OpenMoE: An Early Effort on Open Mixture-of-Experts Language Models
This work provides open-source tools and insights for the ML/AI community to develop more efficient MoE-based language models, though it is incremental as it builds on existing MoE concepts.
The authors tackled the problem of understanding Mixture-of-Experts (MoE) large language models by training and releasing OpenMoE, a series of open-sourced MoE models from 650M to 34B parameters, confirming that MoE-based models offer better cost-effectiveness than dense models. They also analyzed routing mechanisms, finding issues like context-independent specialization and performance degradation in sequential tasks.
To help the open-source community have a better understanding of Mixture-of-Experts (MoE) based large language models (LLMs), we train and release OpenMoE, a series of fully open-sourced and reproducible decoder-only MoE LLMs, ranging from 650M to 34B parameters and trained on up to over 1T tokens. Our investigation confirms that MoE-based LLMs can offer a more favorable cost-effectiveness trade-off than dense LLMs, highlighting the potential effectiveness for future LLM development. One more important contribution of this study is an in-depth analysis of the routing mechanisms within our OpenMoE models, leading to three significant findings: Context-Independent Specialization, Early Routing Learning, and Drop-towards-the-End. We discovered that routing decisions in MoE models are predominantly based on token IDs, with minimal context relevance. The token-to-expert assignments are determined early in the pre-training phase and remain largely unchanged. This imperfect routing can result in performance degradation, particularly in sequential tasks like multi-turn conversations, where tokens appearing later in a sequence are more likely to be dropped. Finally, we rethink our design based on the above-mentioned observations and analysis. To facilitate future MoE LLM development, we propose potential strategies for mitigating the issues we found and further improving off-the-shelf MoE LLM designs.