Modeling FPGA-Based Systems via Few-Shot Learning

Abstract

Machine-learning-based models have recently gained traction as a way to overcome the slow downstream implementation process of FPGAs by building models that provide fast and accurate performance predictions. However, these models suffer from two main limitations: (1) training requires large amounts of data (features extracted from FPGA synthesis and implementation reports), which is costinefficient because of the time-consuming FPGA design cycle;(2) a model trained for a specific environment cannot predict for a new, unknown environment. In a cloud system, where getting access to platforms is typically costly, data collection for ML models can significantly increase the total-costownership (TCO) of a system. To overcome these limitations, we propose LEAPER, a transfer learning-based approach for FPGA-based systems that adapts an existing ML-based model to a new, unknown environment to provide fast and accurate performance and resource utilization predictions. Experimental results show that our approach delivers, on average, 85% accuracy when we use our transferred model for prediction in a cloud environment with 5-shot learning and reduces design-space exploration time by 10×, from days to only a few hours.

Accelerating Weather Prediction Using Near-Memory Reconfigurable Fabric

“…Eighth, we are witnessing an enormous amount of data being generated across multiple application domains [123,151] like weather prediction modeling, radio astronomy, bioinformatics, material science, chemistry, health sciences, etc. The processing of the sheer amount of generated data is one of the biggest challenges to overcome.…”

“…Hence, instead of pruning the design space manually, we formulate the search for the best window size as a multi-objective auto-tuning problem taking into account the datatype precision. We make use of OpenTuner [34], which uses machine-learning techniques to guide the design-space exploration [151].…”