nihuini
61ba8ec68e
mention the community-maintained tensorflow2ncnn tool
2019-11-07 14:30:23 +08:00
nihuini
6ce5da28a0
convert onnx pooling ceil_mode
2019-11-07 10:36:24 +08:00
Guoxia Wang
4336cc15f7
fix bug that MemoryData as multi operations input ( #1357 )
...
* fix onnx2ncnn MemoryData dims problem #1330
* fix ncnn2int8 bug (see #1336 )
* fix bug that MemoryData as multi operations input
* fix ncnn2table find_blob_index_by_name data failed bug, see #1338
* fix ncnn2table find_blob_index_by_name data failed bug, see #1338 , fix change requested
2019-11-05 15:53:22 +08:00
nihui
ced7bcbadc
fix eliminating noop with empty bottom or top blobs
2019-11-04 22:05:32 +08:00
nihui
90f04bab26
minor index optimization
2019-11-04 21:34:24 +08:00
nihuini
e73b06bbb8
fix build with NCNN_STRING=OFF
2019-11-04 11:16:54 +08:00
zyvv
bbcd5db817
Improvement for winograd43 int8 convolution ( #1354 )
...
* Update convolution_3x3_int8.h
* 针对winograd43 int8卷积的小改进
1. 使用非均匀量化系数，解决了权重转换溢出问题
2. 输出转换汇编优化
* Update convolution_3x3_int8.h
* Update convolution_3x3_int8.h
* Update convolution_3x3_int8.h
2019-11-04 10:16:21 +08:00
andywangII
66d7ac2463
arm-hisim100-linux make error convolutiondepthwise_3x3_int8.h for asm ( #1310 )
...
* arm-himix100-linux make error convolutiondepthwise_3x3_int8.h
* arm-hisim100-linux make error convolutiondepthwise_3x3_int8.h for asm
* convolutiondepthwise_3x3_int8.h f
2019-10-31 17:33:54 +08:00
nihuini
c9a9128db7
convert caffe Slice slice_dim
2019-10-30 19:25:13 +08:00
Natsu
f11b772dd1
Fix min/max macro issue ( #1346 )
...
* Fix min/max macro issue
* Make NOMINMAX public, since benchmark is build under CMake, it will no longer be bothered by min/max
2019-10-30 12:36:05 +08:00
nihuini
d5d7886517
resolve onnx Split conversion with empty split attribute
2019-10-30 10:31:24 +08:00
Joshua
96a2499051
fix spp::copy_make_border ( #1345 )
2019-10-29 21:05:31 +08:00
Christopher
2f26697aea
fix err in priorbox ( #1343 )
2019-10-29 17:23:51 +08:00
nihui
a4a162e36d
workaround validation layer complains about Cannot form constants of 8- or 16-bit types, due to specialization constants conversion
2019-10-29 16:33:45 +08:00
Guoxia Wang
31c6cefc44
fix ncnn2int8 bug, see issue #1336 ( #1341 )
...
* fix onnx2ncnn MemoryData dims problem #1330
* fix ncnn2int8 bug (see #1336 )
2019-10-29 15:25:21 +08:00
ShuangLiu1992
3b905cb48a
fix ncnnoptimize bias term bug #1339 ( #1340 )
2019-10-29 07:39:31 +08:00
nihui
ac8c56d6fe
update qcom410 and imx7d benchmark
2019-10-28 22:44:13 +08:00
nihuini
2e59da35a9
fill input and weight data with zero
2019-10-28 19:43:44 +08:00
nihuini
02b07b3e43
update qcom810 and iphone5s benchmark
2019-10-28 15:22:05 +08:00
nihuini
b2758bd9a0
fix deconvolution vulkan output crop
2019-10-28 11:08:57 +08:00
nihui
b77f92dab9
update qcom410 and imx7d benchmark
2019-10-28 08:31:12 +08:00
yehao
60d4ff09d0
add mobilenetv3-ssd ( #1335 )
...
* update
* Update mobilenetv3ssdlite.cpp
* Update mobilenetv3ssdlite.cpp
* Update mobilenetv3ssdlite.cpp
2019-10-27 21:51:38 +08:00
ShuangLiu1992
ecfc09bbc8
position independent code ( #1333 )
...
* web
* add disable pic option
2019-10-27 21:50:39 +08:00
新无止竞博客
4379f30e23
fix benchmark compilation errors ( #1334 )
...
* fix benchmark compilation errors
* some arm-linux toolchain file also defines ANDROID, it will break its build for missing libandroid.so
2019-10-27 21:49:57 +08:00
Guoxia Wang
51bca0b718
fix onnx2ncnn MemoryData dims problem #1330 ( #1331 )
2019-10-26 23:33:03 +08:00
nihui
2905639890
fix null fp handling
2019-10-26 12:28:57 +08:00
nihui
e1d095c7cd
update to android ndk r19c for travis ci
2019-10-26 10:02:15 +08:00
Guocai He
bbbaedda57
fix Crop bugs ( #1326 )
...
* fix crop bugs
* fix crop_arm ang crop_vulkan bug about w_offset,_h_offset,c_offset
2019-10-25 20:32:20 +08:00
nihuini
64333429bb
data reader wrapper, fix #1325
2019-10-25 19:18:40 +08:00
nihuini
567e2bd501
a dirty hack for resolving int8 pack4 crash
2019-10-25 11:48:39 +08:00
bindog
9dfd1b05d3
fix bug in reduction ( #1321 )
2019-10-25 07:16:48 +08:00
nihuini
f8caef7691
add shufflenet_v2 benchmark
2019-10-24 16:53:47 +08:00
nihuini
517d1b2053
improve compability with megvii shufflenet model
2019-10-24 15:07:30 +08:00
nihuini
e8bb88830d
convert mxnet squeeze expanddims, convert onnx squeeze unsqueeze
2019-10-24 11:29:14 +08:00
nihuini
a6c60068e6
convert numpy style slice to crop
2019-10-23 19:43:11 +08:00
nihuini
009c4d9a75
convert mxnet reduce axis and keepdims, pad reflect, fix #739
2019-10-23 15:46:39 +08:00
nihuini
c4b84262d9
fix arm neon sgemm, fix #1283
2019-10-23 15:12:26 +08:00
nihuini
65ce6bccfd
faster weight transform for optimized kernel
2019-10-23 11:23:44 +08:00
BUG1989
69e2693c87
fix the bug of SMP cpu powersave not supported.
2019-10-23 11:00:49 +08:00
nihuini
cd4be6d0fa
call vulkan create_pipeline on the vkdev condition, drop opt_cpu hacks
2019-10-23 10:50:28 +08:00
nihuini
81a028547a
fix bus error on armv7
2019-10-22 19:03:13 +08:00
daichuanliang
6176ada9f0
update ncnn2int8.cpp ( #1315 )
...
Fix compile issue with ncnn2int8
2019-10-22 18:47:30 +08:00
nihuini
19d75955d6
arm neon assembly optimization for conv3x3s1 winograd pack4to1
2019-10-22 17:53:03 +08:00
bindog
04b4b02324
[WIP] add reduce op support for onnx ( #1308 )
...
* [WIP] add reduce op support for onnx
* extend reduction to support 1,2-dim reduction and keepdims
* fix compile error
* split type to 3 flags && split keepdims to another function
2019-10-22 12:53:12 +08:00
nihuini
22a2be4e6c
fix crop pack4 with reference blob
2019-10-22 10:55:54 +08:00
nihuini
6a8e5c58da
fix build on armv7
2019-10-22 10:42:15 +08:00
nihuini
e63e2449fd
arm neon assembly optimization for conv7x7s2 pack1to4
2019-10-21 12:49:33 +08:00
nihui
56fd26a2da
arm neon assembly optimization for conv1x1s1 pack4to1
2019-10-20 17:07:51 +08:00
nihui
7ad514917b
fix potential out of write on unroll 12 remainder
2019-10-20 16:56:29 +08:00
nihuini
15e86dc8e9
reduce pack4 weight memory usage for specialized kernel, reduce runtime memory usage in conv3x3s1 winograd
2019-10-18 10:53:33 +08:00
nihuini