mirror of
https://github.com/deepseek-ai/DeepInfer.git
synced 2026-10-31 00:36:20 +08:00
4a5606ad2d
Groundwork for replacing TileLang with Triton on the H20 path, plus the first ported kernel. The scope is measured rather than guessed. `DEEPINFER_TRACE_KERNELS=<path>` proxies every `*_tl` factory in the dispatcher and records its first use; one full text-generation run on 8xH20 says exactly 24 kernels execute. `H20-TRITON-PLAN.md` lists them, in dependency order. `DEEPINFER_OPS_BACKEND` selects the backend package (unset keeps the platform default, so nothing changes by omission) and `DEEPINFER_OPS_FALLBACK` names individual kernels that must stay on the default. Kernels the selected backend does not export at all are taken from the default with a one-line notice, which lets a partial port run end to end the whole way through instead of only at the end. That coexistence is the reason for a parallel backend rather than an in-place rewrite: the TileLang kernel is a validated baseline, so each Triton kernel can be diffed against it on real shapes through the same public interface function, which is a much stronger check than comparing against a fresh Python reference. `scripts/h20_ops_ab.py` does that swap and reports per case. First kernel: `gather_tl`, all 9 cases bitwise identical (zeros filled for out-of-range indices, int8/bfloat16 payloads, strided rows, int32/int64 indices, empty and single-row inputs).
89 lines
2.6 KiB
TOML
89 lines
2.6 KiB
TOML
[build-system]
|
|
requires = [
|
|
"ninja",
|
|
"setuptools >= 77.0.3",
|
|
"torch >= 2.7",
|
|
]
|
|
build-backend = "setuptools.build_meta"
|
|
|
|
[project]
|
|
name = "deepinfer"
|
|
version = "0.1.0"
|
|
description = "DeepInfer: High-performance, large-scale inference framework for DeepSeek models"
|
|
readme = "README.md"
|
|
license = "Apache-2.0"
|
|
license-files = ["LICENSE"]
|
|
requires-python = ">=3.12"
|
|
authors = [{ name = "DeepSeek-AI" }]
|
|
keywords = ["llm", "inference", "deepseek", "cuda", "ascend"]
|
|
classifiers = [
|
|
"Development Status :: 4 - Beta",
|
|
"Intended Audience :: Developers",
|
|
"Programming Language :: Python :: 3.12",
|
|
"Topic :: Scientific/Engineering :: Artificial Intelligence",
|
|
]
|
|
dynamic = ["dependencies"]
|
|
urls = { Repository = "https://github.com/deepseek-ai/deepinfer", Issues = "https://github.com/deepseek-ai/deepinfer/issues" }
|
|
|
|
[tool.pytest.ini_options]
|
|
# Use importlib import mode so that collecting test files inside the deepinfer
|
|
# package (e.g. deepinfer/model_executor/layers/tilelang_ops/*.py) does NOT
|
|
# trigger the package __init__.py chain (which imports heavy deps like
|
|
# deep_gemm).
|
|
addopts = ["--import-mode=importlib"]
|
|
testpaths = ["tests"]
|
|
markers = [
|
|
"ci: core tests that run on every CI pipeline",
|
|
"ci_npu: core tests that run on the NPU/Ascend CI pipeline",
|
|
"benchmark: performance tests selected by benchmark jobs",
|
|
"long_context: long-input/output decode PPL regression",
|
|
"gpu: unit tests requiring a working GPU",
|
|
]
|
|
|
|
[tool.ruff.lint]
|
|
select = [
|
|
# pycodestyle
|
|
"E",
|
|
# Pyflakes
|
|
"F",
|
|
# pyupgrade
|
|
"UP",
|
|
# flake8-bugbear
|
|
"B",
|
|
# flake8-simplify
|
|
"SIM",
|
|
# isort
|
|
"I",
|
|
# type hint
|
|
"ANN",
|
|
# disallow specific packages
|
|
"TID251",
|
|
]
|
|
ignore = [
|
|
# missing return type for private function; dynamically typed Any
|
|
"ANN202", "ANN401",
|
|
# star imports
|
|
"F405", "F403",
|
|
# lambda expression assignment
|
|
"E731",
|
|
# f-string without any placeholders
|
|
"F541",
|
|
# line too long, handled by black formatting
|
|
"E501",
|
|
# module level import not at top of file
|
|
"E402",
|
|
# function definition does not bind loop variable
|
|
"B023",
|
|
]
|
|
|
|
[tool.ruff.lint.per-file-ignores]
|
|
# Triton kernel bodies: parameters are pointers and scalars that Triton itself types, and the
|
|
# kernels are launched through `kernel[grid](...)` rather than called directly, so the usual
|
|
# signature annotations would be either wrong or rejected by the JIT.
|
|
"deepinfer/model_executor/layers/tilelang_ops/triton/*.py" = ["ANN001", "ANN201", "ANN202"]
|
|
|
|
[tool.ruff]
|
|
line-length = 140
|
|
include = ["./**/*.py"]
|
|
extend-exclude = ["vendor"]
|