Collective Knowledge Aggregator proof-of-concept
Crowd results Raw CK browser Graphs Reports Datasets Models Home

This page is outdated! New version is available here.


Share
b86d40a3849ac9a5
Date: 2017-04-24 09:24:02.715804
License:
Module: experiment.bench.dnn
Repo: upload
CID (DOI-like distributed ID): 1eb2f50d4620903e:b86d40a3849ac9a5

DNN engine and model evaluation statistics per layer (crowd-tuning)


DNN engine name: BVLC Caffe framework (opencl,viennacl)
DNN engine version: master-880111e
DNN engine type: opencl
DNN engine dependencies: cmake=3.7.2
compiler=19.00.23918
lib-blas=0.2.19-85636ff
lib-boost=1.62.0
lib-gflags=2.2.0
lib-glog=development-da816ea
lib-hdf5=1.10.0
lib-opencl=6.3.0.1904
lib-opencv=3.2.0
lib-protobuf-host=3.1.0
lib-viennacl=master-5494bef
DNN model name: Caffe model (net and weights) (bvlc, alexnet)
DNN model version: trunk
Batch size: 1
Platform: LENOVO ThinkPad X240 (20ALCTO1WW)
OS: Windows 10
CPU: Intel(R) Core(TM) i5-4210U CPU @ 1.70GHz
GPU: Intel(R) HD Graphics Family

Name Direction Min time (ms.): Expected time (ms.): Max time (ms.): Repetitions:
16: fc6 backward 185.9 187.2 188.8 3
19: fc7 backward 91.2 91.2 91.6 3
22: fc8 backward 24.3 25.2 25.9 3
16: fc6 forward 14.8 14.8 14.9 3
05: conv2 backward 13.5 13.7 14.2 3
09: conv3 backward 12.7 14.0 14.7 3
11: conv4 backward 11.5 11.7 11.9 3
01: conv1 backward 11.4 13.0 14.0 3
05: conv2 forward 9.8 9.8 9.8 3
13: conv5 backward 7.6 8.4 9.2 3
09: conv3 forward 6.9 6.9 6.9 3
19: fc7 forward 6.9 7.0 7.1 3
11: conv4 forward 6.3 6.3 6.4 3
13: conv5 forward 5.2 5.3 5.3 3
01: conv1 forward 5.2 5.2 5.2 3
15: pool5 backward 4.3 4.6 5.0 3
03: norm1 backward 2.7 2.7 2.7 3
04: pool1 backward 2.6 2.7 2.7 3
22: fc8 forward 2.2 2.2 2.4 3
07: norm2 backward 2.1 3.8 4.0 3
23: prob backward 2.0 2.3 2.8 3
08: pool2 backward 1.9 1.9 4.8 3
21: drop7 forward 1.2 1.3 1.6 3
18: drop6 forward 1.2 1.2 1.3 3
03: norm1 forward 0.7 0.7 0.7 3
07: norm2 forward 0.5 0.6 0.6 3
04: pool1 forward 0.4 0.4 0.5 3
08: pool2 forward 0.4 0.4 0.4 3
06: relu2 forward 0.4 0.4 0.4 3
17: relu6 forward 0.4 0.4 0.4 3
23: prob forward 0.4 0.4 0.5 3
20: relu7 forward 0.4 0.4 0.4 3
02: relu1 forward 0.3 0.4 0.4 3
15: pool5 forward 0.3 0.4 0.4 3
12: relu4 forward 0.3 0.3 0.4 3
10: relu3 forward 0.3 0.3 0.4 3
14: relu5 forward 0.3 0.3 0.3 3
10: relu3 backward 0.3 0.3 0.3 3
00: input backward 0.3 0.3 0.3 3
02: relu1 backward 0.3 0.3 0.3 3
14: relu5 backward 0.2 0.3 0.3 3
06: relu2 backward 0.2 0.3 0.3 3
12: relu4 backward 0.2 0.3 0.3 3
18: drop6 backward 0.2 0.2 0.2 3
17: relu6 backward 0.1 0.1 0.2 3
21: drop7 backward 0.1 0.2 0.2 3
20: relu7 backward 0.1 0.1 0.2 3
00: input forward 0.1 0.1 0.1 3
View entry in raw format

Developed by Grigori Fursin           
Implemented as a CK workflow 		                         
   
                      Hosted at