Commit message (Expand) | Author | Age | |
---|---|---|---|
* | Add more logging to the convolution transformations. | Tim Shen | 2018-10-08 |
* | Add cuDNN fused convolution forward support. | Tim Shen | 2018-09-24 |
* | Remove the public uses of CreateCudnnConv* in the favor of | Tim Shen | 2018-09-24 |
* | Fix a typo in cudnn_convolution_rewriter. | Adrian Kuegel | 2018-09-11 |
* | [XLA:GPU] Don't canonicalize forward convs with constant filters to backwards... | Justin Lebar | 2018-09-10 |
* | Enable grouped convolutions for CudnnConvBackwardInput. | Adrian Kuegel | 2018-09-10 |
* | Call Cudnn also for grouped convolutions. | Adrian Kuegel | 2018-09-03 |
* | [XLA:GPU] Create backwards-input convs when the filter is a constant. | Justin Lebar | 2018-08-28 |
* | [TF:XLA] Split literal_util into {literal, literal_util}. | Kay Zhu | 2018-07-03 |
* | Fix assumptions that a Shape must be a tuple or an array. | Mark Heffernan | 2018-06-13 |
* | Internal change | Justin Lebar | 2018-02-02 |