Local Install

It is recommended to use Docker to build, install, and utilize this repository. The reason for this recommendation is that using Docker allows you to use the repository without needing to change the environment and settings of the entire system. However, there may be reasons why you cannot or do not wish to use Docker, and in such cases, you might want to build and install the repository in your local environment. This article describes how to build and install this repository in your local environment.

1. Install CUDA 12.1

Install CUDA 12.1. For specific instructions, follow the steps described at https://developer.nvidia.com/cuda-12-1-1-download-archive. It is important to note that precisely CUDA 12.1 is required for building and using this repository. Other versions of CUDA, even if they differ only in minor versions, cannot be used.

2. Install Prerequisite Packages

At least, the following packages are necessary. Install them according to the specific methods in each distribution.

• autoconf
• automake
• g++
• libssl-dev
• libtool
• protobuf-compiler
• python3-dev
• python3-venv

3. Checkout This Repository

git checkout 'https://github.com/Cryolite/kanachan.git'

4. Run Installation Script

Execute the following command with the top-level directory of this repository as the current directory.

bin/install.sh /PATH/TO/VIRTUAL_ENVIRONMENT_DIR

Note that this process does not involve changes to the environment and settings of the entire system. Therefore, it can be executed with your user account, and the installation destination will be within the permissions of your account.

5. Perform Environment Configuration

To use this repository, it is necessary to execute the following command to modify the shell environment settings.

source /PATH/TO/VIRTUAL_ENVIRONMENT_DIR/bin/activate

To revert the changes to the shell settings made by the above step, execute the following command.

deactivate