| Presentation | 2021-03-25 Parallelization and Vectorization of SpMM for Sparse Neural Network Yuta Tadokoro, Keiji Kimura, Hironori Kasahara, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | Pruning is one of the well-known model compression techniques in Deep Learning. Eliminating less important weights in the model provides a smaller model size than the original one while keeping high accuracy. As a result of the pruning, the weight matrices are represented as sparse matrices. However, the sparse matrices obtained by pruningare highly randomized, unlike the sparse matrices used in scientific applications. Thus it is difficult to employ acceleration techniques for them relying on the locality of non-zero elements. This paper proposes a method to accelerate SpMM (Sparse Matrix - Dense Matrix Multiplication) for sparse matrices with high randomness. The proposed method is applied to ResNet50 and evaluated on NEC SX-Aurora TSUBASA. The speed-ups were 2.78 times with one processor core for the layer to which the proposed method was used and 1.98 times with eight processor cores for the whole model. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | SpMM / Sparse Neural Network / Vector Processor |
| Paper # | CPSY2020-55,DC2020-85 |
| Date of Issue | 2021-03-18 (CPSY, DC) |
| Conference Information | |
| Committee | CPSY / DC / IPSJ-SLDM / IPSJ-EMB / IPSJ-ARC |
|---|---|
| Conference Date | 2021/3/25(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Online |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | ETNET2021 |
| Chair | Hidetsugu Irie(Univ. of Tokyo) / Hiroshi Takahashi(Ehime Univ.) / Yuichi Nakamura(NEC) / / Hiroshi Inoue(Kyushu Univ.) |
| Vice Chair | Michihiro Koibuchi(NII) / Kota Nakajima(Fujitsu Lab.) / Tatsuhiro Tsuchiya(Osaka Univ.) |
| Secretary | Michihiro Koibuchi(Univ. of Tokyo) / Kota Nakajima(Nagoya Inst. of Tech.) / Tatsuhiro Tsuchiya(Nihon Univ.) / (Chiba Univ.) / (Tokyo City Univ.) / (Kochi Univ. of Tech.) |
| Assistant | Shugo Ogawa(Hitachi) / Eiji Arima(Univ. of Tokyo) |
| Paper Information | |
| Registration To | Technical Committee on Computer Systems / Technical Committee on Dependable Computing / Special Interest Group on System and LSI Design Methodology / Special Interest Group on Embedded Systems / Special Interest Group on System Architecture |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Parallelization and Vectorization of SpMM for Sparse Neural Network |
| Sub Title (in English) | |
| Keyword(1) | SpMM |
| Keyword(2) | Sparse Neural Network |
| Keyword(3) | Vector Processor |
| 1st Author's Name | Yuta Tadokoro |
| 1st Author's Affiliation | Waseda University(Waseda Univ.) |
| 2nd Author's Name | Keiji Kimura |
| 2nd Author's Affiliation | Waseda University(Waseda Univ.) |
| 3rd Author's Name | Hironori Kasahara |
| 3rd Author's Affiliation | Waseda University(Waseda Univ.) |
| Date | 2021-03-25 |
| Paper # | CPSY2020-55,DC2020-85 |
| Volume (vol) | vol.120 |
| Number (no) | CPSY-435,DC-436 |
| Page | pp.pp.31-36(CPSY), pp.31-36(DC), |
| #Pages | 6 |
| Date of Issue | 2021-03-18 (CPSY, DC) |