Build the docker image and run container:
cd torch_bench
(torch_bench) ./build_run.sh [SAVE_NAME (default: torch_benchmark.log)] [TORCH_VERSION (default: 1.13.0)] [CUDA_VERSION (default: 11.6)]
Or only run container:
cd torch_bench
(torch_bench) ./run.sh [SAVE_NAME (default: torch_benchmark.log)]
The result file will be saved in torch_bench/SAVE_NAME
.
If you want to add a new benchmark case, please add the code under torch_bench/codes
.
If you need additional packages, please add it to torch_bench/requirements.txt
.
Before running script, please make sure that your code is runnable with this command:
python3 [YOUR_CODE]