Installing Unsloth on Windows

For Windows, pip install unsloth now works, however you must have Pytorch previously installed.

Method #1 - Docker:

Docker might be the easiest way for Windows users to get started with Unsloth as there is no setup needed or dependency issues. unsloth/unsloth is Unsloth's only Docker image. For Blackwell and 50-series GPUs, use this same image - no separate image needed.

For installation instructions, please follow our Docker guide, otherwise here is a quickstart guide:

export NVIDIA_CONTAINER_TOOLKIT_VERSION=1.17.8-1
sudo apt-get update && sudo apt-get install -y \
  nvidia-container-toolkit=${NVIDIA_CONTAINER_TOOLKIT_VERSION} \
  nvidia-container-toolkit-base=${NVIDIA_CONTAINER_TOOLKIT_VERSION} \
  libnvidia-container-tools=${NVIDIA_CONTAINER_TOOLKIT_VERSION} \
  libnvidia-container1=${NVIDIA_CONTAINER_TOOLKIT_VERSION}

docker run -d -e JUPYTER_PASSWORD="mypassword" \
  -p 8888:8888 -p 2222:22 \
  -v $(pwd)/work:/workspace/work \
  --gpus all \
  unsloth/unsloth

Method #2 - WSL:

wsl.exe --install Ubuntu-24.04
wsl.exe -d Ubuntu-24.04

wsl

sudo apt update
sudo apt install python3 python3-full python3-pip python3-venv -y

pip install torch torchvision --force-reinstall --index-url https://download.pytorch.org/whl/cu130

pip install unsloth jupyter

jupyter notebook

Method #3 - Windows directly:

"C:\Program Files (x86)\Microsoft Visual Studio\Installer\vs_installer.exe" modify ^
--installPath "C:\Program Files\Microsoft Visual Studio\2022\Community" ^
--add Microsoft.Net.Component.4.8.SDK ^
--add Microsoft.Net.Component.4.7.2.TargetingPack ^
--add Microsoft.VisualStudio.Component.Roslyn.Compiler ^
--add Microsoft.Component.MSBuild ^
--add Microsoft.VisualStudio.Component.VC.Tools.x86.x64 ^
--add Microsoft.VisualStudio.Component.VC.Redist.14.Latest ^
--add Microsoft.VisualStudio.Component.VC.CMake.Project ^
--add Microsoft.VisualStudio.Component.VC.CLI.Support ^
--add Microsoft.VisualStudio.Component.VC.Llvm.Clang ^
--add Microsoft.VisualStudio.ComponentGroup.ClangCL ^
--add Microsoft.VisualStudio.Component.Windows11SDK.22621 ^
--add Microsoft.VisualStudio.Component.Windows10SDK.19041 ^
--add Microsoft.VisualStudio.Component.UniversalCRT.SDK ^
--add Microsoft.VisualStudio.Component.VC.Redist.MSM

pip install "unsloth[windows] @ git+https://github.com/unslothai/unsloth.git"

Notes

To run Unsloth directly on Windows:

• Install Triton from this Windows fork and follow the instructions here (be aware that the Windows fork requires PyTorch >= 2.4 and CUDA 12)

• In the SFTTrainer, set dataset_num_proc=1 to avoid a crashing issue:

Advanced/Troubleshooting

For advanced installation instructions or if you see weird errors during installations:

1. Install torch and triton. Go to https://pytorch.org to install it. For example pip install torch torchvision torchaudio triton

2. Confirm if CUDA is installated correctly. Try nvcc. If that fails, you need to install cudatoolkit or CUDA drivers.

3. Install xformers manually. You can try installing vllm and seeing if vllm succeeds. Check if xformers succeeded with python -m xformers.info Go to https://github.com/facebookresearch/xformers. Another option is to install flash-attn for Ampere GPUs.

4. Double check that your versions of Python, CUDA, CUDNN, torch, triton, and xformers are compatible with one another. The PyTorch Compatibility Matrix may be useful.

5. Finally, install bitsandbytes and check it with python -m bitsandbytes

Method #3 - Windows using PowerShell:

Step 1: Install Prerequisites

1. Install NVIDIA CUDA Toolkit:

   • Download and install the appropriate version of the NVIDIA CUDA Toolkit from CUDA Downloads.

   • Reboot your system after installation if prompted.

   • Note: No additional setup is required after installation for Unsloth.

2. Install Microsoft C++ Build Tools:

   • Download and install Microsoft Build Tools for Visual Studio from the official website.

   • During installation, select the C++ build tools workload. Ensure the MSVC compiler toolset is included.

3. Set Environment Variables for the C++ Compiler:

   • Open the System Properties window (search for "Environment Variables" in the Start menu).

   • Click "Environment Variables…".

   • Add or update the following under System variables:

     • CC: Path to the cl.exe C++ compiler. Example (adjust if your version differs):

     • CXX: Same path as CC.

   • Click OK to save changes.

   • Verify: Open a new terminal and type cl. It should show version info.

4. Install Conda

   1. Download and install Miniconda from the official website

   2. Follow installation instruction from the website

   3. To check whether conda is already installed, you can test it with conda in your PowerShell

Step 2: Run the Unsloth Installation Script

1. Download the unsloth_windows.ps1 PowerShell script by going through this link.

2. Open PowerShell as Administrator:

   • Right-click Start and select "Windows PowerShell (Admin)".

3. Navigate to the script’s location using cd:

4. Run the script:

Step 3: Using Unsloth

Activate the environment after the installation completes:

Unsloth and its dependencies are now ready!