On-demand compute reduction with stochastic wav2vec 2.0

• URL: http://arxiv.org/abs/2204.11934v1
• Date: Mon, 25 Apr 2022 19:25:46 GMT
• Title: On-demand compute reduction with stochastic wav2vec 2.0
• Authors: Apoorv Vyas, Wei-Ning Hsu, Michael Auli, Alexei Baevski
• Abstract summary: 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.
• Score: 63.22845151306881
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Squeeze and Efficient Wav2vec (SEW) is a recently proposed architecture that squeezes the input to the transformer encoder for compute efficient pre-training and inference with wav2vec 2.0 (W2V2) models. In this work, we propose stochastic compression for on-demand compute reduction for W2V2 models. As opposed to using a fixed squeeze factor, we sample it uniformly during training. We further introduce query and key-value pooling mechanisms that can be applied to each transformer layer for further compression. Our results for models pre-trained on 960h Librispeech dataset and fine-tuned on 10h of transcribed data show that using the same stochastic model, we get a smooth trade-off between word error rate (WER) and inference time with only marginal WER degradation compared to the W2V2 and SEW models trained for a specific setting. We further show that we can fine-tune the same stochastically pre-trained model to a specific configuration to recover the WER difference resulting in significant computational savings on pre-training models from scratch.

Related papers

• Token Compensator: Altering Inference Cost of Vision Transformer without Re-Tuning [63.43972993473501]
  Token compression expedites the training and inference of Vision Transformers (ViTs) However, when applied to downstream tasks, compression degrees are mismatched between training and inference stages. We propose a model arithmetic framework to decouple the compression degrees between the two stages.
  arXiv  Detail & Related papers  (2024-08-13T10:36:43Z)
• Activations and Gradients Compression for Model-Parallel Training [85.99744701008802]
  We study how simultaneous compression of activations and gradients in model-parallel distributed training setup affects convergence. We find that gradients require milder compression rates than activations. Experiments also show that models trained with TopK perform well only when compression is also applied during inference.
  arXiv  Detail & Related papers  (2024-01-15T15:54:54Z)
• MatFormer: Nested Transformer for Elastic Inference [94.1789252941718]
  MatFormer is a nested Transformer architecture designed to offer elasticity in a variety of deployment constraints. We show that a 2.6B decoder-only MatFormer language model (MatLM) allows us to extract smaller models spanning from 1.5B to 2.6B. We also observe that smaller encoders extracted from a universal MatFormer-based ViT (MatViT) encoder preserve the metric-space structure for adaptive large-scale retrieval.
  arXiv  Detail & Related papers  (2023-10-11T17:57:14Z)
• Task-Agnostic Structured Pruning of Speech Representation Models [18.555223754089905]
  We propose a fine-grained attention head pruning method to compensate for the performance degradation. Experiments on the SUPERB benchmark show that our model can achieve comparable performance to the dense model in multiple tasks.
  arXiv  Detail & Related papers  (2023-06-02T09:11:06Z)
• 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)
• Deep learning model compression using network sensitivity and gradients [3.52359746858894]
  We present model compression algorithms for both non-retraining and retraining conditions. In the first case, we propose the Bin & Quant algorithm for compression of the deep learning models using the sensitivity of the network parameters. In the second case, we propose our novel gradient-weighted k-means clustering algorithm (GWK)
  arXiv  Detail & Related papers  (2022-10-11T03:02:40Z)
• Performance-Efficiency Trade-offs in Unsupervised Pre-training for Speech Recognition [32.61769580342906]
  We focus on wav2vec 2.0, and formalize several architecture designs that influence both the model performance and its efficiency. We introduce SEW (Squeezed and Efficient Wav2vec), a pre-trained model architecture with significant improvements along both performance and efficiency dimensions.
  arXiv  Detail & Related papers  (2021-09-14T17:58:09Z)
• Wav2vec-C: A Self-supervised Model for Speech Representation Learning [40.47940210640496]
  Wav2vec-C is a representation learning technique combining elements from wav2vec 2.0 and VQ-VAE. The proposed self-supervised model is trained on 10k hours of unlabeled data and fine-tuned with 1k hours of labeled data.
  arXiv  Detail & Related papers  (2021-03-09T16:44:45Z)
• Deliberation Model Based Two-Pass End-to-End Speech Recognition [52.45841282906516]
  A two-pass model has been proposed to rescore streamed hypotheses using the non-streaming Listen, Attend and Spell (LAS) model. The model attends to acoustics to rescore hypotheses, as opposed to a class of neural correction models that use only first-pass text hypotheses. A bidirectional encoder is used to extract context information from first-pass hypotheses.
  arXiv  Detail & Related papers  (2020-03-17T22:01:12Z)

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.