LEAPER: Fast and Accurate FPGA-based System Performance Prediction via Transfer Learning
This addresses the problem of high total-cost-ownership and time inefficiency in FPGA-based system design for engineers and cloud providers, though it is incremental as it builds on existing transfer learning methods.
The paper tackles the slow and costly process of training ML models for FPGA accelerator performance prediction by proposing LEAPER, a transfer learning approach that transfers models from low-end edge to high-end cloud environments, achieving 85% accuracy on average and reducing design-space exploration time by 10x.
Machine learning has recently gained traction as a way to overcome the slow accelerator generation and implementation process on an FPGA. It can be used to build performance and resource usage models that enable fast early-stage design space exploration. First, training requires large amounts of data (features extracted from design synthesis and implementation tools), which is cost-inefficient because of the time-consuming accelerator design and implementation process. Second, a model trained for a specific environment cannot predict performance or resource usage for a new, unknown environment. In a cloud system, renting a platform for data collection to build an ML model can significantly increase the total-cost-ownership (TCO) of a system. Third, ML-based models trained using a limited number of samples are prone to overfitting. To overcome these limitations, we propose LEAPER, a transfer learning-based approach for prediction of performance and resource usage in FPGA-based systems. The key idea of LEAPER is to transfer an ML-based performance and resource usage model trained for a low-end edge environment to a new, high-end cloud environment to provide fast and accurate predictions for accelerator implementation. Experimental results show that LEAPER (1) provides, on average across six workloads and five FPGAs, 85% accuracy when we use our transferred model for prediction in a cloud environment with 5-shot learning and (2) reduces design-space exploration time for accelerator implementation on an FPGA by 10x, from days to only a few hours.