FIXAR: A Fixed-Point Deep Reinforcement Learning Platform with Quantization-Aware Training and Adaptive Parallelism
TimeTuesday, December 7th2:15pm - 2:38pm PST
Event Type
Research Manuscript
Virtual Programs
Presented In-Person
AI/ML System Design
DescriptionIn this paper, we present a deep reinforcement learning platform named FIXAR which employs fixed-point data types and arithmetic units for the first time using a SW/HW codesign approach. Starting from 32-bit fixed-point data, Quantization-Aware Training (QAT) reduces its data precision based on the range of activations and performs retraining to minimize the reward degradation. FIXAR proposes the adaptive array engine composed of 16x16 configurable processing elements to support both intra-layer parallelism and intra-batch parallelism for high-throughput inference and training. Finally, FIXAR was implemented on Xilinx U50 and achieves 2638.0 IPS/W, which is 15.4 times more energy efficient than GPU.