Search papers, labs, and topics across Lattice.
This paper introduces the Trajectory-based Quantization Sensitivity Score (TQS), a novel metric that assesses the stability of post-training quantization (PTQ) by modeling neural network rollouts as discrete-time dynamical systems. TQS allows for the a priori estimation of quantization sensitivity independent of quantizer selection, facilitating effective quantization budget planning for black-box networks. The proposed TQS-PTQ framework demonstrates that leveraging a dynamical-systems perspective can achieve robust low-precision deployment without requiring calibration data or complex approximations.
Quantization-induced errors can be effectively managed by a new metric that decouples sensitivity analysis from quantizer selection, paving the way for efficient low-precision model deployment.
We introduce the Trajectory-based Quantization Sensitivity Score (TQS), a metric that reframes post-training quantization (PTQ) through the lens of dynamical-systems stability. By modeling the network's rollout as a discrete-time dynamical system, TQS characterizes how quantization-induced errors propagate and amplify over the rollout horizon. Unlike conventional PTQ methods, where sensitivity analysis is often coupled to the quantization procedure, TQS enables a priori sensitivity estimation decoupled from quantizer selection and bit-width assignment. This separation allows for quantization budget planning even for black-box or compiled networks with fused operators. Building on this, we present TQS-PTQ, a flexible mixed-precision framework that requires no calibration data or costly second-order approximations. Our experiments show that a dynamical-systems perspective provides a robust, high-performing pathway for low-precision deployment in resource-constrained settings.
