Leveraging the true depth of LLMs

• URL: http://arxiv.org/abs/2502.02790v1
• Date: Wed, 05 Feb 2025 00:26:27 GMT
• Title: Leveraging the true depth of LLMs
• Authors: Ramón Calvo González, Daniele Paliotta, Matteo Pagliardini, Martin Jaggi, François Fleuret,
• Abstract summary: Large Language Models demonstrate remarkable capabilities at the cost of high compute requirements. We investigate several potential ways to reduce the depth of pre-trained LLMs without significantly affecting performance. We present a novel approach that exploits this decoupling between layers by grouping some of them into pairs that can be evaluated in parallel.
• Score: 46.81174316936993
• License:
• Abstract: Large Language Models demonstrate remarkable capabilities at the cost of high compute requirements. While recent research has shown that intermediate layers can be removed or have their order shuffled without impacting performance significantly, these findings have not been employed to reduce the computational cost of inference. We investigate several potential ways to reduce the depth of pre-trained LLMs without significantly affecting performance. Leveraging our insights, we present a novel approach that exploits this decoupling between layers by grouping some of them into pairs that can be evaluated in parallel. This modification of the computational graph -- through better parallelism -- results in an average improvement of around 1.20x on the number of tokens generated per second, without re-training nor fine-tuning, while retaining 95%-99% of the original accuracy. Empirical evaluation demonstrates that this approach significantly improves serving efficiency while maintaining model performance, offering a practical improvement for large-scale LLM deployment.

Related papers

• Dynamic Noise Preference Optimization for LLM Self-Improvement via Synthetic Data [51.62162460809116]
  We introduce Dynamic Noise Preference Optimization (DNPO) to ensure consistent improvements across iterations. In experiments with Zephyr-7B, DNPO consistently outperforms existing methods, showing an average performance boost of 2.6%. DNPO shows a significant improvement in model-generated data quality, with a 29.4% win-loss rate gap compared to the baseline in GPT-4 evaluations.
  arXiv  Detail & Related papers  (2025-02-08T01:20:09Z)
• The Journey Matters: Average Parameter Count over Pre-training Unifies Sparse and Dense Scaling Laws [51.608402959163925]
  We present the first systematic exploration of optimal sparse pre-training configurations for large language models. We find that initiating pruning at 25% of total training compute and concluding at 75% achieves near-optimal final evaluation loss. We propose a new scaling law that modifies the Chinchilla scaling law to use the average parameter count over pre-training.
  arXiv  Detail & Related papers  (2025-01-21T20:23:22Z)
• On Effects of Steering Latent Representation for Large Language Model Unlearning [4.058064008234271]
  Representation Misdirection for Unlearning (RMU) is an effective method for large language model (LLM) unlearning. We show that steering forget representations in the intermediate layer reduces token confidence, causing LLMs to generate wrong or nonsense responses. We propose Adaptive RMU--a simple yet effective alternative method that makes unlearning effective with most layers.
  arXiv  Detail & Related papers  (2024-08-12T15:24:50Z)
• Pruning Large Language Models with Semi-Structural Adaptive Sparse Training [17.381160429641316]
  Adaptive Sparse Trainer (AST) is a novel and efficient retraining framework tailored for semi-structured sparse models. AST reduces the perplexity and zero-shot accuracy gap between dense and 2:4 semi-structured sparse models to 0.6 and 1.16%, respectively.
  arXiv  Detail & Related papers  (2024-07-30T06:33:44Z)
• Accelerating LLaMA Inference by Enabling Intermediate Layer Decoding via Instruction Tuning with LITE [62.13435256279566]
  Large Language Models (LLMs) have achieved remarkable performance across a wide variety of natural language tasks. However, their large size makes their inference slow and computationally expensive. We show that it enables these layers to acquire 'good' generation ability without affecting the generation ability of the final layer.
  arXiv  Detail & Related papers  (2023-10-28T04:07:58Z)
• Towards Compute-Optimal Transfer Learning [82.88829463290041]
  We argue that zero-shot structured pruning of pretrained models allows them to increase compute efficiency with minimal reduction in performance. Our results show that pruning convolutional filters of pretrained models can lead to more than 20% performance improvement in low computational regimes.
  arXiv  Detail & Related papers  (2023-04-25T21:49:09Z)
• Extrapolation for Large-batch Training in Deep Learning [72.61259487233214]
  We show that a host of variations can be covered in a unified framework that we propose. We prove the convergence of this novel scheme and rigorously evaluate its empirical performance on ResNet, LSTM, and Transformer.
  arXiv  Detail & Related papers  (2020-06-10T08:22:41Z)

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.