PithTrain: A Compact and Agent-Native MoE Training System
This work addresses the high cost and complexity for AI coding agents to understand and extend existing MoE training frameworks, offering a more efficient development path for framework engineers and researchers.
This paper introduces PithTrain, a compact and agent-native Mixture-of-Experts (MoE) training framework designed to improve the efficiency of AI coding agents in developing and evolving training systems. PithTrain achieves throughput comparable to production frameworks while significantly enhancing agent-task efficiency, demonstrating up to 62% fewer Agent Turns and 64% less Active GPU Time on real-world tasks.
Mixture-of-Experts (MoE) has become the dominant architecture for frontier language models. To meet this demand, production frameworks have built optimized MoE training stacks over years of engineering effort. Yet evolving these stacks for new architectures and system optimizations remains expensive. With the rise of AI coding agents, they could automate parts of training-framework development and accelerate this evolution. But applying them to these existing frameworks carries hidden costs, invisible to today's throughput-only evaluations. We name this missing dimension agent-task efficiency (ATE): the cost of using coding agents to understand, operate, and extend a framework. Grounded in four agent-native design principles, we build PithTrain, a compact, agent-native MoE training framework. We further introduce ATE-Bench, covering real-world training-framework tasks. Our evaluation shows PithTrain matches the throughput of production frameworks, and on ATE-Bench, PithTrain enables higher agent-task efficiency, with up to 62% fewer Agent Turns and 64% less Active GPU Time.