Dynamic Stripes: Exploiting the Dynamic Precision Requirements of Activation Values in Neural Networks
This work addresses efficiency issues in DNN accelerators for hardware designers, though it is incremental as it builds on the existing Stripes method.
The paper tackled the problem of inefficient fixed-precision activation values in neural network accelerators by introducing Dynamic Stripes, which dynamically adjusts precision at runtime, resulting in a 41% performance improvement over the prior Stripes accelerator.
Stripes is a Deep Neural Network (DNN) accelerator that uses bit-serial computation to offer performance that is proportional to the fixed-point precision of the activation values. The fixed-point precisions are determined a priori using profiling and are selected at a per layer granularity. This paper presents Dynamic Stripes, an extension to Stripes that detects precision variance at runtime and at a finer granularity. This extra level of precision reduction increases performance by 41% over Stripes.