mirror of
https://github.com/pytorch/pytorch.git
synced 2026-11-20 08:39:27 +08:00
63bd3d6f53
### Summary Part of #106571. The **tools/ + torchgen/ + functorch/ + .github/ + .ci/** slice of the B007 (unused loop control variable) cleanup: 23 files. Most hunks are ruff's own fix — rename the unused loop variable with a leading underscore. Three dict iterations that only use one half of the pair are rewritten as `for v in d.values()` / `for k in d`, since prefixing the unused target there would just trade B007 for `PERF102`/`SIM118`. In `omniglot_loaders.py` the rewrite also collapses a three-line `for (label, imgs,) in temp.items():` into one line; its trailing comment moves to its own line above the loop so the result stays under the formatter's column limit (a single-line version with the comment attached fails PYFMT). ### Deliberately deferred Two sites in `tools/autograd` are left for the final slice, because the loop variable is used outside the loop body: | file | why | |---|---| | `gen_variable_type.py` | `edge_off` is a for/else index search, read after the `else` clause | | `load_derivatives.py` | `defn` is captured by the nested `set_up_derivatives` closure — renaming it produces F821 | Both need `# noqa: B007`, which `RUF100` rejects while B007 is still in the ignore list, so they have to land in the same PR that turns the code on. Split out of #189319 (whole tree at once, too wide to review). B007 stays in the `pyproject.toml` ignore list until the final slice, so this is a no-op for CI today. ### Test plan ``` uvx ruff@0.14.4 check tools/ torchgen/ functorch/ .github/ .ci/ docs/ \ --config=pyproject.toml --select B007 --config "lint.ignore=[]" # -> only the 2 deferred tools/autograd sites remain uvx ruff@0.14.4 check <changed files> --config=pyproject.toml # All checks passed! uvx ruff@0.14.4 format --check <changed files> --config=pyproject.toml # 23 files already formatted python -m py_compile <changed files> ``` > Authored with the assistance of an AI coding assistant (Claude). 🤖 Generated with [Claude Code](https://claude.com/claude-code) Pull Request resolved: https://github.com/pytorch/pytorch/pull/192151 Approved by: https://github.com/jansel
This folder contains a number of scripts which are used as
part of the PyTorch build process. This directory also doubles
as a Python module hierarchy (thus the __init__.py).
Overview
Modern infrastructure:
- autograd - Code generation for autograd. This includes definitions of all our derivatives.
- jit - Code generation for JIT
- shared - Generic infrastructure that scripts in
tools may find useful.
- module_loader.py - Makes it easier to import arbitrary Python files in a script, without having to add them to the PYTHONPATH first.
Build system pieces:
- setup_helpers - Code generation entry points and related helpers invoked from the CMake build.
- build_libtorch.py - Script for building libtorch, a standalone C++ library without Python support. This build script is tested in CI.
Developer tools which you might find useful:
- git_add_generated_dirs.sh and git_reset_generated_dirs.sh - Use this to force add generated files to your Git index, so that you can conveniently run diffs on them when working on code-generation. (See also generated_dirs.txt which specifies the list of directories with generated files.)
Important if you want to run on AMD GPU:
- amd_build - HIPify scripts, for transpiling CUDA
into AMD HIP. Right now, PyTorch and Caffe2 share logic for how to
do this transpilation, but have separate entry-points for transpiling
either PyTorch or Caffe2 code.
- build_amd.py - Top-level entry point for HIPifying our codebase.
Tools which are only situationally useful:
- docker - Dockerfile for running (but not developing) PyTorch, using the official conda binary distribution. Context: https://github.com/pytorch/pytorch/issues/1619
- download_mnist.py - Download the MNIST dataset; this is necessary if you want to run the C++ API tests.