This research evaluates multi-node and multi-GPU frameworks for training large language models on electrocardiogram data, offering critical insights for medical AI deployment.

• Compares distributed deep learning frameworks including Horovod, DeepSpeed, PyTorch, and TensorFlow
• Analyzes performance across different HPC configurations for medical data processing
• Provides guidance on optimal scalability solutions for ECG-based language models

This research is vital for advancing AI in cardiology diagnostics by identifying efficient computational approaches that can handle the complexity of ECG data while reducing training time and resource requirements.

Scalability Evaluation of HPC Multi-GPU Training for ECG-based LLMs