Skip to content

AGENTS.md

Latest commit

 

History

History
67 lines (50 loc) · 2.08 KB

AGENTS.md

File metadata and controls

67 lines (50 loc) · 2.08 KB
Raw

Repository Guidelines

Project Layout

PyPTO Serving is a local inference stack for Qwen3-14B generation on Ascend NPUs. Runtime code lives under python/, with CLI entry points in python/cli/ and engine, scheduler, KV cache, model loading, sampling, and streaming logic in python/core/. Qwen3-14B examples and executor glue live under examples/model/qwen3_14b/. The pypto-lib/ submodule provides the kernel sources used by the NPU executor. Unit and lint tests live under tests/.

Skills

Repository-local agent skills are stored canonically in .agents/skills/. The .claude/skills and .codex/skills paths are symlinks to ../.agents/skills, so Claude-compatible and Codex-compatible tooling can use the same skill definitions without duplicate files. When adding or updating a skill, edit .agents/skills/<skill-name>/SKILL.md.

Development Commands

Initialize submodules after cloning:

git submodule update --init --recursive

Run the main unit tests:

python -m pytest tests/test_batching.py tests/test_parallel.py

Run lint checks used by pre-commit:

python tests/lint/check_headers.py
python tests/lint/check_english_only.py
ruff check --config ruff.toml .

Style And CI

Use Python 3.10-compatible code. Ruff is configured in ruff.toml with line length 110, target-version = "py310", F checks enabled, and F841 ignored. GitHub CI runs pre-commit plus the CLI and batching unit tests. Keep generated artifacts such as build_output/, caches, and compiled files out of commits.

NPU Verification

Device verification usually runs through task-submit. Preserve the runtime environment variables used by the examples unless the task explicitly changes them:

task-submit --device auto --max-time 0 --run \
  "PTO2_RING_HEAP=536870912 PTO2_RING_TASK_WINDOW=131072 PTO2_RING_DEP_POOL=131072 \
  python examples/model/qwen3_14b/npu_generate.py \
    --model-dir /data/linyifan/models/Qwen3-14B \
    --prompt 'Huawei is' \
    --platform a2a3 \
    --max-seq-len 512 \
    --max-new-tokens 5"

Use --profile for timing reports.