Sparse Upcycling: Inference Inefficient Finetuning

• URL: http://arxiv.org/abs/2411.08968v1
• Date: Wed, 13 Nov 2024 19:02:36 GMT
• Title: Sparse Upcycling: Inference Inefficient Finetuning
• Authors: Sasha Doubov, Nikhil Sardana, Vitaliy Chiley,
• Abstract summary: We show that sparse upcycling can achieve better quality, with improvements of over 20% relative to continued pretraining (CPT) in certain scenarios. However, this comes with a significant inference cost, leading to 40% slowdowns in high-demand inference settings for larger models.
• Score: 4.988895645799531
• License:
• Abstract: Small, highly trained, open-source large language models are widely used due to their inference efficiency, but further improving their quality remains a challenge. Sparse upcycling is a promising approach that transforms a pretrained dense model into a Mixture-of-Experts (MoE) architecture, increasing the model's parameter count and quality. In this work, we compare the effectiveness of sparse upcycling against continued pretraining (CPT) across different model sizes, compute budgets, and pretraining durations. Our experiments show that sparse upcycling can achieve better quality, with improvements of over 20% relative to CPT in certain scenarios. However, this comes with a significant inference cost, leading to 40% slowdowns in high-demand inference settings for larger models. Our findings highlight the trade-off between model quality and inference efficiency, offering insights for practitioners seeking to balance model quality and deployment constraints.

Related papers

• STLM Engineering Report: Dropout [4.3600359083731695]
  We find that dropout remains effective in the overfitting scenario, and that it may have some relevance for improving the fit of models even in the case of excess data. In the process we find that the existing explanation for the mechanism behind this performance gain is not applicable in the case of language modelling.
  arXiv  Detail & Related papers  (2024-09-09T08:24:29Z)
• Co-training and Co-distillation for Quality Improvement and Compression of Language Models [88.94539115180919]
  Knowledge Distillation (KD) compresses expensive pre-trained language models (PLMs) by transferring their knowledge to smaller models. Most smaller models fail to surpass the performance of the original larger model, resulting in sacrificing performance to improve inference speed. We propose Co-Training and Co-Distillation (CTCD), a novel framework that improves performance and inference speed together by co-training two models.
  arXiv  Detail & Related papers  (2023-11-06T03:29:00Z)
• QualEval: Qualitative Evaluation for Model Improvement [82.73561470966658]
  We propose QualEval, which augments quantitative scalar metrics with automated qualitative evaluation as a vehicle for model improvement. QualEval uses a powerful LLM reasoner and our novel flexible linear programming solver to generate human-readable insights. We demonstrate that leveraging its insights, for example, improves the absolute performance of the Llama 2 model by up to 15% points relative.
  arXiv  Detail & Related papers  (2023-11-06T00:21:44Z)
• A Multi-dimensional Evaluation of Tokenizer-free Multilingual Pretrained Models [87.7086269902562]
  We show that subword-based models might still be the most practical choice in many settings. We encourage future work in tokenizer-free methods to consider these factors when designing and evaluating new models.
  arXiv  Detail & Related papers  (2022-10-13T15:47:09Z)
• Feeding What You Need by Understanding What You Learned [54.400455868448695]
  Machine Reading (MRC) reveals the ability to understand a given text passage and answer questions based on it. Existing research works in MRC rely heavily on large-size models and corpus to improve the performance evaluated by metrics such as Exact Match. We argue that a deep understanding of model capabilities and data properties can help us feed a model with appropriate training data.
  arXiv  Detail & Related papers  (2022-03-05T14:15:59Z)
• Knowledge Distillation for Quality Estimation [79.51452598302934]
  Quality Estimation (QE) is the task of automatically predicting Machine Translation quality in the absence of reference translations. Recent success in QE stems from the use of multilingual pre-trained representations, where very large models lead to impressive results. We show that this approach, in combination with data augmentation, leads to light-weight QE models that perform competitively with distilled pre-trained representations with 8x fewer parameters.
  arXiv  Detail & Related papers  (2021-07-01T12:36:21Z)
• MixKD: Towards Efficient Distillation of Large-scale Language Models [129.73786264834894]
  We propose MixKD, a data-agnostic distillation framework, to endow the resulting model with stronger generalization ability. We prove from a theoretical perspective that under reasonable conditions MixKD gives rise to a smaller gap between the error and the empirical error. Experiments under a limited-data setting and ablation studies further demonstrate the advantages of the proposed approach.
  arXiv  Detail & Related papers  (2020-11-01T18:47:51Z)

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.