MoMa: Efficient Early-Fusion Pre-training with Mixture of Modality-Aware Experts

• URL: http://arxiv.org/abs/2407.21770v3
• Date: Mon, 12 Aug 2024 16:20:37 GMT
• Title: MoMa: Efficient Early-Fusion Pre-training with Mixture of Modality-Aware Experts
• Authors: Xi Victoria Lin, Akshat Shrivastava, Liang Luo, Srinivasan Iyer, Mike Lewis, Gargi Ghosh, Luke Zettlemoyer, Armen Aghajanyan,
• Abstract summary: MoMa is a modality-aware mixture-of-experts architecture for pre-training mixed-modal, early-fusion language models. Under a 1-trillion-token training budget, the MoMa 1.4B model, featuring 4 text experts and 4 image experts, achieves impressive FLOPs savings.
• Score: 95.26323548734692
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: We introduce MoMa, a novel modality-aware mixture-of-experts (MoE) architecture designed for pre-training mixed-modal, early-fusion language models. MoMa processes images and text in arbitrary sequences by dividing expert modules into modality-specific groups. These groups exclusively process designated tokens while employing learned routing within each group to maintain semantically informed adaptivity. Our empirical results reveal substantial pre-training efficiency gains through this modality-specific parameter allocation. Under a 1-trillion-token training budget, the MoMa 1.4B model, featuring 4 text experts and 4 image experts, achieves impressive FLOPs savings: 3.7x overall, with 2.6x for text and 5.2x for image processing compared to a compute-equivalent dense baseline, measured by pre-training loss. This outperforms the standard expert-choice MoE with 8 mixed-modal experts, which achieves 3x overall FLOPs savings (3x for text, 2.8x for image). Combining MoMa with mixture-of-depths (MoD) further improves pre-training FLOPs savings to 4.2x overall (text: 3.4x, image: 5.3x), although this combination hurts performance in causal inference due to increased sensitivity to router accuracy. These results demonstrate MoMa's potential to significantly advance the efficiency of mixed-modal, early-fusion language model pre-training, paving the way for more resource-efficient and capable multimodal AI systems.

Related papers

• Mixture-of-Mamba: Enhancing Multi-Modal State-Space Models with Modality-Aware Sparsity [56.0251572416922]
  State Space Models (SSMs) have emerged as efficient alternatives to Transformers for sequential modeling. We propose a novel SSM architecture that introduces modality-aware sparsity through modality-specific parameterization of the Mamba block. We evaluate Mixture-of-Mamba across three multi-modal pretraining settings.
  arXiv  Detail & Related papers  (2025-01-27T18:35:05Z)
• FSMoE: A Flexible and Scalable Training System for Sparse Mixture-of-Experts Models [21.96960353910023]
  We introduce FSMoE, a flexible training system optimizing task scheduling with three novel techniques. We conduct extensive experiments with configured MoE layers and real-world MoE models on two GPU clusters. FSMoE supports four popular types of MoE routing functions and is more efficient than existing implementations.
  arXiv  Detail & Related papers  (2025-01-18T10:14:37Z)
• MoE++: Accelerating Mixture-of-Experts Methods with Zero-Computation Experts [63.67734699877724]
  MoE++ is a general and heterogeneous MoE framework that integrates both Feed-Forward Network(FFN) and zero-computation experts. MoE++ achieves better performance while delivering 1.1-2.1x expert forward throughput compared to a vanilla MoE model of the same size.
  arXiv  Detail & Related papers  (2024-10-09T18:01:27Z)
• MIO: A Foundation Model on Multimodal Tokens [74.85153216521945]
  We introduce MIO, a novel foundation model built on multimodal tokens. MIO is capable of understanding and generating speech, text, images, and videos in an end-to-end, autoregressive manner.
  arXiv  Detail & Related papers  (2024-09-26T09:57:16Z)
• Compact Language Models via Pruning and Knowledge Distillation [61.56557874432008]
  Minitron models exhibit up to a 16% improvement in MMLU scores compared to training from scratch. Deriving 8B and 4B models from an already pretrained 15B model using our approach requires up to 40x fewer training tokens per model compared to training from scratch.
  arXiv  Detail & Related papers  (2024-07-19T21:47:57Z)
• Expert-Token Resonance: Redefining MoE Routing through Affinity-Driven Active Selection [16.539855450082946]
  We introduce a novel framework that redefines MoE routing through affinity-driven active selection. Our theoretical analysis demonstrates that this approach mitigates expert homogenization while enabling substantial capacity boundary reduction. After supervised fine-tuning, it exhibits performance gains of 9.7% to 14.1% across GDAD, C-Eval, and TeleQnA benchmarks.
  arXiv  Detail & Related papers  (2024-05-24T02:50:44Z)
• Toward Inference-optimal Mixture-of-Expert Large Language Models [55.96674056805708]
  We study the scaling law of MoE-based large language models (LLMs) We find that MoEs with a few (4/8) experts are the most serving efficient solution under the same performance, but costs 2.5-3.5x more in training. We propose to amend the scaling law of MoE by introducing inference efficiency as another metric besides the validation loss.
  arXiv  Detail & Related papers  (2024-04-03T16:33:42Z)
• Exploiting Inter-Layer Expert Affinity for Accelerating Mixture-of-Experts Model Inference [3.217776693788795]
  We propose a lightweight optimization technique called ExFlow to largely accelerate the inference of pre-trained MoE models. By exploiting the inter-layer expert affinity, our solution can be directly applied to pre-trained MoE models without any fine-tuning or accuracy degradation. Our solution beats the cutting-edge MoE implementations with experts from 8 to 64, with up to 2.2x improvement in inference throughput.
  arXiv  Detail & Related papers  (2024-01-16T14:16:47Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.