# Sputnik
Sputnik is a library of sparse linear algebra kernels and utilities for deep learning.
## Build
Sputnik uses the CMake build system. Sputnik depends on the CUDA toolkit (v10.1+) and supports SM70+. The only additional dependency for the library is [google/glog](https://github.com/google/glog). To build the library, enter the project directory and run the following commands:
`mkdir build && cd build`
`cmake .. -DCMAKE_BUILD_TYPE=Release`
`make -j12`
The test and benchmark suites additionally depend on [abseil/abseil-cpp](https://github.com/abseil/abseil-cpp), [google/googltest](https://github.com/google/googletest), and [google/benchmark](https://github.com/google/benchmark). These dependencies are includes as submodules in [third_party](https://github.com/google-research/sputnik/tree/os-build/third_party). To build the test suite and/or benchmark suite, set `-DBUILD_TEST=ON` and/or `-DBUILD_BENCHMARK=ON` in your `cmake` command.
`cmake .. -DCMAKE_BUILD_TYPE=Release -DBUILD_TEST=ON -DBUILD_BENCHMARK=ON -DCUDA_ARCHS="70;75"`
## Docker
Sputnik provides a [Dockerfile](https://github.com/google-research/sputnik/blob/os-build/Dockerfile) that builds the proper environment with all dependencies. Note that [nvidia-docker](https://github.com/NVIDIA/nvidia-docker) must be installed to run on GPU. To build the image, run the following command:
`docker build . -t sputnik-dev`
To launch the container with the sputnik source mounted under `/mount` (assuming you're working out of $HOME), run the following:
`sudo docker run --runtime=nvidia -v ~/:/mount/ -it sputnik-dev:latest`
## Citation
If you make use of this library, please cite:
```
@inproceedings{sgk_sc2020,
author = {Trevor Gale and Matei Zaharia and Cliff Young and Erich Elsen},
title = {Sparse {GPU} Kernels for Deep Learning},
booktitle = {Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis, {SC} 2020},
year = {2020},
}
```
The sparse models and dataset of sparse matrices from deep neural networks from the above paper can be found [here](https://github.com/google-research/google-research/tree/master/sgk).
## Disclaimer
This is not an official Google product.
没有合适的资源?快使用搜索试试~ 我知道了~
用于稀疏矩阵运算的GPU内核库。_C++_CMake_下载.zip
共67个文件
h:36个
cc:14个
txt:8个
1.该资源内容由用户上传,如若侵权请联系客服进行举报
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
2.虚拟产品一经售出概不退款(资源遇到问题,请及时私信上传者)
版权申诉
0 下载量 22 浏览量
2023-04-13
23:41:29
上传
评论
收藏 103KB ZIP 举报
温馨提示
用于稀疏矩阵运算的GPU内核库。_C++_CMake_下载.zip
资源推荐
资源详情
资源评论
收起资源包目录
用于稀疏矩阵运算的GPU内核库。_C++_CMake_下载.zip (67个子文件)
sputnik-master
CMakeLists.txt 809B
AUTHORS 306B
LICENSE 11KB
cmake
Dependencies.cmake 1KB
modules
FindGlog.cmake 657B
Cuda.cmake 522B
CONTRIBUTING.md 587B
sputnik
CMakeLists.txt 5KB
vector_utils.h 4KB
type_utils.h 5KB
cuda_utils.h 1KB
matrix_utils.cu.cc 24KB
sputnik.h 1003B
spmm
CMakeLists.txt 964B
compute_utils.h 3KB
spmm_test.cu.cc 13KB
sparse_tile.h 8KB
cuda_spmm.h 9KB
spmm_config.h 4KB
dense_tile.h 10KB
cuda_spmm.cu.cc 29KB
predicate_utils.h 8KB
output_tile.h 4KB
spmm_benchmark.cu.cc 6KB
softmax
CMakeLists.txt 251B
sparse_softmax.cu.cc 3KB
softmax.h 1KB
softmax.cu.cc 3KB
sparse_softmax.h 1KB
utils
index_format.h 2KB
CMakeLists.txt 251B
index_format.cu.cc 2KB
test_utils.h 2KB
tiling_utils.h 2KB
bias_relu
CMakeLists.txt 251B
bias_relu.cu.cc 2KB
bias_relu.h 1KB
common.h 2KB
depthwise
CMakeLists.txt 1000B
width_padding.h 2KB
shape_utils.h 2KB
depthwise_test.cu.cc 11KB
depthwise_benchmark.cu.cc 5KB
filter_tile.h 3KB
cuda_depthwise.h 2KB
cuda_depthwise.cu.cc 15KB
computer.h 6KB
output_tile.h 3KB
input_tile.h 6KB
depthwise_config.h 2KB
matrix_utils.h 13KB
sddmm
CMakeLists.txt 968B
dense_to_shared.h 3KB
compute_utils.h 2KB
sddmm_benchmark.cu.cc 3KB
all_reduce.h 3KB
output_tile.h 2KB
cuda_sddmm.h 2KB
cuda_sddmm.cu.cc 9KB
sddmm_test.cu.cc 6KB
dense_to_reg.h 8KB
barrier.h 5KB
load_store.h 2KB
memory_aligner.h 6KB
Dockerfile 658B
.gitmodules 345B
third_party
benchmark
googletest
abseil-cpp
README.md 2KB
共 67 条
- 1
资源评论
快撑死的鱼
- 粉丝: 1w+
- 资源: 9154
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功