Installation Guide¶
RBLN SDK consists of device driver, compiler stack, and model zoo:
- RBLN Driver (
deb package) - RBLN Compiler (
whl package) - RBLN Model Zoo
Refer to the step-by-step installation guide provided below. Note that you need an RBLN portal account to proceed the installation. Please contact us for more information.
Step 1. RBLN Driver¶
Note
Step 1. RBLN Driver is intended for on-premise server users, and to install the RBLN device driver, users need root privileges on the server. Typically, the stable version of the RBLN device driver is already installed on the cloud server you are currently using. If you can see the RBLN NPU devices by executing the command ls /dev/rbln*, then you can skip Step 1. RBLN Driver and proceed directly to Step 2. RBLN Compiler.
If you are an on-premise server user, you can download the deb package from RBLN portal > RBLN SDK Installation > RBLN Driver. To install the RBLN device driver, run the following command:
Step 2. RBLN Compiler¶
Next step is to install the RBLN compiler, which requires access rights to Rebellions' private PyPI server (RBLN portal > RBLN SDK Installation > RBLN Compiler). To install the RBLN compiler, run the following command:
Step 3. RBLN Model Zoo¶
The last step is to set up the RBLN model zoo. RBLN model zoo is a collection of pre-trained models that you can use with RBLN SDK. You can access RBLN Model Zoo from GitHub.
To learn how to compile and deploy the model, please refer to the tutorials for Tutorial > TensorFlow (Vision) or Tutorial > PyTorch (Vision). The TensorFlow and PyTorch model zoo pages may also be helpful, as they provide simple instructions for running all models included in the model zoo.
HuggingFace transformers and diffusers libraries can be utilized with optimum RBLN. For further details, please refer to the following page Software > HuggingFace Model Support > Optimum RBLN.
If you encounter any issues during installation, please do not hesitate to contact us.