Skip to content

Commit ca37625

Browse files
unknownunknown
unknown
authored and
unknown
committed
Initial commit
0 parents  commit ca37625

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

44 files changed

+9471
-0
lines changed

‎3rd-party-licenses.txt‎

Lines changed: 373 additions & 0 deletions
Large diffs are not rendered by default.

‎LICENSE‎

Lines changed: 21 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,21 @@
1+
The MIT License (MIT)
2+
3+
Copyright (c) 2016 Robert Bosch LLC, USA.
4+
5+
Permission is hereby granted, free of charge, to any person obtaining a copy
6+
of this software and associated documentation files (the "Software"), to deal
7+
in the Software without restriction, including without limitation the rights
8+
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
9+
copies of the Software, and to permit persons to whom the Software is
10+
furnished to do so, subject to the following conditions:
11+
12+
The above copyright notice and this permission notice shall be included in
13+
all copies or substantial portions of the Software.
14+
15+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16+
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17+
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18+
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19+
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20+
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
21+
THE SOFTWARE.

‎README.md‎

Lines changed: 18 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,18 @@
1+
# DL-benchmarks
2+
3+
This is the companion code for DL benchmarking study reported in the paper titled as *Comparative Study of Deep Learning Software Frameworks* authored by *Soheil Bahrampour, Naveen Ramakrishnan, Lukas Schott, and Mohak Shah*. The paper can be found here http://arxiv.org/abs/1511.06435. The code allows the users to reproduce and extend the results reported in the study. More specifically, the code provides timings of forward run and forward+backward (gradient computation) run of several deep learning architecture using Caffe, Neon, TensorFlow, Theano, and Torch. The deep learning architectures used includes LeNet, AlexNet, LSTM, and a stacked AutoEncoder. Please cite the above paper when reporting, reproducing or extending the results.
4+
5+
6+
## Run the benchmarks
7+
See the readme file within each folder to run the experiments.
8+
9+
10+
## Dependencies
11+
Please refer to the corresponding github repository of each framework to install the framework and the corresponding dependencies. Most of the packages require Nvidia cuda and cuDNN to run on GPU. The provided codes have been tested on a system with Ubuntu 14.04 + Nvidia TitanX GPU with CUDA 7.5 and cuDNNv3.
12+
13+
14+
## License
15+
16+
DL-benchmarks is open-sourced under the MIT license. See the [LICENSE](LICENSE) file for details.
17+
18+
For a list of other open source components included in DL-benchmarks, see the file [3rd-party-licenses.txt](3rd-party-licenses.txt).

‎caffe/LeNet/lenet_train_test.prototxt‎

Lines changed: 175 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,175 @@
1+
name: "LeNet"
2+
3+
layer {
4+
type: "DummyData"
5+
name: "data"
6+
top: "data"
7+
include {
8+
phase: TRAIN
9+
}
10+
dummy_data_param {
11+
shape: { dim: 64 dim: 1 dim: 28 dim: 28}
12+
data_filler: { type: "constant"
13+
value: 1
14+
}
15+
}
16+
}
17+
layer {
18+
type: "DummyData"
19+
name: "label"
20+
top: "label"
21+
include {
22+
phase: TRAIN
23+
}
24+
dummy_data_param {
25+
shape: { dim: 64}
26+
data_filler: { type: "constant"
27+
value: 0
28+
}
29+
}
30+
}
31+
layer {
32+
name: "conv1"
33+
type: "Convolution"
34+
bottom: "data"
35+
top: "conv1"
36+
param {
37+
lr_mult: 1
38+
}
39+
param {
40+
lr_mult: 2
41+
}
42+
convolution_param {
43+
num_output: 20
44+
kernel_size: 5
45+
stride: 1
46+
weight_filler {
47+
type: "xavier"
48+
}
49+
bias_filler {
50+
type: "constant"
51+
}
52+
}
53+
}
54+
layer {
55+
name: "pool1"
56+
type: "Pooling"
57+
bottom: "conv1"
58+
top: "pool1"
59+
pooling_param {
60+
pool: MAX
61+
kernel_size: 2
62+
stride: 2
63+
}
64+
}
65+
layer {
66+
name: "tanh1"
67+
bottom: "pool1"
68+
top: "tanh1"
69+
type: "TanH"
70+
}
71+
layer {
72+
name: "conv2"
73+
type: "Convolution"
74+
bottom: "tanh1"
75+
top: "conv2"
76+
param {
77+
lr_mult: 1
78+
}
79+
param {
80+
lr_mult: 2
81+
}
82+
convolution_param {
83+
num_output: 50
84+
kernel_size: 5
85+
stride: 1
86+
weight_filler {
87+
type: "xavier"
88+
}
89+
bias_filler {
90+
type: "constant"
91+
}
92+
}
93+
}
94+
layer {
95+
name: "pool2"
96+
type: "Pooling"
97+
bottom: "conv2"
98+
top: "pool2"
99+
pooling_param {
100+
pool: MAX
101+
kernel_size: 2
102+
stride: 2
103+
}
104+
}
105+
layer {
106+
name: "tanh2"
107+
bottom: "pool2"
108+
top: "tanh2"
109+
type: "TanH"
110+
}
111+
layer {
112+
name: "ip1"
113+
type: "InnerProduct"
114+
bottom: "tanh2"
115+
top: "ip1"
116+
param {
117+
lr_mult: 1
118+
}
119+
param {
120+
lr_mult: 2
121+
}
122+
inner_product_param {
123+
num_output: 500
124+
weight_filler {
125+
type: "xavier"
126+
}
127+
bias_filler {
128+
type: "constant"
129+
}
130+
}
131+
}
132+
layer {
133+
name: "relu1"
134+
type: "ReLU"
135+
bottom: "ip1"
136+
top: "ip1"
137+
}
138+
layer {
139+
name: "ip2"
140+
type: "InnerProduct"
141+
bottom: "ip1"
142+
top: "ip2"
143+
param {
144+
lr_mult: 1
145+
}
146+
param {
147+
lr_mult: 2
148+
}
149+
inner_product_param {
150+
num_output: 10
151+
weight_filler {
152+
type: "xavier"
153+
}
154+
bias_filler {
155+
type: "constant"
156+
}
157+
}
158+
}
159+
layer {
160+
name: "accuracy"
161+
type: "Accuracy"
162+
bottom: "ip2"
163+
bottom: "label"
164+
top: "accuracy"
165+
include {
166+
phase: TEST
167+
}
168+
}
169+
layer {
170+
name: "loss"
171+
type: "SoftmaxWithLoss"
172+
bottom: "ip2"
173+
bottom: "label"
174+
top: "loss"
175+
}

‎caffe/LeNet/readme.txt‎

Lines changed: 6 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,6 @@
1+
to run timing on cpu (n threads only controllable once during setup) for 100 iterations do:
2+
~/git/caffe/build/tools/caffe time -model ./lenet_train_test.prototxt -iterations 100
3+
4+
to run timing on gpu
5+
~/git/caffe/build/tools/caffe time -model ./lenet_train_test.prototxt -iterations 100 -gpu 0
6+

‎caffe/alexnet/readme.txt‎

Lines changed: 10 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,10 @@
1+
To run caffe alexnet:
2+
3+
to run timing on cpu (n threads only controllable once during setup) for 30 iterations do:
4+
~/src/benchmarking_dl/proj-rtc13-dl/release/caffe/alexnet$ ~/git/caffe/build/tools/caffe time -model ./train_val.prototxt -iterations 30
5+
6+
To run on GPU:
7+
~/src/benchmarking_dl/proj-rtc13-dl/release/caffe/alexnet$ ~/git/caffe/build/tools/caffe time -model ./train_val.prototxt -iterations 30 -gpu 0
8+
9+
To run without LRN and without grouping
10+
~/src/benchmarking_dl/proj-rtc13-dl/release/caffe/alexnet$ ~/git/caffe/build/tools/caffe time -model ./train_val_noLRN_noGrouping.prototxt -iterations 30 -gpu 0

0 commit comments

Comments
 (0)