Training and Inference Efficiency of Encoder-Decoder Speech Models

• URL: http://arxiv.org/abs/2503.05931v2
• Date: Wed, 19 Mar 2025 18:08:48 GMT
• Title: Training and Inference Efficiency of Encoder-Decoder Speech Models
• Authors: Piotr Żelasko, Kunal Dhawan, Daniel Galvez, Krishna C. Puvvada, Ankita Pasad, Nithin Rao Koluguri, Ke Hu, Vitaly Lavrukhin, Jagadeesh Balam, Boris Ginsburg,
• Abstract summary: We focus on the efficiency angle and ask the questions of whether we are training these speech models efficiently.<n>We show that negligence in mini-batch sampling leads to more than 50% being spent on padding.<n>We find that adjusting the model architecture to transfer model parameters from the decoder to the encoder results in a 3x inference speedup.
• Score: 25.031622057759492
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Attention encoder-decoder model architecture is the backbone of several recent top performing foundation speech models: Whisper, Seamless, OWSM, and Canary-1B. However, the reported data and compute requirements for their training are prohibitive for many in the research community. In this work, we focus on the efficiency angle and ask the questions of whether we are training these speech models efficiently, and what can we do to improve? We argue that a major, if not the most severe, detrimental factor for training efficiency is related to the sampling strategy of sequential data. We show that negligence in mini-batch sampling leads to more than 50% computation being spent on padding. To that end, we study, profile, and optimize Canary-1B training to show gradual improvement in GPU utilization leading up to 5x increase in average batch sizes versus its original training settings. This in turn allows us to train an equivalent model using 4x less GPUs in the same wall time, or leverage the original resources and train it in 2x shorter wall time. Finally, we observe that the major inference bottleneck lies in the autoregressive decoder steps. We find that adjusting the model architecture to transfer model parameters from the decoder to the encoder results in a 3x inference speedup as measured by inverse real-time factor (RTFx) while preserving the accuracy and compute requirements for convergence. The training code and models will be available as open-source.

Related papers

• 1.5-Pints Technical Report: Pretraining in Days, Not Months -- Your Language Model Thrives on Quality Data [0.0]
  This paper presents a compute-efficient approach to pre-training a Language Model-the "1.5-Pints"-in only 9 days. Based on MT-Bench (a benchmark that emulates human judgments), 1.5-Pints outperforms Apple's OpenELM and Microsoft's Phi. This is achieved by a carefully curated pre-training dataset of 57 billion tokens, using a mix of automated and manual human review.
  arXiv  Detail & Related papers  (2024-08-07T02:14:52Z)
• A Dynamical Model of Neural Scaling Laws [79.59705237659547]
  We analyze a random feature model trained with gradient descent as a solvable model of network training and generalization. Our theory shows how the gap between training and test loss can gradually build up over time due to repeated reuse of data.
  arXiv  Detail & Related papers  (2024-02-02T01:41:38Z)
• Winner-Take-All Column Row Sampling for Memory Efficient Adaptation of Language Model [89.8764435351222]
  We propose a new family of unbiased estimators called WTA-CRS, for matrix production with reduced variance. Our work provides both theoretical and experimental evidence that, in the context of tuning transformers, our proposed estimators exhibit lower variance compared to existing ones.
  arXiv  Detail & Related papers  (2023-05-24T15:52:08Z)
• LegoNet: A Fast and Exact Unlearning Architecture [59.49058450583149]
  Machine unlearning aims to erase the impact of specific training samples upon deleted requests from a trained model. We present a novel network, namely textitLegoNet, which adopts the framework of fixed encoder + multiple adapters'' We show that LegoNet accomplishes fast and exact unlearning while maintaining acceptable performance, synthetically outperforming unlearning baselines.
  arXiv  Detail & Related papers  (2022-10-28T09:53:05Z)
• On-demand compute reduction with stochastic wav2vec 2.0 [63.22845151306881]
  We propose compression for on-demand compute reduction for wav2vec 2.0 (W2V2) models. Our results for models pre-trained on 960h Librispeech dataset and fine-tuned on 10h of transcribed data show that using the same model, we get a smooth trade-off between word error rate (WER) and inference time.
  arXiv  Detail & Related papers  (2022-04-25T19:25:46Z)
• ED2LM: Encoder-Decoder to Language Model for Faster Document Re-ranking Inference [70.36083572306839]
  This paper proposes a new training and inference paradigm for re-ranking. We finetune a pretrained encoder-decoder model using in the form of document to query generation. We show that this encoder-decoder architecture can be decomposed into a decoder-only language model during inference.
  arXiv  Detail & Related papers  (2022-04-25T06:26:29Z)
• Communication-Efficient TeraByte-Scale Model Training Framework for Online Advertising [32.5337643852876]
  Click-Through Rate (CTR) prediction is a crucial component in the online advertising industry. We identify two major challenges in the existing GPU training for massivescale ad models. We propose a hardware-aware training workflow that couples the hardware topology into the algorithm design.
  arXiv  Detail & Related papers  (2022-01-05T18:09:11Z)
• The Right Tool for the Job: Matching Model and Instance Complexities [62.95183777679024]
  As NLP models become larger, executing a trained model requires significant computational resources incurring monetary and environmental costs. We propose a modification to contextual representation fine-tuning which, during inference, allows for an early (and fast) "exit" We test our proposed modification on five different datasets in two tasks: three text classification datasets and two natural language inference benchmarks.
  arXiv  Detail & Related papers  (2020-04-16T04:28:08Z)

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.