zhiliu6
742cc8c571
optimize yolov3 output extraction speed. ( #1398 )
6 years ago
lyfe667
3c7863d341
fix line 1142 bug ( #1397 )
6 years ago
nihuini
279e67bb88
normalize vulkan
6 years ago
nihuini
e917796f97
normalize support inplace
6 years ago
nihuini
095efc5471
load param value type
6 years ago
JackieWu
dbadd15ee4
support OpenCV4 for simplepose example ( #1391 )
6 years ago
nihuini
5dcd5667ad
fix wrong prelu vulkan pack4 slope data
6 years ago
nihuini
50e8b5e4e8
multiple transfers may run concurrently if there is no dependency with each other, do not share staging buffer memory to fix potential data race
6 years ago
nihuini
dce75e1f07
pooling vulkan pad auto SAME mode, fix avgpool_count_include_pad handling
6 years ago
nihuini
d8b0f28dda
optimizer eliminate orphaned memorydata
6 years ago
nihuini
5eae1b8325
always write normalize scale data, do not break optimize tool
6 years ago
nihuini
d430f1e1fe
eliminate noop pooling1x1s1
6 years ago
nihuini
33956cbfc3
pretty error info
6 years ago
nihuini
af3643ed81
normalize eps mode for different training framework conventions
6 years ago
nihuini
c95e7bbb3c
onnx fuse flatten
6 years ago
nihuini
253e54351d
onnx fuse normalize, first try
6 years ago
nihuini
e5a834b299
do not fuse weight reshape
6 years ago
nihuini
66b8a7ee36
fuse onnx unsqueeze-prelu
6 years ago
nihui
b29e8b0e09
check and enable more vulkan extensions
6 years ago
nihui
4bc4a5fa63
webassembly ci ( #1375 )
6 years ago
nihuini
9593783be3
package android x86/x86_64 library
6 years ago
chenyang
fda1c8c530
fix return of get_input_names ( #1373 )
6 years ago
nihuini
f6ddbd9a1c
android x86 x86_64 cpu/gpu build
6 years ago
nihuini
bffb2af2ff
fix build on armv7 without neon
6 years ago
nihuini
dc589351c1
link android
6 years ago
nihuini
a170ef1acf
remove the default option usage in layer interface, fix write out of range in cast arm pack4, handle fp16p conversion on cpu/gpu transfer
6 years ago
Guoxia Wang
5c08ce7db9
fix Reduction Op attr bug ( #1363 )
* 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
* fix Reduction operation attribution bug
* fix #1363 request changes
6 years ago
nihuini
336d1c1edd
remove the ncnn namespace for in source Option
6 years ago
nihuini
61ba8ec68e
mention the community-maintained tensorflow2ncnn tool
6 years ago
nihuini
6ce5da28a0
convert onnx pooling ceil_mode
6 years ago
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
6 years ago
nihui
ced7bcbadc
fix eliminating noop with empty bottom or top blobs
6 years ago
nihui
90f04bab26
minor index optimization
6 years ago
nihuini
e73b06bbb8
fix build with NCNN_STRING=OFF
6 years ago
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
6 years ago
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
6 years ago
nihuini
c9a9128db7
convert caffe Slice slice_dim
6 years ago
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
6 years ago
nihuini
d5d7886517
resolve onnx Split conversion with empty split attribute
6 years ago
Joshua
96a2499051
fix spp::copy_make_border ( #1345 )
6 years ago
Christopher
2f26697aea
fix err in priorbox ( #1343 )
6 years ago
nihui
a4a162e36d
workaround validation layer complains about Cannot form constants of 8- or 16-bit types, due to specialization constants conversion
6 years ago
Guoxia Wang
31c6cefc44
fix ncnn2int8 bug, see issue #1336 ( #1341 )
* fix onnx2ncnn MemoryData dims problem #1330
* fix ncnn2int8 bug (see #1336 )
6 years ago
ShuangLiu1992
3b905cb48a
fix ncnnoptimize bias term bug #1339 ( #1340 )
6 years ago
nihui
ac8c56d6fe
update qcom410 and imx7d benchmark
6 years ago
nihuini
2e59da35a9
fill input and weight data with zero
6 years ago
nihuini
02b07b3e43
update qcom810 and iphone5s benchmark
6 years ago
nihuini
b2758bd9a0
fix deconvolution vulkan output crop
6 years ago
nihui
b77f92dab9
update qcom410 and imx7d benchmark
6 years ago
yehao
60d4ff09d0
add mobilenetv3-ssd ( #1335 )
* update
* Update mobilenetv3ssdlite.cpp
* Update mobilenetv3ssdlite.cpp
* Update mobilenetv3ssdlite.cpp
6 years ago