point-in-rbbox
功能
实现point-in-rbbox算子（支持CUDA和Triton），用于计算3D旋转框内的点云，相比PyTorch版本显存有明显优化
依赖
- CUDA：11.8
- PyTorch：2.0.0
- Triton：2.0.0（使用Triton算子时需要）
安装
- pip安装
export TORCH_CUDA_ARCH_LIST="6.0 6.1 7.0 7.5 8.0 8.6 8.9 9.0+PTX" # 支持不同GPU架构【可选】
export ENABLE_BF16=1 # 支持BF16【可选，部分GPU可能不支持】
python3 -m pip install points-in-rbbox
- 源码安装
git clone https://github.com/lh9171338/points-in-rbbox.git
cd points-in-rbbox
export TORCH_CUDA_ARCH_LIST="6.0 6.1 7.0 7.5 8.0 8.6 8.9 9.0+PTX" # 支持不同GPU架构【可选】
export ENABLE_BF16=1 # 支持BF16【可选，部分GPU可能不支持】
python3 setup.py install
使用
- CUDA算子
from points_in_rbbox import points_in_rbbox_cuda
mask = points_in_rbbox_cuda(points, boxes)
- Triton算子（无需编译，自动适配GPU架构）
from points_in_rbbox import points_in_rbbox_triton
mask = points_in_rbbox_triton(points, boxes)
显存&耗时
| 方法 | dtype | 显存(GB) | 耗时(s) |
|---|---|---|---|
| points_in_rbbox_torch | FP32 | 10.0 | 12 |
| points_in_rbbox_torch | FP16 | 5.3 | 10 |
| points_in_rbbox_torch | BF16 | 5.3 | 10 |
| points_in_rbbox_cuda | FP32 | 1.2 | 9 |
| points_in_rbbox_cuda | FP16 | 1.0 | 9 |
| points_in_rbbox_cuda | BF16 | 1.0 | 9 |
| points_in_rbbox_triton | FP32 | 1.2 | 8 |
| points_in_rbbox_triton | FP16 | 1.0 | 8 |
| points_in_rbbox_triton | BF16 | 1.0 | 8 |
1 maintainer
Metadata
1 maintainer
Release files for points-in-rbbox 1.1.9
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| points_in_rbbox-1.1.9.tar.gz | 6.6 kB | Without provenance | Details for points_in_rbbox-1.1.9.tar.gz |
Release files / points_in_rbbox-1.1.9.tar.gz
| Download URL | points_in_rbbox-1.1.9.tar.gz |
|---|---|
| Size | 6.6 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
c666f0be6ba5391ee63ef393f87be6b11f3fe747e9c317e847e54f2dd5f1f7c8
|
|
BLAKE2b-256 checksum How to use checksums |
91c9c42537a8c9d352977d34d9fe6a7648ce6a37f87146df9645a7d570f0d210
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.8.18
|
Security for points-in-rbbox 1.1.9
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.
