Summary of Disha: Dimension-sharding Adaptation Of Large Language Models with Fast Convergence and Fast Computation, by Jiale Kang
DiSHA: Dimension-Sharding Adaptation of Large Language Models with Fast Convergence and Fast Computation
by Jiale Kang
First submitted to arxiv on: 19 Sep 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper introduces Dimension-Sharding Adaptation (DiSHA), a technique that expands the Parameter-Efficient Fine-Tuning (PEFT) framework to reduce computational burden and enable resource-constrained fine-tuning of Large Language Models (LLMs). DiSHA includes Block Affine Efficient Computation (Bone) and Block Affine Transformation (Bat) to induce nonlinearity in matrix updates without adding parameters. The authors show that Bone outperforms LoRA variants in Natural Language Understanding and Generation tasks, with improved computational efficiency. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper makes large language models more efficient and adaptable for new tasks. It creates a new way of fine-tuning these models called Dimension-Sharding Adaptation (DiSHA). This approach reduces the amount of calculations needed to update the model’s weights. The authors also introduce two new techniques, Block Affine Efficient Computation (Bone) and Block Affine Transformation (Bat), which help the model learn more efficiently. These innovations lead to better performance on natural language tasks like understanding and generating text. |
Keywords
» Artificial intelligence » Fine tuning » Language understanding » Lora » Parameter efficient
