|
|
ae3ce1b181
|
Minor fixes
|
2019-05-20 22:37:42 +03:00 |
|
|
|
4f72fcc015
|
Added grouped convolutional (depth-wise convolutional)
|
2019-05-10 16:46:48 +03:00 |
|
|
|
da74882fe1
|
cleanup in preparation to opencv-4 work
|
2019-03-28 17:54:27 +01:00 |
|
|
|
b6e15f1656
|
ZED 3D Camera support added to ./uselib (yolo_console_cpp.exe) example
|
2019-03-18 02:48:52 +03:00 |
|
|
|
75f2a3e7cf
|
Added object Detection & Tracking using conv-rnn layer on frames from video
|
2019-03-02 03:32:24 +03:00 |
|
|
|
b3579380dc
|
improve compatibility with c++ compilers, prepare for CMake
|
2019-02-15 17:27:12 +01:00 |
|
|
|
3d9c8530a0
|
Use Tensor Cores only when (channels % 8 == 0) and (filters % 8 == 0)
|
2019-02-12 23:13:25 +03:00 |
|
|
|
5448e07445
|
Try to fuse conv_xnor+shortcut -> conv_xnor
|
2019-02-12 02:05:15 +03:00 |
|
|
|
fa1415e3c2
|
CUDNN_HALF and CC 7.5 by default in darknet.sln
|
2019-02-05 20:43:07 +03:00 |
|
|
|
edfdf2c20e
|
Fixed bug in Tensor Cores training
|
2019-02-05 19:33:10 +03:00 |
|
|
|
12b6e93893
|
CHECK_CUDA is used everywhere
|
2019-02-05 16:18:36 +03:00 |
|
|
|
d767e8ca38
|
Minor fixes
|
2019-02-04 23:29:06 +03:00 |
|
|
|
61156239e0
|
Minor performance improvement
|
2019-02-03 00:18:30 +03:00 |
|
|
|
41814fc4b3
|
Minor fixes
|
2019-02-02 15:16:57 +03:00 |
|
|
|
f91d5a5e09
|
Fixed __shfl() and __ballot() warnings
|
2019-02-02 03:16:05 +03:00 |
|
|
|
f09a9c3315
|
XNOR uses Tensor Cores on Turing GPU CC>=7.3 (not Volta)
|
2019-02-02 00:24:34 +03:00 |
|
|
|
640bdbc063
|
LSTM, RNN, GRU - use connected_layer that uses cuDNN. Fixed CRNN for conv-layer with cuDNN.
|
2019-01-28 23:50:51 +03:00 |
|
|
|
85b99872cb
|
Use non-default stream for all CUDA-functions
|
2019-01-28 20:19:26 +03:00 |
|
|
|
17019854c3
|
XNOR minor fix
|
2019-01-19 03:18:50 +03:00 |
|
|
|
0e022d0912
|
Fixed timer
|
2019-01-18 21:29:06 +03:00 |
|
|
|
5343aa4235
|
CUDA minor performance improvement
|
2019-01-16 18:08:11 +03:00 |
|
|
|
4c05166215
|
Temporary experimental XNOR on GPU (repack channels)
|
2019-01-16 02:43:44 +03:00 |
|
|
|
64e478db07
|
Fix training approach (convolutional layer)
|
2018-12-27 00:31:28 +03:00 |
|
|
|
3969ce30ed
|
Speedup Tensor Cores: 1st layer uses FP32 and pre-allocate GPU memory for Tensor Cores
|
2018-12-11 23:48:58 +03:00 |
|
|
|
25f133d6ef
|
Another one minor fix
|
2018-12-11 21:26:36 +03:00 |
|
|
|
cb998db949
|
Some fix for CUDNN_HALF
|
2018-12-11 21:16:18 +03:00 |
|
|
|
a621235783
|
Switch to Tensor Cores after 2000 iterations.
|
2018-12-10 01:35:08 +03:00 |
|
|
|
dc7f8a32ae
|
mAP calculation during training, if is used flag -map
|
2018-12-09 18:18:47 +03:00 |
|
|
|
742bb7c7ce
|
Compile fix
|
2018-12-07 22:52:07 +03:00 |
|
|
|
7c2f302321
|
Fixed nan issue for training with CUDNN_HALF=1 by using Tensor Cores
|
2018-12-07 22:40:10 +03:00 |
|
|
|
21a4ec9390
|
Saving loss-chart for each 100 iterations automatically
|
2018-11-26 11:11:56 +03:00 |
|
|
|
9f7d7c58b5
|
Minor fixes. Use CUDA 10.0
|
2018-11-17 02:48:46 +03:00 |
|
|
|
25f65f6878
|
Added fast_binarize_weights_gpu()
|
2018-11-05 22:38:35 +03:00 |
|
|
|
c0e2512af2
|
Activation improvement, more robust timer.
|
2018-09-27 23:10:54 +03:00 |
|
|
|
7dd97537fb
|
XNOR-net tiny-yolo_xnor.cfg ~2x faster than cuDNN on CUDA (nVidia GPU Maxwell)
|
2018-09-22 02:01:14 +03:00 |
|
|
|
03e95320a1
|
XNOR coalesced memory access, and avoid bank conflicts
|
2018-09-17 23:39:25 +03:00 |
|
|
|
ca43bbdaae
|
Fixed openmp bugs for XNOR
|
2018-09-12 16:22:54 +03:00 |
|
|
|
c0e01fd63c
|
Test for XNOR-conv on CUDA
|
2018-09-08 02:46:05 +03:00 |
|
|
|
b141f85cab
|
Compile fix
|
2018-09-07 15:07:46 +03:00 |
|
|
|
007878393f
|
Temporary Slow implementation of XNOR on CUDA (shared_memory)
|
2018-09-06 23:21:26 +03:00 |
|
|
|
c4a9e3422e
|
Temporary implementation of XNOR on CUDA
|
2018-08-31 02:47:58 +03:00 |
|
|
|
9753b72aeb
|
temp fix, don't use it
|
2018-08-30 17:24:41 +03:00 |
|
|
|
cfc5fedbb6
|
Just used spaces for indents instead of Tabs
|
2018-07-10 23:29:15 +03:00 |
|
|
|
9bae70b225
|
Accelerated by another 5% using FP16/32 Batch-norm for Tensor Cores.
|
2018-04-17 02:51:11 +03:00 |
|
|
|
537d135feb
|
Improve training performance - batch-norm using cuDNN.
|
2018-03-20 02:16:51 +03:00 |
|
|
|
880cf187d8
|
Fixed multi-GPU training for Tensor Cores
|
2018-03-09 19:44:46 +03:00 |
|
|
|
cad4d1618f
|
Added support for Tensor Cores CC >= 7.0 (V100). For FP16/32 (mixed precision) define CUDNN_HALF should be used.
|
2018-02-25 16:29:44 +03:00 |
|
|
|
cd2bdec090
|
Updated to CUDA 9.1. And fixed no_gpu dependecies.
|
2018-02-23 15:05:31 +03:00 |
|
|
|
6332ea99ab
|
one more fix
|
2018-02-23 00:13:08 +03:00 |
|
|
|
b2b5756d86
|
Added __float2half_rn() and __half2float()
|
2018-02-22 23:52:43 +03:00 |
|
