Folder structure:

(1) anonymous.pdf: the paper in the submission with the following citation,
	Anonymous. Large-Scale Recurrent Neural Networks with Fully Homomorphic Encryption for
	  Privacy-Enhanced Speaker Identification. In Submission to ICASSP 2025, 2025. Paper is currently
	  under review for ICASSP 2025, is by the authors of this paper, and is included in the accompanying
	  ZIP file as ”anonymous.pdf” without references to the identity of the authors.

(2) overflow-aware-activity-regularization: the main code for training, quantizing, and extracting plaintext models, as well all plaintext evaluation code in section 4.1.

(3) tfhe_mnist_rnns: the main code for evaluating the models over encrypted data.

