Skip to content

Latest commit

 

History

History
68 lines (53 loc) · 2.12 KB

performance.md

File metadata and controls

68 lines (53 loc) · 2.12 KB
Raw

Performance Analysis

Recommend using Python in Anaconda for better performance.

Information

name value
Training Set MNIST 42000 28x28 gray
Batch Size 100
Iterations 2000
forward_times 2601
backward_times 2001
CPU i7-4790 CPU @ 3.60GHz

Python3 in Anaconda

Accuracy: 0.991667

Iter: 2000, Cost: 0.007383

name forward_time backward_time forward_mean backward_mean
data 0.073538 0.024317 0.000028 0.000012
conv1 50.427243 82.105529 0.019388 0.041032
pool1 95.274342 10.921718 0.036630 0.005458
conv2 71.583208 118.194660 0.027521 0.059068
pool2 14.343701 2.975322 0.005515 0.001487
fc3 9.543055 15.118490 0.003669 0.007555
relu3 1.383795 1.082115 0.000532 0.000541
pred 5.047948 6.669519 0.001941 0.003333
loss 0.458279 0.177457 0.000176 0.000089

Native Python3

Accuracy: 0.991667

Iter: 2000, Cost: 0.007383

name forward_time backward_time forward_mean backward_mean
data 0.039973 0.019009 0.000015 0.000009
conv1 76.094196 114.764927 0.029256 0.057354
pool1 72.345203 10.924688 0.027814 0.005460
conv2 370.081568 476.961305 0.142284 0.238361
pool2 9.614037 2.431894 0.003696 0.001215
fc3 83.233347 85.736211 0.032001 0.042847
relu3 0.765435 0.561773 0.000294 0.000281
pred 1.098624 1.267450 0.000422 0.000633
loss 0.276948 0.102970 0.000106 0.000051

Native Python2

Accuracy: 0.991667

Iter: 2000, Cost: 0.007383

name forward_time backward_time forward_mean backward_mean
data 0.039304 0.018812 0.000015 0.000009
conv1 77.984102 117.541489 0.029982 0.058741
pool1 73.096253 11.021775 0.028103 0.005508
conv2 373.498943 489.150597 0.143598 0.244453
pool2 10.090908 2.586956 0.003880 0.001293
fc3 83.812420 86.976251 0.032223 0.043466
relu3 0.766895 0.567209 0.000295 0.000283
pred 1.100621 1.267404 0.000423 0.000633
loss 0.278426 0.091963 0.000107 0.000046