arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2101.05615cs.LGcs.PF
收藏

FBGEMM:实现高性能低精度深度学习推理

FBGEMM: Enabling High-Performance Low-Precision Deep Learning Inference

Facebook, Inc.(脸书公司)

Daya Khudia, Jianyu Huang, Protonu Basu, Summer Deng, Haixin Liu, Jongsoo Park, Mikhail Smelyanskiy

摘要:

深度学习模型通常使用单精度(FP 32)浮点数据类型来表示激活和权重,但最近的一系列研究工作表明,使用降低精度的数据类型(FP 16,16位整数,8位整数甚至4位或2位整数)进行计算足以达到与FP 32相同的精度,并且效率更高。因此,我们设计了fbgemm,一个高性能的内核库,从头开始在当前一代CPU上执行高性能的量化推理。FBGEMM通过将公共量化操作与高性能GEMM实现融合以及通过在运行时的形状和大小特定的内核代码生成来实现效率。该库已在Facebook部署,相对于我们当前的生产基线,它提供了超过2倍的性能增益。

英文摘要:

Deep learning models typically use single-precision (FP32) floating point data types for representing activations and weights, but a slew of recent research work has shown that computations with reduced-precision data types (FP16, 16-bit integers, 8-bit integers or even 4- or 2-bit integers) are enough to achieve same accuracy as FP32 and are much more efficient. Therefore, we designed fbgemm, a high-performance kernel library, from ground up to perform high-performance quantized inference on current generation CPUs. fbgemm achieves efficiency by fusing common quantization operations with a high-performance gemm implementation and by shape- and size-specific kernel code generation at runtime. The library has been deployed at Facebook, where it delivers greater than 2x performance gains with respect to our current production baseline.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑