0.py
============================= test session starts ==============================
platform linux -- Python 3.10.13, pytest-9.0.1, pluggy-1.6.0 -- /usr/local/bin/python
cachedir: .pytest_cache
rootdir: /workspace/FastDeploy/tests
configfile: cov_pytest.ini
plugins: asyncio-1.3.0, respx-0.22.0, anyio-4.9.0
asyncio: mode=strict, debug=False, asyncio_default_fixture_loop_scope=None, asyncio_default_test_loop_scope=function
collecting ... collected 5 items

tests/e2e/test_ernie_03b_pd_router_v0.py::test_metrics_config Pre-test port cleanup...
Cleaning ports: [8688, 8658, 8678, 8698, 8638, 8689, 8659, 8679, 8699, 8639, 8648]
log dir clean 
model_path: /ModelData/ERNIE-4.5-0.3B-Paddle
start router...
start prefill...
start decode...
Prefill and decode servers are both online
PASSED
tests/e2e/test_ernie_03b_pd_router_v0.py::test_chat_usage_stream 🟢 接收响应中...

Decode Response: 牛顿的三大运动定律是牛顿力学的基本原理，它们是描述物体运动和力的关系的基础。以下是这三条定律的详细解释：

### 一、牛顿第一定律（惯性定律）

**定义**：牛顿第一
PASSED
tests/e2e/test_ernie_03b_pd_router_v0.py::test_chat_usage_non_stream 🟢 接收响应中...

PASSED
tests/e2e/test_ernie_03b_pd_router_v0.py::test_non_chat_usage_stream 🟢 接收响应中...

Decode Response: 

### 解答

牛顿的三大运动定律是：

1. **惯性定律**：物体保持静止或匀速直线运动状态的性质，即**惯性**。
2. **万有引力定律**：
PASSED
tests/e2e/test_ernie_03b_pd_router_v0.py::test_non_chat_usage_non_stream 🟢 接收响应中...

Decode Response: 

### 解答

牛顿的三大运动定律是：

1. **惯性定律**：物体保持静止或匀速直线运动状态的性质，即**惯性**。
2. **万有引力定律**：
PASSED
===== Post-test server cleanup... =====
Cleaning ports: [8688, 8658, 8678, 8698, 8638, 8689, 8659, 8679, 8699, 8639, 8648]
Killed process on port 8658, pid=8672
Killed process on port 8659, pid=8811
Prefill server (pid=8672) terminated
Decode server (pid=8811) terminated


/usr/local/lib/python3.10/site-packages/coverage/control.py:950: CoverageWarning: No data was collected. (no-data-collected); see https://coverage.readthedocs.io/en/7.11.3/messages.html#warning-no-data-collected
  self._warn("No data was collected.", slug="no-data-collected")
============================== 5 passed in 46.41s ==============================
Running pytest file: tests/e2e/test_ernie_03b_pd_splitwise_scheduler.py
============================= test session starts ==============================
platform linux -- Python 3.10.13, pytest-9.0.1, pluggy-1.6.0 -- /usr/local/bin/python
cachedir: .pytest_cache
rootdir: /workspace/FastDeploy/tests
configfile: cov_pytest.ini
plugins: asyncio-1.3.0, respx-0.22.0, anyio-4.9.0
asyncio: mode=strict, debug=False, asyncio_default_fixture_loop_scope=None, asyncio_default_test_loop_scope=function
collecting ... collected 5 items

tests/e2e/test_ernie_03b_pd_splitwise_scheduler.py::test_metrics_config Pre-test port cleanup...
Cleaning ports: [8688, 8658, 8678, 8698, 8638, 8689, 8659, 8679, 8699, 8639, 8533]
log dir clean 
model_path: /ModelData/ERNIE-4.5-0.3B-Paddle
start redis...
start prefill...
start decode...
Prefill server is up on port 8688
Decode server is up on port 8689
PASSED
tests/e2e/test_ernie_03b_pd_splitwise_scheduler.py::test_chat_usage_stream 🟢 接收响应中...

Decode Response: 牛顿的三大运动定律是牛顿力学的基本原理，它们是描述物体运动和力的关系的基础。以下是这三条定律的详细解释：

### 一、牛顿第一定律（惯性定律）

**定义**：牛顿第一
PASSED
tests/e2e/test_ernie_03b_pd_splitwise_scheduler.py::test_chat_usage_non_stream ❌ 请求超时（超过 600 秒）
FAILED
tests/e2e/test_ernie_03b_pd_splitwise_scheduler.py::test_non_chat_usage_stream 🟢 接收响应中...

PASSED
tests/e2e/test_ernie_03b_pd_splitwise_scheduler.py::test_non_chat_usage_non_stream 🟢 接收响应中...

PASSED
===== Post-test server cleanup... =====
Cleaning ports: [8688, 8658, 8678, 8698, 8638, 8689, 8659, 8679, 8699, 8639, 8533]
Killed process on port 8658, pid=10328
Killed process on port 8658, pid=10404
Prefill server (pid=10328) terminated
Decode server (pid=10404) terminated


=================================== FAILURES ===================================
__________________________ test_chat_usage_non_stream __________________________

api_url = ('http://0.0.0.0:8688/v1/chat/completions', 'http://0.0.0.0:8689/v1/chat/completions')

    def test_chat_usage_non_stream(api_url):
        """测试非流式chat usage"""
        payload = {
            "model": "default",
            "temperature": 0,
            "top_p": 0,
            "seed": 33,
            "messages": [
                {"role": "system", "content": "You are a helpful assistant."},
                {"role": "user", "content": "牛顿的三大运动定律是什么？"},
            ],
            "max_tokens": 50,
            "stream": False,
            "metadata": {"min_tokens": 10},
        }
    
        p_url, d_url = api_url
>       response = send_request(url=p_url, payload=payload).json()
E       AttributeError: 'NoneType' object has no attribute 'json'

tests/e2e/test_ernie_03b_pd_splitwise_scheduler.py:366: AttributeError

我的已通过

Running pytest file: tests/model_executor/test_tp_utils.py
============================= test session starts ==============================
platform linux -- Python 3.10.13, pytest-9.0.1, pluggy-1.6.0 -- /usr/local/bin/python
cachedir: .pytest_cache
rootdir: /workspace/FastDeploy/tests
configfile: cov_pytest.ini
plugins: asyncio-1.3.0, respx-0.22.0, anyio-4.9.0
asyncio: mode=strict, debug=False, asyncio_default_fixture_loop_scope=None, asyncio_default_test_loop_scope=function
collecting ... collected 26 items

tests/model_executor/test_tp_utils.py::CheckTensorParallelPrerequisitesTest::test_inconsistent_tensor_parallel_keys_logs_error PASSED
tests/model_executor/test_tp_utils.py::CheckTensorParallelPrerequisitesTest::test_missing_tensor_parallel_map_logs_error PASSED
tests/model_executor/test_tp_utils.py::CheckTensorParallelPrerequisitesTest::test_tensor_parallel_disabled_noop PASSED
tests/model_executor/test_tp_utils.py::CheckTensorParallelPrerequisitesTest::test_tensor_parallel_mappings_populated PASSED
tests/model_executor/test_tp_utils.py::HelperFunctionTest::test_extract_placeholders PASSED
tests/model_executor/test_tp_utils.py::HelperFunctionTest::test_extract_prefix_variants PASSED
tests/model_executor/test_tp_utils.py::HelperFunctionTest::test_has_placeholders PASSED
tests/model_executor/test_tp_utils.py::HelperFunctionTest::test_has_prefix PASSED
tests/model_executor/test_tp_utils.py::HelperFunctionTest::test_safe_dict_preserves_unknown PASSED
tests/model_executor/test_tp_utils.py::HelperFunctionTest::test_update_final_actions_formats_keys PASSED
tests/model_executor/test_tp_utils.py::BuildExpandedKeysTest::test_ffn_layer_id_requires_start PASSED
tests/model_executor/test_tp_utils.py::BuildExpandedKeysTest::test_invalid_placeholder_raises PASSED
tests/model_executor/test_tp_utils.py::BuildExpandedKeysTest::test_layer_id_placeholder PASSED
tests/model_executor/test_tp_utils.py::BuildExpandedKeysTest::test_moe_layer_and_expert_id PASSED
tests/model_executor/test_tp_utils.py::BuildExpandedKeysTest::test_moe_layer_and_image_expert_id PASSED
tests/model_executor/test_tp_utils.py::BuildExpandedKeysTest::test_moe_layer_and_text_expert_id PASSED
tests/model_executor/test_tp_utils.py::BuildExpandedKeysTest::test_moe_layer_only PASSED
tests/model_executor/test_tp_utils.py::BuildExpandedKeysTest::test_no_placeholder_keys_pass_through PASSED
tests/model_executor/test_tp_utils.py::GQATensorOpsTest::test_gqa_merge_reconstructs_weights PASSED
tests/model_executor/test_tp_utils.py::GQATensorOpsTest::test_gqa_split_on_matrix_rows PASSED
tests/model_executor/test_tp_utils.py::GQATensorOpsTest::test_gqa_split_returns_all_partitions PASSED
tests/model_executor/test_tp_utils.py::GQATensorOpsTest::test_gqa_split_with_rank_and_repeat_kv PASSED
tests/model_executor/test_tp_utils.py::GQATensorOpsTest::test_split_or_merge_func_v1_default_path PASSED
tests/model_executor/test_tp_utils.py::GQATensorOpsTest::test_split_or_merge_func_v1_gqa_path PASSED
tests/model_executor/test_tp_utils.py::GQATensorOpsTest::test_split_or_merge_func_v1_row_bias PASSED
tests/model_executor/test_tp_utils.py::GQATensorOpsTest::test_split_or_merge_qkv_dispatch PASSED

============================== 26 passed in 0.16s ==============================

YuanRisheng

把减少的miss行数在描述里补充一下，另外把ai的review建议都改一下

…dels/tp_utils.py

…ean-up-tests Update tp_utils tests

codecov-commenter · 2025-11-24T15:05:59Z

Codecov Report

✅ All modified and coverable lines are covered by tests.
⚠️ Please upload report for BASE (develop@95b3931). Learn more about missing BASE report.

Additional details and impacted files

@@            Coverage Diff             @@
##             develop    #5055   +/-   ##
==========================================
  Coverage           ?   58.86%           
==========================================
  Files              ?      317           
  Lines              ?    38679           
  Branches           ?     5812           
==========================================
  Hits               ?    22767           
  Misses             ?    14081           
  Partials           ?     1831

Flag	Coverage Δ
GPU	`58.86% <ø> (?)`

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

Add tp_utils tests

94791ff

Copilot AI review requested due to automatic review settings November 15, 2025 10:16

paddle-bot bot added the contributor External developers label Nov 15, 2025

Copilot started reviewing on behalf of xunyoyo November 15, 2025 10:16 View session

Copilot finished reviewing on behalf of xunyoyo November 15, 2025 10:18

Copilot AI reviewed Nov 15, 2025

View reviewed changes

luotao1 mentioned this pull request Nov 17, 2025

【Hackathon 9th】开源贡献个人挑战赛冲刺赛 PaddlePaddle/Paddle#76333

Open

luotao1 added the PaddlePaddle Hackathon label Nov 17, 2025

luotao1 assigned luotao1 and YuanRisheng Nov 17, 2025

YuanRisheng reviewed Nov 24, 2025

View reviewed changes

xunyoyo added 3 commits November 24, 2025 21:45

Merge branch 'PaddlePaddle:develop' into fastdeploy/model_executor/mo…

e3653d3

…dels/tp_utils.py

Add header and tidy tp_utils test stubs

2732078

Merge pull request #23 from xunyoyo/codex/add-copyright-header-and-cl…

d677963

…ean-up-tests Update tp_utils tests

YuanRisheng approved these changes Nov 25, 2025

View reviewed changes

YuanRisheng merged commit edf0d09 into PaddlePaddle:develop Nov 25, 2025
17 checks passed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[CI] 【Hackathon 9th Sprint No.24】NO.24 功能模块单测补充 #5055

[CI] 【Hackathon 9th Sprint No.24】NO.24 功能模块单测补充 #5055

xunyoyo commented Nov 15, 2025 •

edited

Loading

Uh oh!

paddle-bot bot commented Nov 15, 2025

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Nov 15, 2025

Uh oh!

Copilot AI Nov 15, 2025

Uh oh!

Copilot AI Nov 15, 2025

Uh oh!

xunyoyo commented Nov 17, 2025

Uh oh!

YuanRisheng left a comment

Uh oh!

codecov-commenter commented Nov 24, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

		"""Unit tests for tensor parallel utility helpers."""

		from __future__ import annotations

	def _fn(weight, *, is_column=True, is_naive_2fuse=False): # pylint: disable=unused-argument
	def _fn(weight, *, is_column=True):

		np.testing.assert_array_equal(merged, np.array([0, 1, 2, 3], dtype=np.float32))


		if __name__ == "__main__": # pragma: no cover - entry point for python -m unittest

	if __name__ == "__main__": # pragma: no cover - entry point for python -m unittest
	if __name__ == "__main__":

[CI] 【Hackathon 9th Sprint No.24】NO.24 功能模块单测补充 #5055

[CI] 【Hackathon 9th Sprint No.24】NO.24 功能模块单测补充 #5055

Conversation

xunyoyo commented Nov 15, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Motivation

Modifications

Usage or Command

Accuracy Tests

Checklist

Uh oh!

paddle-bot bot commented Nov 15, 2025

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Uh oh!

Copilot AI Nov 15, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 15, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Nov 15, 2025

Choose a reason for hiding this comment

Uh oh!

xunyoyo commented Nov 17, 2025

Uh oh!

YuanRisheng left a comment

Choose a reason for hiding this comment

Uh oh!

codecov-commenter commented Nov 24, 2025

Codecov Report

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

xunyoyo commented Nov 15, 2025 •

edited

Loading