nihuini
296e0022df
deconvolution output adj and output shape
6 years ago
nihuini
e4b44d293e
more autopad SAME_LOWER
6 years ago
nihuini
9a6ee37eef
asymmetric padding parameter for convolution and deconvolution family
6 years ago
nihui
29324771b1
Implemented hard swish layer ( #1195 )
6 years ago
nihuini
3d5b7f20ff
avgpool count_include_pad
6 years ago
Hao Zeng
1f6919fd40
Implemented hard swish layer
6 years ago
nihuini
1cce18bdde
binaryop broadcast vulkan
6 years ago
nihui
0b6d7b7096
use underscored offset
6 years ago
nihuini
163fb92537
concat vulkan pack1to4 and pack4to1to4
6 years ago
nihuini
1b910efea5
convert slice properly
6 years ago
nihui
cf42e7c254
deconvolutiondepthwise pack4 arm neon
6 years ago
nihui
60c0890eaf
crop region -234 is rarely used, fix out of channel range write, crop pack4 arm neon
6 years ago
nihuini
c4f23ae8ad
rename Mat packing to elempack
6 years ago
nihui
c013bd9b7e
vulkan convolution winograd f63
6 years ago
nihuini
c769437533
fix fp16p deconvolution and convolution-typed innerproduct
7 years ago
PENGUINLIONG
084053fed8
Implemented hard sigmoid ( #1046 )
* Implementation for hard sigmoid
Not yet tested.
* Resolve requests
7 years ago
nihui
21f79b8546
prefer cpu fp16 casting to reduce upload/download overhead on discrete gpu
7 years ago
nihuini
e09607bc22
add option to upload model function, pipeline creation honors option use flags, setting allocator per extractor do not make much sense
7 years ago
nihui
fe4b00f7a2
unroll outh 4 for winograd gemm
7 years ago
nihuini
74276314bb
unroll size 4 for conv1x1s1 pack4
7 years ago
nihuini
cd7559c639
more fix for fp16p, still disabled by default
7 years ago
harhar539
5e317b98c5
fix illegal memory access at conv layer of vulkan ( #1011 )
* 1.fix pad tail bug in commit d1ea2a3 at pooling layer
* fix illegal memory access at conv layer of vulkan
fix illegal memory access at conv layer of vulkan when bias term is 0
7 years ago
nihui
25b9736f82
shader fp16 packed
7 years ago
nihuini
4b50a97e31
implement vulkan winograd23
7 years ago
nihuini
aa94e77e68
fix pipeline object leak
7 years ago
nihui
3e003ffd98
fuse sigmoid
7 years ago
nihui
5adfa290a5
1x1s1d1_lds_4_4_4 is non-optimal, delete it
7 years ago
nihuini
8ac300c3a2
mat4 type in shared memory makes some driver unhappy ..
7 years ago
nihuini
f5ba97e7c6
lds optimize for conv3x3s1, conv1x1s1 and fc
7 years ago
nihuini
7a8f68aca6
move vulkan code to subdir, new layer interface create_pipeline and destroy_pipeline for post-loading works
7 years ago