skills.rest
PaddlePaddle
PaddlePaddleOFFICIAL
Explore all skills by this author →
WebsiteGitHub

write-fastdeploy-unittest

Generates CI-compliant pytest unit tests for FastDeploy source files.

3.7k|759|Updated Jun 27, 2022
Search Tags:#Code Coverage#ci#PyTest#paddlepaddle#fastdeploy#Unit testing#gpu testing
One-click install
npx skills add https://github.com/PaddlePaddle/FastDeploy --skill write-fastdeploy-unittest
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: write-fastdeploy-unittest
Source: https://github.com/PaddlePaddle/FastDeploy/tree/main/.claude/skills/write-fastdeploy-unittest
Command: npx skills add https://github.com/PaddlePaddle/FastDeploy --skill write-fastdeploy-unittest

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Writing unit tests for the FastDeploy inference toolkit requires following strict CI conventions: correct test classification (multi-GPU sequential vs single-GPU parallel), port variable usage, log isolation, resource cleanup, and an 80% diff coverage threshold. This Skill automates that process so contributors get compliant tests without memorizing the CI rules.

Core Features & Use Cases

  • Automatic Test Pattern Selection: Chooses among four patterns (pure logic, GPU kernel, offline inference, E2E serving) based on the code under test.
  • Coverage-Driven Test Generation: Accepts coverage report lines or incremental diff-coverage JSON and writes tests targeting the exact uncovered lines.
  • CI Convention Compliance: Enforces naming conventions, port environment variables, log isolation via FD_LOG_DIR, resource cleanup, and black/isort/flake8 style checks.
  • Use Case: Paste a coverage line like fastdeploy/model_executor/model_loader/default_loader.py 48 32 14 0 26% 37-38, 42 and the agent appends tests to the existing test file covering those missing branches.

Quick Start

Use the write-fastdeploy-unittest skill to add unit tests for fastdeploy/cache_manager/transfer_factory/file_store/file_store.py.

Frequently Asked Questions about write-fastdeploy-unittest

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I write unit tests for FastDeploy that pass CI?▼

Use pytest or unittest, place the file in the matching tests/ subdirectory, and follow the four test patterns: pure logic with mocks, GPU kernel tests with @pytest.mark.gpu, offline inference via the LLM API, or E2E serving with subprocess servers. Verify with coverage run and pre-commit style checks.

How to generate tests from a coverage report for uncovered lines?▼

Paste the coverage report line or incremental diff-coverage JSON directly into the request. The agent parses the violation_lines or missing line ranges, analyzes the source logic at those lines, and writes tests specifically targeting those uncovered branches.

Why is my FastDeploy test classified as multi-GPU sequential in CI?▼

Any test referencing port variables like FD_API_PORT or FD_ENGINE_QUEUE_PORT, or containing tensor_parallel_size patterns, is classified as multi-GPU and runs sequentially. This applies even to single-GPU service tests with tensor_parallel_size=1.

Does FastDeploy testing require a GPU machine?▼

Tests run on CI runners with 2x NVIDIA H20 GPUs. Pure logic tests need no GPU and mock external dependencies, while GPU kernel tests require a GPU and are marked with @pytest.mark.gpu so they skip on non-GPU platforms.

What coverage threshold do FastDeploy PRs need to meet?▼

Pull requests require 80% diff coverage, enforced with diff-cover using --fail-under=80. Hardware-dependent branches that cannot be mocked, such as NUMA topology or RDMA paths, are acceptable coverage gaps.

Should I create a new test file or append to an existing one?▼

Always search for an existing test file covering the same module first and append new test cases to it. Only create a new file when none exists, following the naming convention of test_<module>.py with parent prefixes for generic names.

Related Skills

Explore More Software Engineering Skills→
tmalcolm-0607tmalcolm-0607

coverage-fix

Diagnose and resolve Azure DevOps diff coverage failures for .NET pull requests.

Community
Advanced
TechlemariamTechlemariam

coverage-check

Validates test coverage thresholds from nyc and Jest reports in CI pipelines and pull requests.

Community
Intermediate
LincyawLincyaw

verify

Run Black, isort, Flake8, and pytest checks on Python projects.

Community
Intermediate
NVIDIANVIDIA
14

k8s-launch-kit-config

Explain, create, and edit k8s-launch-kit configuration files for Kubernetes network deployment.

Official
Advanced