/cuda-rt-hook

Primary LanguageC++

The plt hook technology used refers to plthook

mock pytorch cuda runtime interface

  • build wheel package
    pip wheel .

  • direct install
    pip install .

collect cuda operator call stack

  • find nvcc installed path
    which nvcc
  • replace nvcc with my nvcc
    mv /usr/local/bin/nvcc /usr/local/bin/nvcc_b
    chmod 777 tools/nvcc
    cp tools/nvcc /usr/local/bin/nvcc
  • build and install pytorch
  • build and install cuda_mock
  • import cuda_mock after import torch
  • run your torch train script
  • we will dump the stack into console

收集cuda 算子调用堆栈

  • 找到nvcc安装路径 which nvcc
  • 用我们的nvcc替换系统的nvcc（我们只是在编译选项加了-g）
    mv /usr/local/bin/nvcc /usr/local/bin/nvcc_b
    chmod 777 tools/nvcc
    cp tools/nvcc /usr/local/bin/nvcc
  • 构建并且安装pytorch
  • 构建并且安装cuda_mock
  • 注意要在import torch之后import cuda_mock
  • 开始跑你的训练脚本
  • 我们将会把堆栈打印到控制台

example

python test/test_import_mock.py

debug

  • export LOG_LEVEL=0