From 10d6b3b3fa594b9aca4a72f002acea9f927f9c60 Mon Sep 17 00:00:00 2001 From: Raul Farkas Date: Mon, 30 Jan 2023 12:58:46 +0000 Subject: MLBEDSW-7283: Add opt cases for strided CONV2D * Implement a general optimization solution for strided CONV2D that supports a stride_w with no upper bound. * Implement filter zero padding to allow for optimization in those cases in which the filter width is not divisible by the stride width. E.g.: Filter width = 8, stride width = 3 -> Filter width = 8 + 1 (0 padding) = 9, stride width = 3 * Implement partial optimization to reduce the stride to hw supported strides (i.e. 2 and 3) when optimizing to reach a stride = 1 is not possible due to the IFM width not being divisible by the stride width. * Implement optimization for when SAME padding is used. If the pre-opt and post-opt padding do not match, add zero padding to the filter so that the post-opt IFM padding matches. Change-Id: Ia66b0d107281fa9993f6bf4d0c26627ee743253b Signed-off-by: Raul Farkas --- SUPPORTED_OPS.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'SUPPORTED_OPS.md') diff --git a/SUPPORTED_OPS.md b/SUPPORTED_OPS.md index 85ac0367..4c2a098a 100644 --- a/SUPPORTED_OPS.md +++ b/SUPPORTED_OPS.md @@ -18,7 +18,7 @@ limitations under the License. # Supported Ops This file was automatically generated by Vela using the `--supported-ops-report` parameter. -Vela version: `3.7.1.dev23+g3734897.d20230427` +Vela version: `3.7.1.dev17+geeff1bcf` This file complies with [**Gitiles Markdown syntax**](https://github.com/google/gitiles/blob/master/Documentation/markdown.md) @@ -153,7 +153,7 @@ This is a list of constraints that the CONV_2D operator must satisfy in order to - Stride values for both width and height must be integer types - Dilation factor values for both width and height must be integer types -- Stride values for height must be between 1 and 3 and for width between 1 and 4 +- Stride width must be greater than or equal to 1 and stride height must be between 1 and 3 - Dilated kernel height must be in the range [1, 64] - Product of dilated kernel width and height must be in the range [1, 4096] - Weight tensor must be 8-bit -- cgit v1.2.1