Fast Convolutional Nets With fbfft: A GPU Performance Evaluation · HackerPapers discussions

Fast Convolutional Nets With fbfft: A GPU Performance Evaluation — mentioned by ajtulloch

Tue, 21 Apr 2015 01:28:42 GMT

FWIW, FFT-based convolutional methods aren't better for a lot of common sizes (including a bunch of AlexNet/VGG/GoogLeNet sizes). http://arxiv.org/abs/1412.7580 has a pretty thorough evaluation of im2col + gemm vs fused gemm (cudnn) vs fft.

Paper and discussion history. A mention is not an endorsement.

Fast Convolutional Nets With fbfft: A GPU Performance Evaluation — mentioned by ajtulloch

Fri, 16 Jan 2015 20:39:13 GMT

http://arxiv.org/pdf/1412.7580v2.pdf compares the convolutional implementation with the cuDNN layers. For the FC layers, it's just CuBLAS `sgemm`.

Paper and discussion history. A mention is not an endorsement.

Fast Convolutional Nets With fbfft: A GPU Performance Evaluation — mentioned by contingencies

Fri, 16 Jan 2015 19:37:43 GMT

Paper with proper summary: http://arxiv.org/abs/1412.7580

Paper and discussion history. A mention is not an endorsement.