Development Environment for Apple Silicon¶

Environment¶

(mypy310) kaikailiu@kaikais-mbp ~ % nano ~/.zshrc
(mypy310) kaikailiu@kaikais-mbp ~ % source ~/.zshrc

Xcode Tools¶

Install Xcode command-line tools (CLT), this includes all the useful open-source tools Unix developers expect, like git, clang, and more.

xcode-select --install #git will be installed

Git¶

% git config --global user.email "kaikai.liu@sjsu.edu"
% git config --global user.name "Kaikai Liu"

Brew¶

Install Brew: package manager for Macs

/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

The script will install Homebrew to the correct directory for Apple silicon, which is /opt/homebrew

After the installation is complete, the terminal will give you next steps. The terminal will likely give you a line to add to your .zshrc file. To do this, open your .zshrc file in a text editor.

echo >> /Users/kaikailiu/.zprofile
echo 'eval "$(/opt/homebrew/bin/brew shellenv)"' >> /Users/kaikailiu/.zprofile
eval "$(/opt/homebrew/bin/brew shellenv)"
brew help #test brew

Install LLVM:

brew install llvm
brew --prefix llvm

If you need to have llvm first in your PATH, run: echo ‘export PATH=“/opt/homebrew/opt/llvm/bin:$PATH”’ >> ~/.zshrc

For compilers to find llvm you may need to set: export LDFLAGS=“-L/opt/homebrew/opt/llvm/lib” export CPPFLAGS=“-I/opt/homebrew/opt/llvm/include”

Then append /lib/libLLVM.dylib to the output of that command. For example, if brew –prefix llvm outputs /usr/local/opt/llvm, then the full path would be /usr/local/opt/llvm/lib/libLLVM.dylib.

Set the DRJIT_LIBLLVM_PATH Environment Variable:

export DRJIT_LIBLLVM_PATH=/path/to/libLLVM.dylib

Open the file in a text editor (e.g., nano ~/.zshrc). Add the line: export DRJIT_LIBLLVM_PATH=/path/to/libLLVM.dylib

Python Environment¶

To install conda environment in Mac, there are several options. You can install miniforge or miniconda.

Install python via miniforge¶

Miniforge is a minimal installer for Conda that works well with Apple Silicon. You can download it from miniforge. Miniforge is a community-driven, minimal Conda installer focused on the conda-forge channel.

% bash Miniforge3-MacOSX-arm64.sh #installed to default location: /Users/xxx/miniforge3
#bash Miniforge3-MacOSX-x86_64.sh #if you are using Intel CPU
% source ~/miniforge3/bin/activate
% conda init bash #restart the terminal

If you’d prefer that conda’s base environment not be activated on startup, run the following command when conda is activated: conda config --set auto_activate_base false

Install Python packages in Macbook Pro M4Pro:

(base) kaikai@MacBook-Pro ~ % conda create --name py312 python=3.12
conda activate py312
conda install pandas numpy matplotlib seaborn scikit-learn jupyter
conda install -y Pillow scipy pyyaml scikit-image 
pip install opencv-python-headless
pip install PyQt5
pip3 install torch torchvision torchaudio

Follow the installation guide, and do not forget to restart your terminal after it is completed.

Install python via miniconda¶

Download miniconda from link

% curl https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-x86_64.sh -o Miniconda3-latest-MacOSX-x86_64.sh
% bash Miniconda3-latest-MacOSX-x86_64.sh -b -u
lkk@kaikais-mbp2019 Developers % source ~/miniconda3/bin/activate
(base) lkk@kaikais-mbp2019 Developers % conda init bash

Restart your terminal after it is completed.

In Mac, if you face md5sum: command not found problem, install it via brew

brew update
brew upgrade
brew install md5sha1sum

Create Conda environment with python¶

After miniconda or miniforge3 installed, you can use the conda command to create python environment

conda search python #check existing python versions
#remove existing environment if needed
#conda remove -n ENV_NAME --all
% conda create --name py310 python=3.10 #install python3.10
#% conda create --name py312 python=3.12 #install python3.12
% conda activate py310
% python -V
% conda deactivate 
conda info --envs #check existing conda environment
% conda env list
% conda env remove --name myenv

A Minimal Setup for Data science:

conda install pandas numpy numba matplotlib seaborn scikit-learn jupyter
conda install -y Pillow scipy pyyaml scikit-image 
pip install opencv-python-headless
pip install PyQt5
pip install pyqtgraph
pip install pyqt6
pip install pyside6 #Side6 and QT5 works in Mac

Setup for ADI IIO:

(py312) kaikailiu@kaikais-mbp Documents % conda install -c conda-forge pylibiio
Channels:
 - conda-forge
 - defaults
Platform: osx-arm64
% pip install pyadi-iio
pip install pyqt5 pyqt6 PySide6 pyqtgraph opencv-python-headless PyOpenGL PyOpenGL_accelerate pyopengl
pip install sionna DeepMIMO
(mypy310) (base) kaikailiu@Kaikais-MBP radarsensing % pip install tensorflow==2.14.0
#Test tensorflow
python3 -c "import tensorflow as tf; print(tf.reduce_sum(tf.random.normal([1000, 1000])))"
pip install sionna

Jupyter notebook setup¶

conda install -y -c conda-forge jupyterlab
conda install -y ipykernel
jupyter kernelspec list #view current jupyter kernels
ipython kernel install --user --name=py312

Google Cloud¶

Install the Google Cloud CLI: https://cloud.google.com/sdk/docs/install-sdk

curl https://sdk.cloud.google.com | bash
% ls ~/google-cloud-sdk
google-cloud-sdk % ./bin/gcloud init

Network¶

Install zrok tunnel:

kaikailiu@Kaikais-MacBook-Pro Downloads % mkdir -p /tmp/zrok && tar -xf ./zrok_0.4.39_darwin_arm64.tar.gz -C /tmp/zrok
% cd /tmp/zrok
% mkdir -p bin && install /tmp/zrok/zrok bin/
#nano ~/.zshrc
% export PATH=/tmp/zrok/bin:$PATH
% zrok version
% zrok enable DmXXX #add environment
ssh lkk@127.0.0.1 -p 9191

If you want to ssh into a remote linux machine via ssh, enter share in the Linux machine

$ zrok share private --backend-mode tcpTunnel 192.168.137.14:22
#it will show access your share with: zrok access private gdjf3oz1pudh

In your local Mac, enter zrok access private gdjf3oz1pudh, it will show tcp://127.0.0.1:9191 -> gdjf3oz1pudh. In another Mac terminal, ssh into the remote machine

% ssh lkk@127.0.0.1 -p 9191
# same to this: ssh lkk@localhost -p 9191

Problems¶

“torchvision.ops box_ops Couldn’t load custom C++ ops.”: upgrade torch and torchvision to the same version.

Ollama¶

Download Ollama for Mac, Once the download is complete, locate the .zip file and extract its contents. This should create Ollama.app. Drag Ollama.app to your Applications folder. Open the Applications folder and double-click on Ollama.app. Follow the setup wizard to complete the installation. The wizard will prompt you to install the command line version (ollama). Verify Olllama and make sure its working

  % ollama --version
  ollama version is 0.5.7
  % ollama run llama3.2
  >>> /bye
  % ollama show llama3.2
  % ollama list #List models on your computer
  % ollama ps
  % ollama stop llama3.2

Try accessing it as well : http://127.0.0.1:11434/

https://github.com/ollama/ollama

https://ollama.com/library/deepseek-r1

Install OpenWebui: https://github.com/open-webui/open-webui

% conda create --name mypy311 python=3.11
% conda activate mypy311
pip install open-webui
pip install ollama
% open-webui serve                   
Loading WEBUI_SECRET_KEY from file, not provided as an environment variable.
Generating a new secret key and saving it to /Users/kaikailiu/.webui_secret_key
Loading WEBUI_SECRET_KEY from /Users/kaikailiu/.webui_secret_key

To manage your Ollama instance in Open WebUI, Go to Admin Settings in Open WebUI. Navigate to Connections > Ollama > Manage (click the wrench icon).

ollama run deepseek-r1:32b --verbose
(base) kaikailiu@Kaikais-MacBook-Pro ~ % ollama ps                 
NAME               ID              SIZE     PROCESSOR    UNTIL              
deepseek-r1:32b    38056bbcbb2d    21 GB    100% GPU     4 minutes from now

On Mac, the models will be download to ~/.ollama/models; On Linux (or WSL), the models will be stored at /usr/share/ollama/.ollama/models

Ollama with Langchain: https://python.langchain.com/docs/integrations/chat/ollama/