Kai Zhen

Amazon AGI, Aaron Langford, Aayush Shah et al. · amazon-science

We present Amazon Nova, a new generation of state-of-the-art foundation models that deliver frontier intelligence and industry-leading price performance. Amazon Nova Pro is a highly-capable multimodal model with the best combination of accuracy, speed, and cost for a wide range of tasks. Amazon Nova Lite is a low-cost multimodal model that is lightning fast for processing images, video, documents and text. Amazon Nova Micro is a text-only model that delivers our lowest-latency responses at very low cost. Amazon Nova Canvas is an image generation model that creates professional grade images with rich customization controls. Amazon Nova Reel is a video generation model offering high-quality outputs, customization, and motion control. Our models were built responsibly and with a commitment to customer trust, security, and reliability. We report benchmarking results for core capabilities, agentic performance, long context, functional adaptation, runtime performance, and human evaluation.

Kai Zhen, Hieu Duy Nguyen, Raviteja Chinta et al.

We present a novel sub-8-bit quantization-aware training (S8BQAT) scheme for 8-bit neural network accelerators. Our method is inspired from Lloyd-Max compression theory with practical adaptations for a feasible computational overhead during training. With the quantization centroids derived from a 32-bit baseline, we augment training loss with a Multi-Regional Absolute Cosine (MRACos) regularizer that aggregates weights towards their nearest centroid, effectively acting as a pseudo compressor. Additionally, a periodically invoked hard compressor is introduced to improve the convergence rate by emulating runtime model weight quantization. We apply S8BQAT on speech recognition tasks using Recurrent Neural NetworkTransducer (RNN-T) architecture. With S8BQAT, we are able to increase the model parameter size to reduce the word error rate by 4-16% relatively, while still improving latency by 5%.

Kai Zhen, Martin Radfar, Hieu Duy Nguyen et al.

For on-device automatic speech recognition (ASR), quantization aware training (QAT) is ubiquitous to achieve the trade-off between model predictive performance and efficiency. Among existing QAT methods, one major drawback is that the quantization centroids have to be predetermined and fixed. To overcome this limitation, we introduce a regularization-free, "soft-to-hard" compression mechanism with self-adjustable centroids in a mu-Law constrained space, resulting in a simpler yet more versatile quantization scheme, called General Quantizer (GQ). We apply GQ to ASR tasks using Recurrent Neural Network Transducer (RNN-T) and Conformer architectures on both LibriSpeech and de-identified far-field datasets. Without accuracy degradation, GQ can compress both RNN-T and Conformer into sub-8-bit, and for some RNN-T layers, to 1-bit for fast and accurate inference. We observe a 30.73% memory footprint saving and 31.75% user-perceived latency reduction compared to 8-bit QAT via physical device benchmarking.

Yifan Yang, Kai Zhen, Bhavana Ganesh et al.

Large Language Models (LLMs) pruning seeks to remove unimportant weights for inference speedup with minimal accuracy impact. However, existing methods often suffer from accuracy degradation without full-model sparsity-aware fine-tuning. This paper presents Wanda++, a novel pruning framework that outperforms the state-of-the-art methods by utilizing decoder-block-level \textbf{regional} gradients. Specifically, Wanda++ improves the pruning score with regional gradients for the first time and proposes an efficient regional optimization method to minimize pruning-induced output discrepancies between the dense and sparse decoder output. Notably, Wanda++ improves perplexity by up to 32\% over Wanda in the language modeling task and generalizes effectively to downstream tasks. Moreover, despite updating weights with regional optimization, Wanda++ remains orthogonal to sparsity-aware fine-tuning, further reducing perplexity with LoRA in great extend. Our approach is lightweight, pruning a 7B LLaMA model in under 10 minutes on a single H100 GPU.

Jiajun Zhou, Yifan Yang, Kai Zhen et al.

Language Models (LLMs) are often quantized to lower precision to reduce the memory cost and latency in inference. However, quantization often degrades model performance, thus fine-tuning is required for various down-stream tasks. Traditional fine-tuning methods such as stochastic gradient descent and Adam optimization require backpropagation, which are error-prone in the low-precision settings. To overcome these limitations, we propose the Quantized Zeroth-Order (QuZO) framework, specifically designed for fine-tuning LLMs through low-precision (e.g., 4- or 8-bit) forward passes. Our method can avoid the error-prone low-precision straight-through estimator, and utilizes optimized stochastic rounding to mitigate the increased bias. QuZO simplifies the training process, while achieving results comparable to first-order methods in ${\rm FP}8$ and superior accuracy in ${\rm INT}8$ and ${\rm INT}4$ training. Experiments demonstrate that low-bit training QuZO achieves performance comparable to MeZO optimization on GLUE, Multi-Choice, and Generation tasks, while reducing memory cost by $2.94 \times$ in LLaMA2-7B fine-tuning compared to quantized first-order methods.

Zhen Zhang, Yifan Yang, Kai Zhen et al.

Large language models have demonstrated exceptional capabilities across diverse tasks, but their fine-tuning demands significant memory, posing challenges for resource-constrained environments. Zeroth-order (ZO) optimization provides a memory-efficient alternative by eliminating the need for backpropagation. However, ZO optimization suffers from high gradient variance, and prior research has largely focused on single-task learning, leaving its application to multi-task learning unexplored. Multi-task learning is crucial for leveraging shared knowledge across tasks to improve generalization, yet it introduces unique challenges under ZO settings, such as amplified gradient variance and collinearity. In this paper, we present MaZO, the first framework specifically designed for multi-task LLM fine-tuning under ZO optimization. MaZO tackles these challenges at the parameter level through two key innovations: a weight importance metric to identify critical parameters and a multi-task weight update mask to selectively update these parameters, reducing the dimensionality of the parameter space and mitigating task conflicts. Experiments demonstrate that MaZO achieves state-of-the-art performance, surpassing even multi-task learning methods designed for first-order optimization.

Ryan Solgi, Kai Zhen, Rupak Vignesh Swaminathan et al.

The efficient implementation of large language models (LLMs) is crucial for deployment on resource-constrained devices. Low-rank tensor compression techniques, such as tensor-train (TT) networks, have been widely studied for over-parameterized neural networks. However, their applications to compress pre-trained large language models (LLMs) for downstream tasks (post-training) remains challenging due to the high-rank nature of pre-trained LLMs and the lack of access to pretraining data. In this study, we investigate low-rank tensorized LLMs during fine-tuning and propose sparse augmented tensor networks (Saten) to enhance their performance. The proposed Saten framework enables full model compression. Experimental results demonstrate that Saten enhances both accuracy and compression efficiency in tensorized language models, achieving state-of-the-art performance.

Yifan Yang, Kai Zhen, Ershad Banijamal et al.

Fine-tuning large language models (LLMs) has achieved remarkable performance across various natural language processing tasks, yet it demands more and more memory as model sizes keep growing. To address this issue, the recently proposed Memory-efficient Zeroth-order (MeZO) methods attempt to fine-tune LLMs using only forward passes, thereby avoiding the need for a backpropagation graph. However, significant performance drops and a high risk of divergence have limited their widespread adoption. In this paper, we propose the Adaptive Zeroth-order Tensor-Train Adaption (AdaZeta) framework, specifically designed to improve the performance and convergence of the ZO methods. To enhance dimension-dependent ZO estimation accuracy, we introduce a fast-forward, low-parameter tensorized adapter. To tackle the frequently observed divergence issue in large-scale ZO fine-tuning tasks, we propose an adaptive query number schedule that guarantees convergence. Detailed theoretical analysis and extensive experimental results on Roberta-Large and Llama-2-7B models substantiate the efficacy of our AdaZeta framework in terms of accuracy, memory efficiency, and convergence speed.

Kai Zhen, Jongmo Sung, Mi Suk Lee et al.

We present a scalable and efficient neural waveform coding system for speech compression. We formulate the speech coding problem as an autoencoding task, where a convolutional neural network (CNN) performs encoding and decoding as a neural waveform codec (NWC) during its feedforward routine. The proposed NWC also defines quantization and entropy coding as a trainable module, so the coding artifacts and bitrate control are handled during the optimization process. We achieve efficiency by introducing compact model components to NWC, such as gated residual networks and depthwise separable convolution. Furthermore, the proposed models are with a scalable architecture, cross-module residual learning (CMRL), to cover a wide range of bitrates. To this end, we employ the residual coding concept to concatenate multiple NWC autoencoding modules, where each NWC module performs residual coding to restore any reconstruction loss that its preceding modules have created. CMRL can scale down to cover lower bitrates as well, for which it employs linear predictive coding (LPC) module as its first autoencoder. The hybrid design integrates LPC and NWC by redefining LPC's quantization as a differentiable process, making the system training an end-to-end manner. The decoder of proposed system is with either one NWC (0.12 million parameters) in low to medium bitrate ranges (12 to 20 kbps) or two NWCs in the high bitrate (32 kbps). Although the decoding complexity is not yet as low as that of conventional speech codecs, it is significantly reduced from that of other neural speech coders, such as a WaveNet-based vocoder. For wide-band speech coding quality, our system yields comparable or superior performance to AMR-WB and Opus on TIMIT test utterances at low and medium bitrates. The proposed system can scale up to higher bitrates to achieve near transparent performance.

Kai Zhen, Hieu Duy Nguyen, Feng-Ju Chang et al.

In order to achieve high accuracy for machine learning (ML) applications, it is essential to employ models with a large number of parameters. Certain applications, such as Automatic Speech Recognition (ASR), however, require real-time interactions with users, hence compelling the model to have as low latency as possible. Deploying large scale ML applications thus necessitates model quantization and compression, especially when running ML models on resource constrained devices. For example, by forcing some of the model weight values into zero, it is possible to apply zero-weight compression, which reduces both the model size and model reading time from the memory. In the literature, such methods are referred to as sparse pruning. The fundamental questions are when and which weights should be forced to zero, i.e. be pruned. In this work, we propose a compressed sensing based pruning (CSP) approach to effectively address those questions. By reformulating sparse pruning as a sparsity inducing and compression-error reduction dual problem, we introduce the classic compressed sensing process into the ML model training process. Using ASR task as an example, we show that CSP consistently outperforms existing approaches in the literature.

Kai Zhen, Mi Suk Lee, Jongmo Sung et al.

Conventional audio coding technologies commonly leverage human perception of sound, or psychoacoustics, to reduce the bitrate while preserving the perceptual quality of the decoded audio signals. For neural audio codecs, however, the objective nature of the loss function usually leads to suboptimal sound quality as well as high run-time complexity due to the large model size. In this work, we present a psychoacoustic calibration scheme to re-define the loss functions of neural audio coding systems so that it can decode signals more perceptually similar to the reference, yet with a much lower model complexity. The proposed loss function incorporates the global masking threshold, allowing the reconstruction error that corresponds to inaudible artifacts. Experimental results show that the proposed model outperforms the baseline neural codec twice as large and consuming 23.4% more bits per second. With the proposed method, a lightweight neural codec, with only 0.9 million parameters, performs near-transparent audio coding comparable with the commercial MPEG-1 Audio Layer III codec at 112 kbps.

Kai Zhen, Mi Suk Lee, Jongmo Sung et al.

Scalability and efficiency are desired in neural speech codecs, which supports a wide range of bitrates for applications on various devices. We propose a collaborative quantization (CQ) scheme to jointly learn the codebook of LPC coefficients and the corresponding residuals. CQ does not simply shoehorn LPC to a neural network, but bridges the computational capacity of advanced neural network models and traditional, yet efficient and domain-specific digital signal processing methods in an integrated manner. We demonstrate that CQ achieves much higher quality than its predecessor at 9 kbps with even lower model complexity. We also show that CQ can scale up to 24 kbps where it outperforms AMR-WB and Opus. As a neural waveform codec, CQ models are with less than 1 million parameters, significantly less than many other generative models.

Kai Zhen, Mi Suk Lee, Minje Kim

In speech enhancement, an end-to-end deep neural network converts a noisy speech signal to a clean speech directly in time domain without time-frequency transformation or mask estimation. However, aggregating contextual information from a high-resolution time domain signal with an affordable model complexity still remains challenging. In this paper, we propose a densely connected convolutional and recurrent network (DCCRN), a hybrid architecture, to enable dual-staged temporal context aggregation. With the dense connectivity and cross-component identical shortcut, DCCRN consistently outperforms competing convolutional baselines with an average STOI improvement of 0.23 and PESQ of 1.38 at three SNR levels. The proposed method is computationally efficient with only 1.38 million parameters. The generalizability performance on the unseen noise types is still decent considering its low complexity, although it is relatively weaker comparing to Wave-U-Net with 7.25 times more parameters.

Kai Zhen, Jongmo Sung, Mi Suk Lee et al.

Speech codecs learn compact representations of speech signals to facilitate data transmission. Many recent deep neural network (DNN) based end-to-end speech codecs achieve low bitrates and high perceptual quality at the cost of model complexity. We propose a cross-module residual learning (CMRL) pipeline as a module carrier with each module reconstructing the residual from its preceding modules. CMRL differs from other DNN-based speech codecs, in that rather than modeling speech compression problem in a single large neural network, it optimizes a series of less-complicated modules in a two-phase training scheme. The proposed method shows better objective performance than AMR-WB and the state-of-the-art DNN-based speech codec with a similar network architecture. As an end-to-end model, it takes raw PCM signals as an input, but is also compatible with linear predictive coding (LPC), showing better subjective quality at high bitrates than AMR-WB and OPUS. The gain is achieved by using only 0.9 million trainable parameters, a significantly less complex architecture than the other DNN-based codecs in the literature.

Kai Zhen, Aswin Sivaraman, Jongmo Sung et al.

We present a psychoacoustically enhanced cost function to balance network complexity and perceptual performance of deep neural networks for speech denoising. While training the network, we utilize perceptual weights added to the ordinary mean-squared error to emphasize contribution from frequency bins which are most audible while ignoring error from inaudible bins. To generate the weights, we employ psychoacoustic models to compute the global masking threshold from the clean speech spectra. We then evaluate the speech denoising performance of our perceptually guided neural network by using both objective and perceptual sound quality metrics, testing on various network structures ranging from shallow and narrow ones to deep and wide ones. The experimental results showcase our method as a valid approach for infusing perceptual significance to deep neural network operations. In particular, the more perceptually sensible enhancement in performance seen by simple neural network topologies proves that the proposed method can lead to resource-efficient speech denoising implementations in small devices without degrading the perceived signal fidelity.

Kai Zhen, Mridul Birla, David Crandall et al.

Given the progress in image recognition with recent data driven paradigms, it's still expensive to manually label a large training data to fit a convolutional neural network (CNN) model. This paper proposes a hybrid supervised-unsupervised method combining a pre-trained AlexNet with Latent Dirichlet Allocation (LDA) to extract image topics from both an unlabeled life-logging dataset and the COCO dataset. We generate the bag-of-words representations of an egocentric dataset from the softmax layer of AlexNet and use LDA to visualize the subject's living genre with duplicated images. We use a subset of COCO on 4 categories as ground truth, and define consistent rate to quantitatively analyze the performance of the method, it achieves 84% for consistent rate on average comparing to 18.75% from a raw CNN model. The method is capable of detecting false labels and multi-labels from COCO dataset. For scalability test, parallelization experiments are conducted with Harp-LDA on a Intel Knights Landing cluster: to extract 1,000 topic assignments for 241,035 COCO images, it takes 10 minutes with 60 threads.