BitTTS：使用1.58位量化与权重索引的高紧凑度文本到语音合成
BitTTS: Highly Compact Text-to-Speech Using 1.58-bit Quantization and Weight Indexing
AI总结：
本文提出BitTTS，通过1.58位量化感知训练和权重索引方法，在端侧实现模型规模缩减83%，且合成质量优于同规模未量化基线。
摘要：
本文提出了一种面向端侧应用的高紧凑度、轻量级文本到语音（TTS）模型。为减小模型规模，所提模型引入两项技术。第一，我们引入量化感知训练（QAT），在训练过程中将模型参数量化至低至1.58位。此时，大部分32位模型参数被量化为三值{-1, 0, 1}。第二，我们提出一种名为权重索引的方法。在该方法中，我们将一组1.58位权重保存为单个int8索引。这使得即使在以8位为单位处理数值的硬件上，也能高效存储模型参数。实验结果表明，所提方法实现了83%的模型规模缩减，同时在合成质量上优于具有相似模型规模但未量化的基线模型。
英文摘要：
This paper proposes a highly compact, lightweight text-to-speech (TTS) model for on-device applications. To reduce the model size, the proposed model introduces two techniques. First, we introduce quantization-aware training (QAT), which quantizes model parameters during training to as low as 1.58-bit. In this case, most of 32-bit model parameters are quantized to ternary values {-1, 0, 1}. Second, we propose a method named weight indexing. In this method, we save a group of 1.58-bit weights as a single int8 index. This allows for efficient storage of model parameters, even on hardware that treats values in units of 8-bit. Experimental results demonstrate that the proposed method achieved 83 % reduction in model size, while outperforming the baseline of similar model size without quantization in synthesis quality.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
