GPU-dev

1. download cuda-specific libtorch (2.3.1 with cuda 11.8)

   $ wget https://download.pytorch.org/libtorch/cu118/libtorch-shared-with-deps-2.3.1%2Bcu118.zip

2. download typhoon-blade, which is the latest version

   $ git clone git@github.com:anqin/blade-build.git

3. download the GPU docker

   $ docker pull registry.cn-hangzhou.aliyuncs.com/anqindev/ubuntu22-dev:ubuntu22-tf2.15-dev1

CPU-dev

1. download cuda-specific libtorch (2.3.1 with cuda 11.8)

   $ wget https://download.pytorch.org/libtorch/cpu/libtorch-shared-with-deps-2.3.1%2Bcpu.zip

2. download typhoon-blade, which is the v2.0-alpha

   $ git clone -b v2.0-alpha git@github.com:anqin/blade-build.git

3. download the GPU docker

   $ docker pull registry.cn-hangzhou.aliyuncs.com/anqindev/ubuntu22-dev:

Build

4. path tree would be:

   copy the BUILD.libtorh (for GPU or CPU) to corrent dir:

   pytorch_gpu_dev |- BLADE_ROOT |- BUILD |- .cpp/.h |- libtorch_cpu |- BUILD |- |- libtorch_gpu |- BUILD |-

5. mount the dirs to docker image

   • list the docker images and find the

   • for GPU

     $ bash docker_env_setup_gpu.sh

   • for CPU

     $ bash docker_env_setup.sh

6. build the codes

   $ blade build ...

7. run the example codes

   $ export LD_LIBRARY_PATH=<path/to/libtorch/lib64> $ ./mnist_sample_main $ ./load_model_main net.pt