#
Prepare Fine-tuning on MoAI Platform
The MoAI Platform can be configured with various GPUs, yet it provides a consistent user experience through a unified interface (CLI). This uniform access allows all users to interact with the system in the same way, making it more efficient and intuitive.
The MoAI Platform supports Python-based programming, similar to typical AI training environments. This document focuses on setting up and using a conda virtual environment as the standard configuration for AI training.
#
Setting up a Conda Environment
To begin training, first create a conda environment:
$ conda create --name <my-env> python=3.8
Replace
<my-env>
with your desired environment name.Activate the conda environment:
$ conda activate <my-env>
The MoAI Platform supports various PyTorch versions, allowing you to choose the one that fits your needs.
$ pip install torch==1.13.1+cu116.moreh24.5.0
Use the
moreh-smi
command to check the version of the installed Moreh solution and the details of the MoAI Accelerator in use. The current MoAI Accelerator is 4xLarge.2048GB . For more information about the MoAI Accelerator, refer to the specifications.$ moreh-smi +-----------------------------------------------------------------------------------------------------+ | Current Version: 24.5.0 Latest Version: 24.5.0 | +-----------------------------------------------------------------------------------------------------+ | Device | Name | Model | Memory Usage | Total Memory | Utilization | +=====================================================================================================+ | * 0 | MoAI Accelerator | 4xLarge.2048GB | - | - | - | +-----------------------------------------------------------------------------------------------------+
For optimal parameters recommended for fine-tuning each model on the MoAI Platform, refer to the LLM Fine-tuning parameter guide
For detailed usage of the moreh toolkit, including moreh-smi
and moreh-switch-model
, please refer to the Using the MoAI Platform Toolkit