PyTorch HammerBlade Port

This work aims to port PyTorch to HammerBlade.

How to build

This assumes that you have a working COSIM installed. Then you can either put hb-pytorch under bsg_bladerunner, or set BRG_BSG_BLADERUNNER_DIR to your bladerunner path.

Clone hb-pytorch repo git clone -b hb-device git@github.com:cornell-brg/hb-pytorch.git
Create python virtual environment python3.6 -m venv ./venv_pytorch python3.6 -m venv ./venv_pytorch
Install dependencies pip install numpy pyyaml mkl mkl-include setuptools cmake cffi typing sklearn tqdm pytest
Init pytorch third party dependencies git submodule update --init --recursive
Setup building environment variables. You need to edit hb-pytorch/setup_cosim_build_env.sh and set BSG_MANYCORE_DIR to <bsg_bladerunner>/bsg_replicant/libraries
Build pytorch. This step can take up to 15 minutes cd hb-pytorch && python setup.py install

Run Pytests

Goto hb-pytorch directory cd hb-pytorch
Run pytest source ./run-hb-pytest.sh

Important files and directories related to HammerBlade

files used to run pytest (adapted from Baseline)

fragments/
environment.mk
baseline-README.md
run-hb-pytest.sh (source this one to run pytest!)
hb_device/torch/

HammerBlade device code

hb_device/torch/kernel

Pytest tests

hb_device/torch/tests/
hb_device/torch/tests/targets.py (Register a new test here!)

files that interacts with HammerBlade CUDALite runtime

c10/hammerblade/

zzy82518996/hb-pytorch