god-ml-data-training

What problem does it solve?

This Skill addresses the complex challenges of machine learning data preparation and model training, providing a comprehensive toolkit for data scientists and engineers.

Core Features & Use Cases

• Data Labeling: Offers tools for high-quality data annotation, including Label Studio, Prodigy, and Scale AI.
• Dataset Quality: Ensures data quality through principles like quality over quantity, class balance, and diversity.
• Training Loop Design: Provides guidance on designing robust training loops with PyTorch, including mixed precision and gradient accumulation.
• Loss Functions and Optimizers: Explains the selection of appropriate loss functions and optimizers for different scenarios.
• Learning Rate Scheduling: Offers strategies for effective learning rate adjustments during training.
• Regularization and Safety: Discusses techniques like dropout, batch normalization, and early stopping to prevent overfitting.
• Distributed Training: Covers distributed training methods like DDP, FSDP, and DeepSpeed ZeRO for large-scale models.
• Debugging and GPU Memory Management: Provides guidance on common debugging issues and efficient GPU memory management.

Quick Start

Load the god-ml-data-training skill to begin preparing your machine learning datasets and training your models.