Spartus: A 9.4 TOp/s FPGA-based LSTM Accelerator Exploiting Spatio-Temporal Sparsity
| Autor: | Gao, Chang, Delbruck, Tobi, Liu, Shih-Chii |
|---|---|
| Rok vydání: | 2021 |
| Předmět: | |
| Zdroj: | IEEE Transactions on Neural Networks and Learning Systems, 2022 |
| Druh dokumentu: | Working Paper |
| DOI: | 10.1109/TNNLS.2022.3180209 |
| Popis: | Long Short-Term Memory (LSTM) recurrent networks are frequently used for tasks involving time-sequential data such as speech recognition. Unlike previous LSTM accelerators that either exploit spatial weight sparsity or temporal activation sparsity, this paper proposes a new accelerator called "Spartus" that exploits spatio-temporal sparsity to achieve ultra-low latency inference. Spatial sparsity is induced using a new Column-Balanced Targeted Dropout (CBTD) structured pruning method, producing structured sparse weight matrices for a balanced workload. The pruned networks running on Spartus hardware achieve weight sparsity levels of up to 96% and 94% with negligible accuracy loss on the TIMIT and the Librispeech datasets. To induce temporal sparsity in LSTM, we extend the previous DeltaGRU method to the DeltaLSTM method. Combining spatio-temporal sparsity with CBTD and DeltaLSTM saves on weight memory access and associated arithmetic operations. The Spartus architecture is scalable and supports real-time online speech recognition when implemented on small and large FPGAs. Spartus per-sample latency for a single DeltaLSTM layer of 1024 neurons averages 1 us. Exploiting spatio-temporal sparsity on our test LSTM network using the TIMIT dataset leads to 46X speedup of Spartus over its theoretical hardware performance to achieve 9.4 TOp/s effective batch-1 throughput and 1.1 TOp/s/W power efficiency. Comment: Accepted for publication in IEEE Transactions on Neural Networks and Learning Systems, 2022 |
| Databáze: | arXiv |
| Externí odkaz: |
|