Introduction
mxTuningKit is a toolkit designed for foundation models, helping users quickly start fine-tuning, evaluation, and inference jobs. It encapsulates downstream job startup scripts of foundation models and provides CLI and Python SDK for quick job startup.
Software Architecture
As shown in Figure 1, mxTuningKit consists of four modules:
- Interface function module: implements interfaces for fine-tuning, evaluation, and inference, and supports CLI and Python SDKs.
- Parameter verification module: verifies interface and configuration parameters.
- Basic tool module: implements parameter definition, task process, log, and tool classes.
- Configuration example module: provides configuration examples for customizing model parameters and freezing networks.
