Skip to content

Instantly share code, notes, and snippets.

@AlexDBlack

AlexDBlack/gist:e02b99d259f610de61163640d9602a05 Secret

Created December 18, 2018 00:35
Show Gist options
  • Save AlexDBlack/e02b99d259f610de61163640d9602a05 to your computer and use it in GitHub Desktop.
Save AlexDBlack/e02b99d259f610de61163640d9602a05 to your computer and use it in GitHub Desktop.
Download ZIP
Raw
gistfile1.txt
========== Memory Information ==========
----- Version Information -----
Deeplearning4j Version 1.0.0-beta3
Deeplearning4j CUDA deeplearning4j-cuda-9.2
----- System Information -----
Operating System Microsoft Windows 10
CPU Intel(R) Core(TM) i7-5960X CPU @ 3.00GHz
CPU Cores - Physical 8
CPU Cores - Logical 16
Total System Memory 31.89 GB (34246447104)
Number of GPUs Detected 1
Name CC Total Memory Used Memory Free Memory
TITAN X (Pascal) 6.1 12 GB (12884901888) 7.62 GB (8179207373) 4.38 GB (4705694515)
----- ND4J Environment Information -----
Data Type FLOAT
backend CUDA
blas.vendor CUBLAS
os Windows 10
----- Memory Configuration -----
JVM Memory: XMX 7.09 GB (7611613184)
JVM Memory: current 856.50 MB (898105344)
JavaCPP Memory: Max Bytes 7.09 GB (7611613184)
JavaCPP Memory: Max Physical 14.18 GB (15223226368)
JavaCPP Memory: Current Bytes 221.53 MB (232285900)
JavaCPP Memory: Current Physical 6.42 GB (6894276608)
Periodic GC Enabled true
Periodic GC Frequency 100 ms
----- Workspace Information -----
Workspaces: # for current thread 4
Current thread workspaces:
Name State Size # Cycles
WS_LAYER_WORKING_MEM CLOSED 199.92 MB (209631313) 430
WS_ALL_LAYERS_ACT CLOSED 2.01 GB (2161293440) 20
WS_LAYER_ACT_0 CLOSED 199.92 MB (209631313) 110
WS_LAYER_ACT_1 CLOSED 199.92 MB (209631313) 100
Workspaces total size 2.60 GB (2790187379)
Helper Workspaces
CUDNN_WORKSPACE 25.00 MB (26216448)
----- Network Information -----
Network # Parameters 135079944
Parameter Memory 515.29 MB (540319776)
Parameter Gradients Memory 515.29 MB (540319776)
Updater Number of Elements 270159888
Updater Memory 1.01 GB (1080639552)
Updater Classes:
org.nd4j.linalg.learning.AdaDeltaUpdater
Params + Gradient + Updater Memory 1.51 GB (1620959328)
Iteration Count 10
Epoch Count 1
Backprop Type Standard
Workspace Mode: Training ENABLED
Workspace Mode: Inference ENABLED
Number of Layers 21
Layer Counts
ConvolutionLayer 13
DenseLayer 2
OutputLayer 1
SubsamplingLayer 5
Layer Parameter Breakdown
Idx Name Layer Type Layer # Parameters Layer Parameter Memory
1 0 ConvolutionLayer 1792 7 KB (7168)
2 1 ConvolutionLayer 36928 144.25 KB (147712)
3 2 SubsamplingLayer 0 0 B
4 3 ConvolutionLayer 73856 288.50 KB (295424)
5 4 ConvolutionLayer 147584 576.50 KB (590336)
6 5 SubsamplingLayer 0 0 B
7 6 ConvolutionLayer 295168 1.13 MB (1180672)
8 7 ConvolutionLayer 590080 2.25 MB (2360320)
9 8 ConvolutionLayer 590080 2.25 MB (2360320)
10 9 SubsamplingLayer 0 0 B
11 10 ConvolutionLayer 1180160 4.50 MB (4720640)
12 11 ConvolutionLayer 2359808 9.00 MB (9439232)
13 12 ConvolutionLayer 2359808 9.00 MB (9439232)
14 13 SubsamplingLayer 0 0 B
15 14 ConvolutionLayer 2359808 9.00 MB (9439232)
16 15 ConvolutionLayer 2359808 9.00 MB (9439232)
17 16 ConvolutionLayer 2359808 9.00 MB (9439232)
18 17 SubsamplingLayer 0 0 B
19 18 DenseLayer 102764544 392.02 MB (411058176)
20 19 DenseLayer 16781312 64.02 MB (67125248)
21 20 OutputLayer 819400 3.13 MB (3277600)
----- Layer Helpers - Memory Use -----
Total Helper Count 18
Helper Count w/ Memory 0
Total Helper Persistent Memory Use 0 B
----- Network Activations: Inferred Activation Shapes -----
Current Minibatch Size 16
Current Input Shape (Input 0) [16, 3, 224, 224]
Idx Name Layer Type Activations Type Activations Shape # Elements Memory
0 in InputVertex InputTypeConvolutional(h=224,w=224,c=3) [16, 3, 224, 224] 2408448 9.19 MB (9633792)
1 0 ConvolutionLayer InputTypeConvolutional(h=224,w=224,c=64) [16, 64, 224, 224] 51380224 196 MB (205520896)
2 1 ConvolutionLayer InputTypeConvolutional(h=224,w=224,c=64) [16, 64, 224, 224] 51380224 196 MB (205520896)
3 2 SubsamplingLayer InputTypeConvolutional(h=112,w=112,c=64) [16, 64, 112, 112] 12845056 49 MB (51380224)
4 3 ConvolutionLayer InputTypeConvolutional(h=112,w=112,c=128) [16, 128, 112, 112] 25690112 98 MB (102760448)
5 4 ConvolutionLayer InputTypeConvolutional(h=112,w=112,c=128) [16, 128, 112, 112] 25690112 98 MB (102760448)
6 5 SubsamplingLayer InputTypeConvolutional(h=56,w=56,c=128) [16, 128, 56, 56] 6422528 24.50 MB (25690112)
7 6 ConvolutionLayer InputTypeConvolutional(h=56,w=56,c=256) [16, 256, 56, 56] 12845056 49 MB (51380224)
8 7 ConvolutionLayer InputTypeConvolutional(h=56,w=56,c=256) [16, 256, 56, 56] 12845056 49 MB (51380224)
9 8 ConvolutionLayer InputTypeConvolutional(h=56,w=56,c=256) [16, 256, 56, 56] 12845056 49 MB (51380224)
10 9 SubsamplingLayer InputTypeConvolutional(h=28,w=28,c=256) [16, 256, 28, 28] 3211264 12.25 MB (12845056)
11 10 ConvolutionLayer InputTypeConvolutional(h=28,w=28,c=512) [16, 512, 28, 28] 6422528 24.50 MB (25690112)
12 11 ConvolutionLayer InputTypeConvolutional(h=28,w=28,c=512) [16, 512, 28, 28] 6422528 24.50 MB (25690112)
13 12 ConvolutionLayer InputTypeConvolutional(h=28,w=28,c=512) [16, 512, 28, 28] 6422528 24.50 MB (25690112)
14 13 SubsamplingLayer InputTypeConvolutional(h=14,w=14,c=512) [16, 512, 14, 14] 1605632 6.13 MB (6422528)
15 14 ConvolutionLayer InputTypeConvolutional(h=14,w=14,c=512) [16, 512, 14, 14] 1605632 6.13 MB (6422528)
16 15 ConvolutionLayer InputTypeConvolutional(h=14,w=14,c=512) [16, 512, 14, 14] 1605632 6.13 MB (6422528)
17 16 ConvolutionLayer InputTypeConvolutional(h=14,w=14,c=512) [16, 512, 14, 14] 1605632 6.13 MB (6422528)
18 17 SubsamplingLayer InputTypeConvolutional(h=7,w=7,c=512) [16, 512, 7, 7] 401408 1.53 MB (1605632)
19 18 DenseLayer InputTypeFeedForward(4096) [16, 4096] 65536 256 KB (262144)
20 19 DenseLayer InputTypeFeedForward(4096) [16, 4096] 65536 256 KB (262144)
21 20 OutputLayer InputTypeFeedForward(200) [16, 200] 3200 12.50 KB (12800)
Total Activations Memory 929.98 MB (975155712)
Total Activation Gradient Memory 929.97 MB (975142912)
----- Network Training Listeners -----
Number of Listeners 1
Listener 0 org.deeplearning4j.optimize.listeners.PerformanceListener@7fc6de5b
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment