Acknowledgments

Oumi makes use of several libraries and tools from the open-source community. 🚀

We would like to acknowledge these projects:

Project	Usage in Oumi
Accelerate	Distributed training and mixed precision computations
bitsandbytes	Quantization and efficient optimizers for QLora
Datasets	Dataset loading and processing
jsonlines	JSON Lines data format handling
LLama.cpp	Efficient inference capabilities for quantized models
LM Evaluation Harness	Comprehensive suite for evaluating language models
NumPy	Numerical operations and data manipulation
OmegaConf	Configuration system for managing model and training parameters
Pandas	Data manipulation and analysis
Peft	Parameter-efficient fine-tuning techniques
Pydantic	Ensures type checking and validation for configuration objects
PyTorch	Primary deep learning framework for model training and inference
SkyPilot	Cloud-agnostic deployment and management of training jobs
TensorBoard	Training visualization and monitoring
Tqdm	Progress bars to enhance user experience during long-running operations
Transformers	Core model architectures and utilities for working with transformer-based models
TRL	SFT and DPO training implementations
Typer	Command-line interface for Oumi commands
vllm	Fast inference for large language models
Weights & Biases	Experiment tracking and visualization

We are grateful to the developers and maintainers of these projects for their valuable contributions to the open-source community. 🙏