From cad4d1618fee74471d335314cb77070fee951a42 Mon Sep 17 00:00:00 2001
From: AlexeyAB <alexeyab84@gmail.com>
Date: Sun, 25 Feb 2018 13:29:44 +0000
Subject: [PATCH] Added support for Tensor Cores CC >= 7.0 (V100). For FP16/32 (mixed precision) define CUDNN_HALF should be used.
---
src/layer.h | 4 ++++
1 files changed, 4 insertions(+), 0 deletions(-)
diff --git a/src/layer.h b/src/layer.h
index 285abe3..93aca6c 100644
--- a/src/layer.h
+++ b/src/layer.h
@@ -63,6 +63,7 @@
int out_h, out_w, out_c;
int n;
int max_boxes;
+ int small_object;
int groups;
int size;
int side;
@@ -241,6 +242,9 @@
float * weights_gpu;
float * weight_updates_gpu;
+ float * weights_gpu16;
+ float * weight_updates_gpu16;
+
float * biases_gpu;
float * bias_updates_gpu;
--
Gitblit v1.10.0