Generative machine learning methods such as large-language models are
revolutionizing the creation of text and images. While these models are
powerful they also harness a large amount of computational resources. The
transformer is a key component in large language models that aims to generate a
suitable completion of a given partial sequence. In this work, we investigate
transformer architectures under the lens of fault-tolerant quantum computing.
The input model is one where pre-trained weight matrices are given as block
encodings to construct the query, key, and value matrices for the transformer.
As a first step, we show how to prepare a block encoding of the self-attention
matrix, with a row-wise application of the softmax function using the Hadamard
product. In addition, we combine quantum subroutines to construct important
building blocks in the transformer, the residual connection, layer
normalization, and the feed-forward neural network. Our subroutines prepare an
amplitude encoding of the transformer output, which can be measured to obtain a
prediction. We discuss the potential and challenges for obtaining a quantum
advantage.

生成式机器学习方法如大型语言模型正在革新文本和图像的创作。本研究通过容错性量子计算的视角来调查变压器架构。通过将预训练的权重矩阵作为区块编码来构建变压器的查询、关键和值矩阵。我们展示了如何使用 Hadamard 乘积对自注意矩阵的行逐行应用 softmax 函数来准备区块编码。此外，我们结合了量子子程序来构建变压器中的重要构件，包括残余连接、层归一化和前馈神经网络。我们的子程序准备了变压器输出的振幅编码，可以进行测量以获得预测结果。我们讨论了获取量子优势的潜力和挑战。