Skip to content
Snippets Groups Projects
Commit e847a50b authored by Hashim Sharif's avatar Hashim Sharif
Browse files

Adding Tuner Config and Runtime quantization ranges file

parent b1e20669
No related branches found
No related tags found
No related merge requests found
Hide whitespace changes
Inline Side-by-side
Showing
with 480 additions and 0 deletions
llvm/test/VISC/DNN_Benchmarks/benchmarks/alexnet/data/tuner_confs_base.txt 0 → 100644
+ 20
− 0
View file @ e847a50b
+++++
conf1 1 0 79.9 0
1 gpu conv fp32 1 add fp32 1 tanh fp32 1 pool_max fp32 1
2 gpu conv fp32 1 add fp32 1 tanh fp32 1 pool_max fp32 1
3 gpu conv fp32 1 add fp32 1 tanh fp32 1
4 gpu conv fp32 1 add fp32 1 tanh fp32 1
5 gpu conv fp32 1 add fp32 1 tanh fp32 1 pool_max fp32 1
6 gpu mul fp32 1 add fp32 1
7 gpu softmax fp32 1
-----
+++++
conf2 1.5 0 79.9 0
1 gpu conv fp16 1 add fp16 1 tanh fp16 1 pool_max fp16 1
2 gpu conv fp16 1 add fp16 1 tanh fp16 1 pool_max fp16 1
3 gpu conv fp16 1 add fp16 1 tanh fp16 1
4 gpu conv fp16 1 add fp16 1 tanh fp16 1
5 gpu conv fp16 1 add fp16 1 tanh fp16 1 pool_max fp16 1
6 gpu mul fp16 1 add fp16 1
7 gpu softmax fp32 1
-----
llvm/test/VISC/DNN_Benchmarks/benchmarks/lenet/data/tuner_confs_base.txt 0 → 100644
+ 16
− 0
View file @ e847a50b
+++++
conf1 1 0 98.9 0
1 gpu conv fp32 1 add fp32 1 pool_max fp32 1 tanh fp32 1
2 gpu conv fp32 1 add fp32 1 pool_max fp32 1 tanh fp32 1
3 gpu mul fp32 1 add fp32 1 tanh fp32 1
4 gpu mul fp32 1 add fp32 1 tanh fp32 1
5 gpu softmax fp32 1
-----
+++++
conf2 1.5 0 98.9 0
1 gpu conv fp16 1 add fp16 1 pool_max fp16 1 tanh fp16 1
2 gpu conv fp16 1 add fp16 1 pool_max fp16 1 tanh fp16 1
3 gpu mul fp16 1 add fp16 1 tanh fp16 1
4 gpu mul fp16 1 add fp16 1 tanh fp16 1
5 gpu softmax fp32 1
-----
llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet/data/quant_ranges_rt.txt 0 → 100644
+ 15
− 0
View file @ e847a50b
1 -1.9892114 2.126797 -2.19630692005 1.34758170414 0.0 0.0 -60.892750473 51.9925691605
2 0.0 5.71354155397 -0.931772116065 1.07742589378 0.0 0.0 -6.51858950329 6.81084251881
3 0.0 4.93213940287 -0.531654466152 0.57537904036 0.0 0.0 -4.48263123512 3.96730119753
4 0.0 4.10326339769 -0.362340988219 0.407691390038 0.0 0.0 -4.04261828327 3.8867793293
5 0.0 5.38322130251 -0.313120054901 0.293576799393 0.0 0.0 -5.92146921539 4.33867932415
6 0.0 4.31673815441 -0.232992478013 0.258029025793 0.0 0.0 -4.20778994751 3.93243697071
7 0.0 5.8304081068 -0.202337772191 0.189983081758 0.0 0.0 -6.29828691578 4.84813511753
8 0.0 4.44641780996 -0.174427356511 0.176958308667 0.0 0.0 -4.34791088581 3.61443646955
9 0.0 4.5180956049 -0.145467961878 0.15256431669 0.0 0.0 -3.02877027559 2.94873657799
10 0.0 6.34857563496 -0.130258745223 0.135582433432 0.0 0.0 -4.22931008053 3.53150463724
11 0.0 5.22100311041 -0.119001727596 0.125363747835 0.0 0.0 -4.03820378017 4.00400940704
12 0.0 5.73249834776 -0.108397216856 0.116256686077 0.0 0.0 -3.31110151148 4.46293323326
13 0.0 7.24049821186 -0.0862374496162 0.0885944995135 0.0 0.0 -4.17543139458 6.2043294754
14 0.0 7.81395883465 -0.0681302513927 0.0700202777982 0.0 0.0 -10.9205664234 2.64429125786
15 0.0 2.86920666504 -0.223010196954 0.14426593782 -0.1654396 0.23336112 -12.2459499588 23.8053251343
llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet/data/tuner_confs_base.txt 0 → 100644
+ 174
− 0
View file @ e847a50b
+++++
conf1 1 0 84.8 0
1 gpu conv fp32 1
2 gpu batchnorm fp32 1
3 gpu relu fp32 1
4 gpu group_conv fp32 1
5 gpu batchnorm fp32 1
6 gpu relu fp32 1
7 gpu conv fp32 1
8 gpu batchnorm fp32 1
9 gpu relu fp32 1
10 gpu group_conv fp32 1
11 gpu batchnorm fp32 1
12 gpu relu fp32 1
13 gpu conv fp32 1
14 gpu batchnorm fp32 1
15 gpu relu fp32 1
16 gpu group_conv fp32 1
17 gpu batchnorm fp32 1
18 gpu relu fp32 1
19 gpu conv fp32 1
20 gpu batchnorm fp32 1
21 gpu relu fp32 1
22 gpu group_conv fp32 1
23 gpu batchnorm fp32 1
24 gpu relu fp32 1
25 gpu conv fp32 1
26 gpu batchnorm fp32 1
27 gpu relu fp32 1
28 gpu group_conv fp32 1
29 gpu batchnorm fp32 1
30 gpu relu fp32 1
31 gpu conv fp32 1
32 gpu batchnorm fp32 1
33 gpu relu fp32 1
34 gpu group_conv fp32 1
35 gpu batchnorm fp32 1
36 gpu relu fp32 1
37 gpu conv fp32 1
38 gpu batchnorm fp32 1
39 gpu relu fp32 1
40 gpu group_conv fp32 1
41 gpu batchnorm fp32 1
42 gpu relu fp32 1
43 gpu conv fp32 1
44 gpu batchnorm fp32 1
45 gpu relu fp32 1
46 gpu group_conv fp32 1
47 gpu batchnorm fp32 1
48 gpu relu fp32 1
49 gpu conv fp32 1
50 gpu batchnorm fp32 1
51 gpu relu fp32 1
52 gpu group_conv fp32 1
53 gpu batchnorm fp32 1
54 gpu relu fp32 1
55 gpu conv fp32 1
56 gpu batchnorm fp32 1
57 gpu relu fp32 1
58 gpu group_conv fp32 1
59 gpu batchnorm fp32 1
60 gpu relu fp32 1
61 gpu conv fp32 1
62 gpu batchnorm fp32 1
63 gpu relu fp32 1
64 gpu group_conv fp32 1
65 gpu batchnorm fp32 1
66 gpu relu fp32 1
67 gpu conv fp32 1
68 gpu batchnorm fp32 1
69 gpu relu fp32 1
70 gpu group_conv fp32 1
71 gpu batchnorm fp32 1
72 gpu relu fp32 1
73 gpu conv fp32 1
74 gpu batchnorm fp32 1
75 gpu relu fp32 1
76 gpu group_conv fp32 1
77 gpu batchnorm fp32 1
78 gpu relu fp32 1
79 gpu conv fp32 1
80 gpu batchnorm fp32 1
81 gpu relu fp32 1
82 gpu pool_mean fp32 1
83 gpu mul fp32 1 add fp32 1
84 gpu softmax fp32 1
-----
+++++
conf2 1.5 0 84.8 0
1 gpu conv fp16 1
2 gpu batchnorm fp16 1
3 gpu relu fp16 1
4 gpu group_conv fp16 1
5 gpu batchnorm fp16 1
6 gpu relu fp16 1
7 gpu conv fp16 1
8 gpu batchnorm fp16 1
9 gpu relu fp16 1
10 gpu group_conv fp16 1
11 gpu batchnorm fp16 1
12 gpu relu fp16 1
13 gpu conv fp16 1
14 gpu batchnorm fp16 1
15 gpu relu fp16 1
16 gpu group_conv fp16 1
17 gpu batchnorm fp16 1
18 gpu relu fp16 1
19 gpu conv fp16 1
20 gpu batchnorm fp16 1
21 gpu relu fp16 1
22 gpu group_conv fp16 1
23 gpu batchnorm fp16 1
24 gpu relu fp16 1
25 gpu conv fp16 1
26 gpu batchnorm fp16 1
27 gpu relu fp16 1
28 gpu group_conv fp16 1
29 gpu batchnorm fp16 1
30 gpu relu fp16 1
31 gpu conv fp16 1
32 gpu batchnorm fp16 1
33 gpu relu fp16 1
34 gpu group_conv fp16 1
35 gpu batchnorm fp16 1
36 gpu relu fp16 1
37 gpu conv fp16 1
38 gpu batchnorm fp16 1
39 gpu relu fp16 1
40 gpu group_conv fp16 1
41 gpu batchnorm fp16 1
42 gpu relu fp16 1
43 gpu conv fp16 1
44 gpu batchnorm fp16 1
45 gpu relu fp16 1
46 gpu group_conv fp16 1
47 gpu batchnorm fp16 1
48 gpu relu fp16 1
49 gpu conv fp16 1
50 gpu batchnorm fp16 1
51 gpu relu fp16 1
52 gpu group_conv fp16 1
53 gpu batchnorm fp16 1
54 gpu relu fp16 1
55 gpu conv fp16 1
56 gpu batchnorm fp16 1
57 gpu relu fp16 1
58 gpu group_conv fp16 1
59 gpu batchnorm fp16 1
60 gpu relu fp16 1
61 gpu conv fp16 1
62 gpu batchnorm fp16 1
63 gpu relu fp16 1
64 gpu group_conv fp16 1
65 gpu batchnorm fp16 1
66 gpu relu fp16 1
67 gpu conv fp16 1
68 gpu batchnorm fp16 1
69 gpu relu fp16 1
70 gpu group_conv fp16 1
71 gpu batchnorm fp16 1
72 gpu relu fp16 1
73 gpu conv fp16 1
74 gpu batchnorm fp16 1
75 gpu relu fp16 1
76 gpu group_conv fp16 1
77 gpu batchnorm fp16 1
78 gpu relu fp16 1
79 gpu conv fp16 1
80 gpu batchnorm fp16 1
81 gpu relu fp16 1
82 gpu pool_mean fp16 1
83 gpu mul fp16 1 add fp16 1
84 gpu softmax fp32 1
-----
llvm/test/VISC/DNN_Benchmarks/benchmarks/mobilenet_shallow/data/tuner_confs_base.txt 0 → 100644
+ 90
− 0
View file @ e847a50b
+++++
conf1 1 0 87.59 0
1 gpu conv fp32 1
2 gpu batchnorm fp32 1
3 gpu relu fp32 1
4 gpu group_conv fp32 1
5 gpu batchnorm fp32 1
6 gpu relu fp32 1
7 gpu conv fp32 1
8 gpu batchnorm fp32 1
9 gpu relu fp32 1
10 gpu group_conv fp32 1
11 gpu batchnorm fp32 1
12 gpu relu fp32 1
13 gpu conv fp32 1
14 gpu batchnorm fp32 1
15 gpu relu fp32 1
16 gpu group_conv fp32 1
17 gpu batchnorm fp32 1
18 gpu relu fp32 1
19 gpu conv fp32 1
20 gpu batchnorm fp32 1
21 gpu relu fp32 1
22 gpu group_conv fp32 1
23 gpu batchnorm fp32 1
24 gpu relu fp32 1
25 gpu conv fp32 1
26 gpu batchnorm fp32 1
27 gpu relu fp32 1
28 gpu group_conv fp32 1
29 gpu batchnorm fp32 1
30 gpu relu fp32 1
31 gpu conv fp32 1
32 gpu batchnorm fp32 1
33 gpu relu fp32 1
34 gpu group_conv fp32 1
35 gpu batchnorm fp32 1
36 gpu relu fp32 1
37 gpu conv fp32 1
38 gpu batchnorm fp32 1
39 gpu relu fp32 1
40 gpu pool_mean fp32 1
41 gpu mul fp32 1 add fp32 1
42 gpu softmax fp32 1
-----
+++++
conf2 1.5 0 87.59 0
1 gpu conv fp16 1
2 gpu batchnorm fp16 1
3 gpu relu fp16 1
4 gpu group_conv fp16 1
5 gpu batchnorm fp16 1
6 gpu relu fp16 1
7 gpu conv fp16 1
8 gpu batchnorm fp16 1
9 gpu relu fp16 1
10 gpu group_conv fp16 1
11 gpu batchnorm fp16 1
12 gpu relu fp16 1
13 gpu conv fp16 1
14 gpu batchnorm fp16 1
15 gpu relu fp16 1
16 gpu group_conv fp16 1
17 gpu batchnorm fp16 1
18 gpu relu fp16 1
19 gpu conv fp16 1
20 gpu batchnorm fp16 1
21 gpu relu fp16 1
22 gpu group_conv fp16 1
23 gpu batchnorm fp16 1
24 gpu relu fp16 1
25 gpu conv fp16 1
26 gpu batchnorm fp16 1
27 gpu relu fp16 1
28 gpu group_conv fp16 1
29 gpu batchnorm fp16 1
30 gpu relu fp16 1
31 gpu conv fp16 1
32 gpu batchnorm fp16 1
33 gpu relu fp16 1
34 gpu group_conv fp16 1
35 gpu batchnorm fp16 1
36 gpu relu fp16 1
37 gpu conv fp16 1
38 gpu batchnorm fp16 1
39 gpu relu fp16 1
40 gpu pool_mean fp16 1
41 gpu mul fp16 1 add fp16 1
42 gpu softmax fp32 1
-----
llvm/test/VISC/DNN_Benchmarks/benchmarks/resnet18/data/quant_ranges_rt.txt 0 → 100644
+ 22
− 0
View file @ e847a50b
1 -0.5500815 0.60786617 -1.0248864 1.2929907 -0.36291853 0.2533059 0.0 0.753551840782
2 0.0 0.753551840782 -0.69884616 0.71849966 -0.2781147 0.45571187 0.0 1.01057458043
3 0.0 1.01057458043 -0.59568167 0.7714691 -0.8602873 0.19743633 -1.84771883726 1.87930787086
4 0.0 2.33981014252 -0.41976976 0.43748936 -0.7021962 0.3033103 0.0 1.04317724705
5 0.0 1.04317724705 -0.46757826 0.4635873 -0.20662616 0.1778044 -0.829483509064 0.786805033684
6 0.0 2.49733686686 -0.64404047 0.45383143 -0.819547 0.38550296 0.0 0.897360802293
7 0.0 0.897360802293 -0.41986948 0.33654243 -0.3563013 0.22371122 -0.957150224447 0.54919362247
8 0.0 2.37362146616 -0.4805263 0.50655717 -0.296758 0.7742441 0.0 3.01592136621
9 0.0 3.01592136621 -0.52083415 0.45517674 -0.20242067 0.8236838 -5.2759475708 5.79733039856
10 0.0 2.37362146616 -0.5338656 1.3395424 -0.20242067 0.8236838 -0.738995380998 2.33600783587
11 0.0 7.07933432579 -0.34429058 0.43629733 -1.0744808 0.056708273 0.0 1.58645607233
12 0.0 1.58645607233 -0.30342352 0.39493486 -0.44630566 0.6492069 -1.49672914267 1.29970229745
13 0.0 7.11914063454 -0.38351893 0.45775774 -1.4733055 -0.014426912 0.0 1.52876508832
14 0.0 1.52876508832 -0.25695276 0.45372736 -0.5259744 0.26591402 -1.59576894164 1.08074297309
15 0.0 6.94405080318 -0.55299705 0.5443531 -0.71790683 1.2730768 0.0 10.3651468277
16 0.0 10.3651468277 -0.4203967 0.48641303 -0.90653443 1.3546854 -22.372925148 17.2033731079
17 0.0 6.94405080318 -0.4365755 0.84913826 -0.90653443 1.3546851 -3.66810325861 4.87814051151
18 0.0 18.8401451111 -0.38657624 0.5228989 -1.2083547 0.76361173 0.0 19.1229192352
19 0.0 19.1229192352 -0.40857902 0.575035 -1.8731614 1.0960501 -31.3229312897 14.8234729958
20 0.0 23.7382488823 -0.33079496 0.5893278 -1.0234511 1.0016295 0.0 19.5892774963
21 0.0 19.5892774963 -0.27897888 0.38280907 -2.2086356 1.0066502 -34.4416886902 20.9890329933
22 0.0 10.8541981602 -1.5092047 1.0279838 -0.49379802 0.61032647 -40.9121678543 25.7082381058
llvm/test/VISC/DNN_Benchmarks/benchmarks/resnet18/data/tuner_confs_base.txt 0 → 100644
+ 90
− 0
View file @ e847a50b
+++++
conf1 1 0 89.59 0
1 gpu conv fp32 1 add fp32 1 relu fp32 1
2 gpu conv fp32 1 add fp32 1 relu fp32 1
3 gpu conv fp32 1 add fp32 1
4 gpu add fp32 1
5 gpu relu fp32 1
6 gpu conv fp32 1 add fp32 1 relu fp32 1
7 gpu conv fp32 1 add fp32 1
8 gpu add fp32 1
9 gpu relu fp32 1
10 gpu conv fp32 1 add fp32 1 relu fp32 1
11 gpu conv fp32 1 add fp32 1
12 gpu add fp32 1
13 gpu relu fp32 1
14 gpu conv fp32 1 add fp32 1 relu fp32 1
15 gpu conv fp32 1 add fp32 1
16 gpu conv fp32 1 add fp32 1
17 gpu add fp32 1
18 gpu relu fp32 1
19 gpu conv fp32 1 add fp32 1 relu fp32 1
20 gpu conv fp32 1 add fp32 1
21 gpu add fp32 1
22 gpu relu fp32 1
23 gpu conv fp32 1 add fp32 1 relu fp32 1
24 gpu conv fp32 1 add fp32 1
25 gpu add fp32 1
26 gpu relu fp32 1
27 gpu conv fp32 1 add fp32 1 relu fp32 1
28 gpu conv fp32 1 add fp32 1
29 gpu conv fp32 1 add fp32 1
30 gpu add fp32 1
31 gpu relu fp32 1
32 gpu conv fp32 1 add fp32 1 relu fp32 1
33 gpu conv fp32 1 add fp32 1
34 gpu add fp32 1
35 gpu relu fp32 1
36 gpu conv fp32 1 add fp32 1 relu fp32 1
37 gpu conv fp32 1 add fp32 1
38 gpu add fp32 1
39 gpu relu fp32 1
40 gpu pool_mean fp32 1
41 gpu mul fp32 1 add fp32 1
42 gpu softmax fp32 1
-----
+++++
conf2 1.5 0 89.59 0
1 gpu conv fp16 1 add fp16 1 relu fp16 1
2 gpu conv fp16 1 add fp16 1 relu fp16 1
3 gpu conv fp16 1 add fp16 1
4 gpu add fp16 1
5 gpu relu fp16 1
6 gpu conv fp16 1 add fp16 1 relu fp16 1
7 gpu conv fp16 1 add fp16 1
8 gpu add fp16 1
9 gpu relu fp16 1
10 gpu conv fp16 1 add fp16 1 relu fp16 1
11 gpu conv fp16 1 add fp16 1
12 gpu add fp16 1
13 gpu relu fp16 1
14 gpu conv fp16 1 add fp16 1 relu fp16 1
15 gpu conv fp16 1 add fp16 1
16 gpu conv fp16 1 add fp16 1
17 gpu add fp16 1
18 gpu relu fp16 1
19 gpu conv fp16 1 add fp16 1 relu fp16 1
20 gpu conv fp16 1 add fp16 1
21 gpu add fp16 1
22 gpu relu fp16 1
23 gpu conv fp16 1 add fp16 1 relu fp16 1
24 gpu conv fp16 1 add fp16 1
25 gpu add fp16 1
26 gpu relu fp16 1
27 gpu conv fp16 1 add fp16 1 relu fp16 1
28 gpu conv fp16 1 add fp16 1
29 gpu conv fp16 1 add fp16 1
30 gpu add fp16 1
31 gpu relu fp16 1
32 gpu conv fp16 1 add fp16 1 relu fp16 1
33 gpu conv fp16 1 add fp16 1
34 gpu add fp16 1
35 gpu relu fp16 1
36 gpu conv fp16 1 add fp16 1 relu fp16 1
37 gpu conv fp16 1 add fp16 1
38 gpu add fp16 1
39 gpu relu fp16 1
40 gpu pool_mean fp16 1
41 gpu mul fp16 1 add fp16 1
42 gpu softmax fp32 1
-----
llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar10/data/quant_ranges_rt.txt 0 → 100644
+ 15
− 0
View file @ e847a50b
1 -1.8816367 2.0934217 -0.53275156 0.49437004 -0.6403629 0.2490165 0.0 1.35908746719
2 0.0 1.35908746719 -0.2688396 0.20639156 -0.7745511 0.82006615 0.0 2.52123117924
3 0.0 2.52123117924 -0.16776876 0.14878987 -0.35283303 0.5154362 0.0 1.20119857848
4 0.0 1.20119857848 -0.088948585 0.114222586 -0.30250227 0.36856708 0.0 1.03598809302
5 0.0 1.03598809302 -0.07739562 0.10973293 -0.15568458 0.17634983 0.0 0.300495595038
6 0.0 0.300495595038 -0.051649556 0.05435231 -0.07395447 0.07996062 0.0 0.11490475405
7 0.0 0.11490475405 -0.043513633 0.07577866 -0.06921874 0.02660573 0.0 0.16232508488
8 0.0 0.16232508488 -0.033842053 0.045218028 -0.022827804 0.023845317 0.0 0.124249965735
9 0.0 0.124249965735 -0.02211613 0.032084666 -0.02699063 0.03773564 0.0 0.174634486511
10 0.0 0.174634486511 -0.01979376 0.034854397 -0.036107242 0.07056531 0.0 0.575175762177
11 0.0 0.575175762177 -0.03452098 0.046055835 -0.051925894 0.07039055 0.0 0.771875114441
12 0.0 0.771875114441 -0.025946895 0.040090334 -0.06049362 0.12658806 0.0 1.17285169065
13 0.0 1.17285169065 -0.021766115 0.03315237 -0.20705001 0.117947325 0.0 2.00157693863
14 0.0 2.00157693863 -0.042597745 0.046707444 -0.21937433 0.2545502 0.0 2.00236111879
15 0.0 2.00236111879 -0.32550547 0.30829763 -1.1787822 1.2378151 -18.2514705467 24.1736344528
llvm/test/VISC/DNN_Benchmarks/benchmarks/vgg16_cifar10/data/tuner_confs_base.txt 0 → 100644
+ 38
− 0
View file @ e847a50b
+++++
conf1 1 0 90.19 0
1 gpu conv fp32 1 add fp32 1 relu fp32 1
2 gpu conv fp32 1 add fp32 1 relu fp32 1 pool_max fp32 1
3 gpu conv fp32 1 add fp32 1 relu fp32 1
4 gpu conv fp32 1 add fp32 1 relu fp32 1 pool_max fp32 1
5 gpu conv fp32 1 add fp32 1 relu fp32 1
6 gpu conv fp32 1 add fp32 1 relu fp32 1
7 gpu conv fp32 1 add fp32 1 relu fp32 1 pool_max fp32 1
8 gpu conv fp32 1 add fp32 1 relu fp32 1
9 gpu conv fp32 1 add fp32 1 relu fp32 1
10 gpu conv fp32 1 add fp32 1 relu fp32 1 pool_max fp32 1
11 gpu conv fp32 1 add fp32 1 relu fp32 1
12 gpu conv fp32 1 add fp32 1 relu fp32 1
13 gpu conv fp32 1 add fp32 1 relu fp32 1 pool_max fp32 1
14 gpu mul fp32 1 add fp32 1 relu fp32 1
15 gpu mul fp32 1 add fp32 1
16 gpu softmax fp32 1
-----
+++++
conf2 1.5 0 90.19 0
1 gpu conv fp16 1 add fp16 1 relu fp16 1
2 gpu conv fp16 1 add fp16 1 relu fp16 1 pool_max fp16 1
3 gpu conv fp16 1 add fp16 1 relu fp16 1
4 gpu conv fp16 1 add fp16 1 relu fp16 1 pool_max fp16 1
5 gpu conv fp16 1 add fp16 1 relu fp16 1
6 gpu conv fp16 1 add fp16 1 relu fp16 1
7 gpu conv fp16 1 add fp16 1 relu fp16 1 pool_max fp16 1
8 gpu conv fp16 1 add fp16 1 relu fp16 1
9 gpu conv fp16 1 add fp16 1 relu fp16 1
10 gpu conv fp16 1 add fp16 1 relu fp16 1 pool_max fp16 1
11 gpu conv fp16 1 add fp16 1 relu fp16 1
12 gpu conv fp16 1 add fp16 1 relu fp16 1
13 gpu conv fp16 1 add fp16 1 relu fp16 1 pool_max fp16 1
14 gpu mul fp16 1 add fp16 1 relu fp16 1
15 gpu mul fp16 1 add fp16 1
16 gpu softmax fp32 1
-----
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment