Skip to content

Environment Setup

Prerequisites

Our Environment Setup for Reference

We run all the experiments on linux machines and did not test on Windows.

Here are a few environment we have tested for reference:

  • NVIDIA A6000 48GBx8, Ubuntu 24.04, CUDA 12.4
  • NVIDIA H100 96GBx2, Red Hat Enterprise Linux 9.5, CUDA 12.6.
  • NVIDIA H100 80GBx8, Ubuntu 24.04, CUDA 12.4
  • NVIDIA H200 141GBx8, Ubuntu 24.04, CUDA 12.4

Environment Setup

  1. Config SSH key for GitHub. One of the dependencies, MASE, requires SSH to clone and install. Please set up ~/.ssh/config accordingly (refer to Connecting to GitHub with SSH).

  2. Clone the project repository

    git clone https://github.com/AICrossSim/NewComputeBench.git
cd NewComputeBench
git submodule update --init

  3. Create a new conda environment

    conda env create -f environment.yaml

  4. Activate the new environment and install required packages

    conda activate new-compute

    We recommend check if the python and pip in $PATH are from the conda environment: 

    which python
which pip

    Then install the required packages: 

    pip install -r requirements.txt
pip install -e ./submodules/mase

    The MASE submodule provides the quantization backend used by PIM and other hardware simulation passes.

  5. (Optional) You may want to log in Wandb to track the training logs.

    wandb login