FetchFortune
Abstract
It is widely acknowledged that the performance of Transformer models is exponentially related to their number of parameters and computational complexity. While approaches like Mixture of Experts (MoE) decouple parameter count from computational complexity, they still face challenges in inference due to high memory access costs. This work introduces UltraMem, incorporating large-scale, ultra-sparse memory layer to address these limitations. Our approach significantly reduces inference latency while maintaining model performance. We also investigate the scaling laws of this new architecture, demonstrating that it not only exhibits favorable scaling properties but outperforms traditional models. In our experiments, we train networks with up to 20 million memory slots. The results show that our method achieves state-of-the-art inference speed and model performance within a given computational budget.
Community
TL;DR: We propose UltraMem, a model that significantly accelerates inference speeds while maintaining comparable performance to Mixture of Experts (MoE). This improvement is primarily attributed to the substantially reduced memory access during inference compared to MoE. Furthermore, by increasing the number of sparse parameters while keeping the activated parameters constant, UltraMem ensures that the inference speed does not significantly increase.
This is an automated message from the Librarian Bot. I found the following papers similar to this paper.
The following papers were recommended by the Semantic Scholar API
- Sparse Upcycling: Inference Inefficient Finetuning (2024)
- FluidML: Fast and Memory Efficient Inference Optimization (2024)
- Beyond Linear Approximations: A Novel Pruning Approach for Attention Matrix (2024)
- Communication Compression for Tensor Parallel LLM Inference (2024)
- ED-ViT: Splitting Vision Transformer for Distributed Inference on Edge Devices (2024)
- SLiM: One-shot Quantized Sparse Plus Low-rank Approximation of LLMs (2024)
- Learning Parameter Sharing with Tensor Decompositions and Sparsity (2024)
Please give a thumbs up to this comment if you found it helpful!
If you want recommendations for any Paper on Hugging Face checkout this Space
You can directly ask Librarian Bot for paper recommendations by tagging it in a comment:
@librarian-bot
recommend
Models citing this paper 0
No model linking this paper
Datasets citing this paper 0
No dataset linking this paper
Spaces citing this paper 0
No Space linking this paper