TDA2EXEVM: [TIDL] Dense model is faster than Sparse model

Taro Shibata

Expert 8110 points

Part Number: TDA2EXEVM

Hi,

Here are another query about TIDL_OD performance on TDA2x.

- My customer is evaluating TIDL_OD performance by changing sparsity settings.

- Prepared following two caffe-jacinto models

(A) Sparsity enabled in part of layers (same configuration with the one in TIDL User’s Guide)

(B) Dense configuration. Sparsity disabled in all layers (all conv2dKernelType are set to 1)

- Assuming the number of MACs is larger in case (B) than (A), customer expected (A) is faster than (B).

- But the result is unexpected one. “ALG_TIDL” processing time on EVE in (B) is shorter than in (A)

(A) 360 ms

(B) 316 ms

Are there any reason why (B) is faster than (A)?

I referred below thread. In my customer case, image size is 768 x 320.

Is this the same case?

https://e2e.ti.com/support/processors/f/791/p/694669/2561198?tisearch=e2e-quicksearch&keymatch=[TIDL]%20Dense%20model%20is%20faster%20than%20Sparse%20model#2561198

FYI, I attach configuration files, logs.

logs_cfgs.zip

Thanks & Regards,

-Shibata

over 5 years ago

0 Praveen Eppa1 over 5 years ago

TI__Genius 17580 points

Hi Shibata,

>> Assuming the number of MACs is larger in case (B) than (A), customer expected (A) is faster than (B).

Can you print "Total Giga Macs" of both the models in the import tool and confirm that MACs is larger in case (B) than (A), and then use the same import config file for both the models (don't change all conv2dKernelType are set to 1) and check profile numbers again.

Thanks,

Praveen

0 Taro Shibata over 5 years ago in reply to Praveen Eppa1

TI__Expert 8110 points

Hello Praveen,

Customer captured the import tool logs for both (A) and (B).

Please find attachments.

nt-1a84591e31034fac832d29ed8584666c1431701329_code-editor-fs">Fullscreen ://e2e.ti.com/cfs-filesystemfile/__key/communityserver-discussions-components-files/791/Log_5F00_Importtool_5F005F00_JDetNet_5F00_768x320_5F00_ini_5F00_A_2D00_manual.txt?_=637205558560000000" Log_Importtool__JDetNet_768x320_ini_A-manual.txt">Log_Importtool__JDetNet_768x320_ini_A-manual.txt ://e2e.ti.com/cfs-filesystemfile/__key/communityserver-discussions-components-files/791/Log_5F00_Importtool_5F005F00_JDetNet_5F00_768x320_5F00_ini_5F00_A_2D00_manual.txt?_=637205558560000000" Log_Importtool__JDetNet_768x320_ini_A-manual.txt">Download a84591e31034fac832d29ed8584666c1431701329_code-editor" 922px;min-width:1230px;margin:0px;" Network File : ..\..\test\testvecs\config\caffe-jacinto\set\deploy_v2.prototxt : ..\..\test\testvecs\config\caffe-jacinto\set\voc0712_ssdJacintoNetV2_iter_120000.caffemodel : ..\..\test\testvecs\config\tidl_models\jdetnet768_test\tidl_net_jdetNet_ssd_768_ini_A-manual.bin : ..\..\test\testvecs\config\tidl_models\jdetnet768_test\tidl_param_jdetNet_ssd_768_ini_A-manual.bin : ssdJacintoNetV2_deploy , data 0, -1 , 1 , x , x , x , x , x , x , x , x , 0 , 0 , 0 , 0 , 0 , 1 , 3 , 320 , 768 , 0 , , data/bias 1, 1 , 1 , 0 , x , x , x , x , x , x , x , 1 , 1 , 3 , 320 , 768 , 1 , 3 , 320 , 768 , 737280 , , conv1a 1, 1 , 1 , 1 , x , x , x , x , x , x , x , 2 , 1 , 3 , 320 , 768 , 1 , 32 , 160 , 384 , 147456000 , , conv1b 1, 1 , 1 , 2 , x , x , x , x , x , x , x , 3 , 1 , 32 , 160 , 384 , 1 , 32 , 80 , 192 , 141557760 , , res2a_branch2a 1, 1 , 1 , 3 , x , x , x , x , x , x , x , 4 , 1 , 32 , 80 , 192 , 1 , 64 , 80 , 192 , 283115520 , , res2a_branch2b 1, 1 , 1 , 4 , x , x , x , x , x , x , x , 5 , 1 , 64 , 80 , 192 , 1 , 64 , 40 , 96 , 141557760 , , res3a_branch2a 1, 1 , 1 , 5 , x , x , x , x , x , x , x , 6 , 1 , 64 , 40 , 96 , 1 , 128 , 40 , 96 , 283115520 , , res3a_branch2b 1, 1 , 1 , 6 , x , x , x , x , x , x , x , 7 , 1 , 128 , 40 , 96 , 1 , 128 , 20 , 48 , 141557760 , , res4a_branch2a 1, 1 , 1 , 7 , x , x , x , x , x , x , x , 8 , 1 , 128 , 20 , 48 , 1 , 256 , 20 , 48 , 283115520 , , res4a_branch2b 1, 1 , 1 , 8 , x , x , x , x , x , x , x , 9 , 1 , 256 , 20 , 48 , 1 , 256 , 20 , 48 , 141557760 , , pool4 1, 1 , 1 , 9 , x , x , x , x , x , x , x , 10 , 1 , 256 , 20 , 48 , 1 , 256 , 10 , 24 , 245760 , , res5a_branch2a 1, 1 , 1 , 10 , x , x , x , x , x , x , x , 11 , 1 , 256 , 10 , 24 , 1 , 512 , 10 , 24 , 283115520 , , res5a_branch2b 1, 1 , 1 , 11 , x , x , x , x , x , x , x , 12 , 1 , 512 , 10 , 24 , 1 , 512 , 10 , 24 , 141557760 , , pool6 1, 1 , 1 , 12 , x , x , x , x , x , x , x , 13 , 1 , 512 , 10 , 24 , 1 , 512 , 5 , 12 , 122880 , , pool7 1, 1 , 1 , 13 , x , x , x , x , x , x , x , 14 , 1 , 512 , 5 , 12 , 1 , 512 , 3 , 6 , 36864 , , pool8 1, 1 , 1 , 14 , x , x , x , x , x , x , x , 15 , 1 , 512 , 3 , 6 , 1 , 512 , 2 , 3 , 12288 , , ctx_output1 1, 1 , 1 , 9 , x , x , x , x , x , x , x , 16 , 1 , 256 , 20 , 48 , 1 , 256 , 20 , 48 , 62914560 , , ctx_output2 1, 1 , 1 , 12 , x , x , x , x , x , x , x , 17 , 1 , 512 , 10 , 24 , 1 , 256 , 10 , 24 , 31457280 , , ctx_output3 1, 1 , 1 , 13 , x , x , x , x , x , x , x , 18 , 1 , 512 , 5 , 12 , 1 , 256 , 5 , 12 , 7864320 , , ctx_output4 1, 1 , 1 , 14 , x , x , x , x , x , x , x , 19 , 1 , 512 , 3 , 6 , 1 , 256 , 3 , 6 , 2359296 , , ctx_output5 1, 1 , 1 , 15 , x , x , x , x , x , x , x , 20 , 1 , 512 , 2 , 3 , 1 , 256 , 2 , 3 , 786432 , , ctx_output1/relu_mbox_loc 1, 1 , 1 , 16 , x , x , x , x , x , x , x , 21 , 1 , 256 , 20 , 48 , 1 , 16 , 20 , 48 , 3932160 , , ctx_output1/relu_mbox_loc_perm 1, 1 , 1 , 21 , x , x , x , x , x , x , x , 22 , 1 , 16 , 20 , 48 , 1 , 1 , 1 , 15360 , 1 , , ctx_output1/relu_mbox_conf 1, 1 , 1 , 16 , x , x , x , x , x , x , x , 23 , 1 , 256 , 20 , 48 , 1 , 84 , 20 , 48 , 20643840 , , ctx_output1/relu_mbox_conf_perm 1, 1 , 1 , 23 , x , x , x , x , x , x , x , 24 , 1 , 84 , 20 , 48 , 1 , 1 , 1 , 80640 , 1 , , ctx_output2/relu_mbox_loc 1, 1 , 1 , 17 , x , x , x , x , x , x , x , 26 , 1 , 256 , 10 , 24 , 1 , 24 , 10 , 24 , 1474560 , , ctx_output2/relu_mbox_loc_perm 1, 1 , 1 , 26 , x , x , x , x , x , x , x , 27 , 1 , 24 , 10 , 24 , 1 , 1 , 1 , 5760 , 1 , , ctx_output2/relu_mbox_conf 1, 1 , 1 , 17 , x , x , x , x , x , x , x , 28 , 1 , 256 , 10 , 24 , 1 , 126 , 10 , 24 , 7741440 , , ctx_output2/relu_mbox_conf_perm 1, 1 , 1 , 28 , x , x , x , x , x , x , x , 29 , 1 , 126 , 10 , 24 , 1 , 1 , 1 , 30240 , 1 , , ctx_output3/relu_mbox_loc 1, 1 , 1 , 18 , x , x , x , x , x , x , x , 31 , 1 , 256 , 5 , 12 , 1 , 24 , 5 , 12 , 368640 , , ctx_output3/relu_mbox_loc_perm 1, 1 , 1 , 31 , x , x , x , x , x , x , x , 32 , 1 , 24 , 5 , 12 , 1 , 1 , 1 , 1440 , 1 , , ctx_output3/relu_mbox_conf 1, 1 , 1 , 18 , x , x , x , x , x , x , x , 33 , 1 , 256 , 5 , 12 , 1 , 126 , 5 , 12 , 1935360 , , ctx_output3/relu_mbox_conf_perm 1, 1 , 1 , 33 , x , x , x , x , x , x , x , 34 , 1 , 126 , 5 , 12 , 1 , 1 , 1 , 75 1 �̃t�@�C��R�s�[��܂��B file .\tempDir\qunat_stats_config.txt ! , 0, -1 , 1 , x , x , x , x , x , x , x , x , 0 , 0 , 0 , 0 , 0 , 1 , 3 , 320 , 768 , , 1, 1 , 1 , 0 , x , x , x , x , x , x , x , 1 , 1 , 3 , 320 , 768 , 1 , 3 , 320 , 768 , , 1, 1 , 1 , 1 , x , x , x , x , x , x , x , 2 , 1 , 3 , 320 , 768 , 1 , 32 , 160 , 384 , , 1, 1 , 1 , 2 , x , x , x , x , x , x , x , 3 , 1 , 32 , 160 , 384 , 1 , 32 , 80 , 192 , , 1, 1 , 1 , 3 , x , x , x , x , x , x , x , 4 , 1 , 32 , 80 , 192 , 1 , 64 , 80 , 192 , , 1, 1 , 1 , 4 , x , x , x , x , x , x , x , 5 , 1 , 64 , 80 , 192 , 1 , 64 , 40 , 96 , , 1, 1 , 1 , 5 , x , x , x , x , x , x , x , 6 , 1 , 64 , 40 , 96 , 1 , 128 , 40 , 96 , , 1, 1 , 1 , 6 , x , x , x , x , x , x , x , 7 , 1 , 128 , 40 , 96 , 1 , 128 , 20 , 48 , , 1, 1 , 1 , 7 , x , x , x , x , x , x , x , 8 , 1 , 128 , 20 , 48 , 1 , 256 , 20 , 48 , , 1, 1 , 1 , 8 , x , x , x , x , x , x , x , 9 , 1 , 256 , 20 , 48 , 1 , 256 , 20 , 48 , , 1, 1 , 1 , 9 , x , x , x , x , x , x , x , 10 , 1 , 256 , 20 , 48 , 1 , 256 , 10 , 24 , , 1, 1 , 1 , 10 , x , x , x , x , x , x , x , 11 , 1 , 256 , 10 , 24 , 1 , 512 , 10 , 24 , , 1, 1 , 1 , 11 , x , x , x , x , x , x , x , 12 , 1 , 512 , 10 , 24 , 1 , 512 , 10 , 24 , , 1, 1 , 1 , 12 , x , x , x , x , x , x , x , 13 , 1 , 512 , 10 , 24 , 1 , 512 , 5 , 12 , , 1, 1 , 1 , 13 , x , x , x , x , x , x , x , 14 , 1 , 512 , 5 , 12 , 1 , 512 , 3 , 6 , , 1, 1 , 1 , 14 , x , x , x , x , x , x , x , 15 , 1 , 512 , 3 , 6 , 1 , 512 , 2 , 3 , , 1, 1 , 1 , 9 , x , x , x , x , x , x , x , 16 , 1 , 256 , 20 , 48 , 1 , 256 , 20 , 48 , , 1, 1 , 1 , 12 , x , x , x , x , x , x , x , 17 , 1 , 512 , 10 , 24 , 1 , 256 , 10 , 24 , , 1, 1 , 1 , 13 , x , x , x , x , x , x , x , 18 , 1 , 512 , 5 , 12 , 1 , 256 , 5 , 12 , , 1, 1 , 1 , 14 , x , x , x , x , x , x , x , 19 , 1 , 512 , 3 , 6 , 1 , 256 , 3 , 6 , , 1, 1 , 1 , 15 , x , x , x , x , x , x , x , 20 , 1 , 512 , 2 , 3 , 1 , 256 , 2 , 3 , , 1, 1 , 1 , 16 , x , x , x , x , x , x , x , 21 , 1 , 256 , 20 , 48 , 1 , 16 , 20 , 48 , , 1, 1 , 1 , 21 , x , x , x , x , x , x , x , 22 , 1 , 16 , 20 , 48 , 1 , 1 , 1 ,15360 , , 1, 1 , 1 , 16 , x , x , x , x , x , x , x , 23 , 1 , 256 , 20 , 48 , 1 , 84 , 20 , 48 , , 1, 1 , 1 , 23 , x , x , x , x , x , x , x , 24 , 1 , 84 , 20 , 48 , 1 , 1 , 1 ,80640 , , 1, 1 , 1 , 17 , x , x , x , x , x , x , x , 26 , 1 , 256 , 10 , 24 , 1 , 24 , 10 , 24 , , 1, 1 , 1 , 26 , x , x , x , x , x , x , x , 27 , 1 , 24 , 10 , 24 , 1 , 1 , 1 , 5760 , , 1, 1 , 1 , 17 , x , x , x , x , x , x , x , 28 , 1 , 256 , 10 , 24 , 1 , 126 , 10 , 24 , , 1, 1 , 1 , 28 , x , x , x , x , x , x , x , 29 , 1 , 126 , 10 , 24 , 1 , 1 , 1 ,30240 , , 1, 1 , 1 , 18 , x , x , x , x , x , x , x , 31 , 1 , 256 , 5 , 12 , 1 , 24 , 5 , 12 , , 1, 1 , 1 , 31 , x , x , x , x , x , x , x , 32 , 1 , 24 , 5 , 12 , 1 , 1 , 1 , 1440 , , 1, 1 , 1 , 18 , x , x , x , x , x , x , x , 33 , 1 , 256 , 5 , 12 , 1 , 126 , 5 , 12 , , 1, 1 , 1 , 33 , x , x , x , x , x , x , x , 34 , 1 , 126 , 5 , 12 , 1 , 1 , 1 , 7560 , , 1, 1 , 1 , 19 , x , x , x , x , x , x , x , 36 , 1 , 256 , 3 , 6 , 1 , 24 , 3 , 6 , , 1, 1 , 1 , 36 , x , x , x , x , x , x , x , 37 , 1 , 24 , 3 , 6 , 1 , 1 , 1 , 432 , , 1, 1 , 1 , 19 , x , x , x , x , x , x , x , 38 , 1 , 256 , 3 , 6 , 1 , 126 , 3 , 6 , , 1, 1 , 1 , 38 , x , x , x , x , x , x , x , 39 , 1 , 126 , 3 , 6 , 1 , 1 , 1 , 2268 , , 1, 1 , 1 , 20 , x , x , x , x , x , x , x , 41 , 1 , 256 , 2 , 3 , 1 , 16 , 2 , 3 , , 1, 1 , 1 , 41 , x , x , x , x , x , x , x , 42 , 1 , 16 , 2 , 3 , 1 , 1 , 1 , 96 , , 1, 1 , 1 , 20 , x , x , x , x , x , x , x , 43 , 1 , 256 , 2 , 3 , 1 , 84 , 2 , 3 , , 1, 1 , 1 , 43 , x , x , x , x , x , x , x , 44 , 1 , 84 , 2 , 3 , 1 , 1 , 1 , 504 , , 1, 5 , 1 , 22 , 27 , 32 , 37 , 42 , x , x , x , 46 , 1 , 1 , 1 ,15360 , 1 , 1 , 1 ,23088 , , 1, 5 , 1 , 24 , 29 , 34 , 39 , 44 , x , x , x , 47 , 1 , 1 , 1 ,80640 , 1 , 1 , 1 ,121212 , , 1, 2 , 1 , 46 , 47 , x , x , x , x , x , x , 49 , 1 , 1 , 1 ,23088 , 1 , 1 , 1 , 560 , , 0, 1 , -1 , 49 , x , x , x , x , x , x , x , 0 , 1 , 1 , 1 , 560 , 0 , 0 , 0 , 0 , ,inBlkHeight ,inBlkPitch ,outBlkWidth ,outBlkHeight,outBlkPitch ,numInChs ,numOutChs ,numProcInChs,numLclInChs ,numLclOutChs,numProcItrs ,numAccItrs ,numHorBlock ,numVerBlock ,inBlkChPitch,outBlkChPitc,alignOrNot 72 72 32 32 32 3 32 3 1 8 1 3 12 5 5184 1024 1 34 40 32 32 32 8 8 8 4 8 1 2 12 5 1360 1024 1 22 40 32 20 32 32 64 32 8 8 1 4 6 4 880 640 1 22 40 32 20 32 16 16 16 8 8 1 2 6 4 880 640 1 22 40 32 20 32 64 128 64 8 8 1 8 3 2 880 640 1 22 40 32 20 32 32 32 32 8 8 1 4 3 2 880 640 1 22 56 48 20 48 128 256 128 7 8 1 19 1 1 1232 960 1 22 56 48 20 48 64 64 64 7 8 1 10 1 1 1232 960 1 12 40 32 10 32 256 512 256 8 8 1 32 1 1 480 320 1 12 40 32 10 32 128 128 128 8 8 1 16 1 1 480 320 1 4 48 48 4 48 256 256 256 32 8 1 8 1 5 192 192 1 10 24 24 10 24 512 256 512 32 32 1 16 1 1 240 240 1 5 12 12 5 12 512 256 512 32 32 1 16 1 1 60 60 1 3 6 6 3 6 512 256 512 32 32 1 16 1 1 18 18 1 2 3 3 2 3 512 256 512 32 32 1 16 1 1 6 6 1 4 48 48 4 48 256 16 256 32 8 1 8 1 5 192 192 1 4 48 48 4 48 256 88 256 32 8 1 8 1 5 192 192 1 10 24 24 10 24 256 24 256 32 24 1 8 1 1 240 240 1 10 24 24 10 24 256 128 256 32 32 1 8 1 1 240 240 1 5 12 12 5 12 256 24 256 32 24 1 8 1 1 60 60 1 5 12 12 5 12 256 128 256 32 32 1 8 1 1 60 60 1 3 6 6 3 6 256 24 256 32 24 1 8 1 1 18 18 1 3 6 6 3 6 256 128 256 32 32 1 8 1 1 18 18 1 2 3 3 2 3 256 16 256 32 16 1 8 1 1 6 6 1 2 3 3 2 3 256 96 256 32 32 1 8 1 1 6 6 1 254 , TIDL_BatchNormLayer , PASSED #MMACs = 0.74, 0.74, Sparsity : 0.00 5662 , TIDL_ConvolutionLayer, PASSED #MMACs = 147.46, 127.80, Sparsity : 13.33 6568 , TIDL_ConvolutionLayer, PASSED #MMACs = 141.56, 122.14, Sparsity : 13.72 11145 , TIDL_ConvolutionLayer, PASSED #MMACs = 283.12, 270.03, Sparsity : 4.62 11360 , TIDL_ConvolutionLayer, PASSED #MMACs = 141.56, 138.24, Sparsity : 2.34 13400 , TIDL_ConvolutionLayer, PASSED #MMACs = 283.12, 271.46, Sparsity : 4.12 15675 , TIDL_ConvolutionLayer, PASSED #MMACs = 141.56, 138.12, Sparsity : 2.43 16548 , TIDL_ConvolutionLayer, PASSED #MMACs = 283.12, 263.52, Sparsity : 6.92 14680 , TIDL_ConvolutionLayer, PASSED #MMACs = 141.56, 136.82, Sparsity : 3.35 PASSED #MMACs = 0.06, 0.06, Sparsity : 0.00 21005 , TIDL_ConvolutionLayer, PASSED #MMACs = 283.12, 254.26, Sparsity : 10.19 5819 , TIDL_ConvolutionLayer, PASSED #MMACs = 141.56, 119.73, Sparsity : 15.42 PASSED #MMACs = 0.03, 0.03, Sparsity : 0.00 PASSED #MMACs = 0.01, 0.01, Sparsity : 0.00 PASSED #MMACs = 0.00, 0.00, Sparsity : 0.00 21250 , TIDL_ConvolutionLayer, PASSED #MMACs = 62.91, 62.91, Sparsity : 0.00 15487 , TIDL_ConvolutionLayer, PASSED #MMACs = 31.46, 31.46, Sparsity : 0.00 20361 , TIDL_ConvolutionLayer, PASSED #MMACs = 7.86, 7.86, Sparsity : 0.00 22575 , TIDL_ConvolutionLayer, PASSED #MMACs = 2.36, 2.36, Sparsity : 0.00 27565 , TIDL_ConvolutionLayer, PASSED #MMACs = 0.79, 0.79, Sparsity : 0.00 4644 , TIDL_ConvolutionLayer, PASSED #MMACs = 3.93, 3.93, Sparsity : 0.00 PASSED #MMACs = 0.02, 0.02, Sparsity : 0.00 3760 , TIDL_ConvolutionLayer, PASSED #MMACs = 21.63, 21.63, Sparsity : 0.00 PASSED #MMACs = 0.08, 0.08, Sparsity : 0.00 8166 , TIDL_ConvolutionLayer, PASSED #MMACs = 1.47, 1.47, Sparsity : 0.00 PASSED #MMACs = 0.01, 0.01, Sparsity : 0.00 3123 , TIDL_ConvolutionLayer, PASSED #MMACs = 7.86, 7.86, Sparsity : 0.00 PASSED #MMACs = 0.03, 0.03, Sparsity : 0.00 10031 , TIDL_ConvolutionLayer, PASSED #MMACs = 0.37, 0.37, Sparsity : 0.00 PASSED #MMACs = 0.00, 0.00, Sparsity : 0.00 2884 , TIDL_ConvolutionLayer, PASSED #MMACs = 1.97, 1.97, Sparsity : 0.00 PASSED #MMACs = 0.01, 0.01, Sparsity : 0.00 8777 , TIDL_ConvolutionLayer, PASSED #MMACs = 0.11, 0.11, Sparsity : 0.00 PASSED #MMACs = 0.00, 0.00, Sparsity : 0.00 3649 , TIDL_ConvolutionLayer, PASSED #MMACs = 0.59, 0.59, Sparsity : 0.00 PASSED #MMACs = 0.00, 0.00, Sparsity : 0.00 9392 , TIDL_ConvolutionLayer, PASSED #MMACs = 0.02, 0.02, Sparsity : 0.00 PASSED #MMACs = 0.00, 0.00, Sparsity : 0.00 4266 , TIDL_ConvolutionLayer, PASSED #MMACs = 0.15, 0.15, Sparsity : 0.00 PASSED #MMACs = 0.00, 0.00, Sparsity : 0.00 4662 , TIDL_ConcatLayer, PASSED #MMACs = 0.00, 0.00, Sparsity : -1.#J 2895 , TIDL_ConcatLayer, PASSED #MMACs = 0.00, 0.00, Sparsity : -1.#J 0.00, 0.00, Sparsity : 0.00 , ctx_output4/relu_mbox_loc 1, 1 , 1 , 19 , x , x , x , x , x , x , x , 36 , 1 , 256 , 3 , 6 , 1 , 24 , 3 , 6 , 110592 , , ctx_output4/relu_mbox_loc_perm 1, 1 , 1 , 36 , x , x , x , x , x , x , x , 37 , 1 , 24 , 3 , 6 , 1 , 1 , 1 , 432 , 1 , , ctx_output4/relu_mbox_conf 1, 1 , 1 , 19 , x , x , x , x , x , x , x , 38 , 1 , 256 , 3 , 6 , 1 , 126 , 3 , 6 , 580608 , , ctx_output4/relu_mbox_conf_perm 1, 1 , 1 , 38 , x , x , x , x , x , x , x , 39 , 1 , 126 , 3 , 6 , 1 , 1 , 1 , 2268 , 1 , , ctx_output5/relu_mbox_loc 1, 1 , 1 , 20 , x , x , x , x , x , x , x , 41 , 1 , 256 , 2 , 3 , 1 , 16 , 2 , 3 , 24576 , , ctx_output5/relu_mbox_loc_perm 1, 1 , 1 , 41 , x , x , x , x , x , x , x , 42 , 1 , 16 , 2 , 3 , 1 , 1 , 1 , 96 , 1 , , ctx_output5/relu_mbox_conf 1, 1 , 1 , 20 , x , x , x , x , x , x , x , 43 , 1 , 256 , 2 , 3 , 1 , 84 , 2 , 3 , 129024 , , ctx_output5/relu_mbox_conf_perm 1, 1 , 1 , 43 , x , x , x , x , x , x , x , 44 , 1 , 84 , 2 , 3 , 1 , 1 , 1 , 504 , 1 , , mbox_loc 1, 5 , 1 , 22 , 27 , 32 , 37 , 42 , x , x , x , 46 , 1 , 1 , 1 , 15360 , 1 , 1 , 1 , 23088 , 1 , , mbox_conf 1, 5 , 1 , 24 , 29 , 34 , 39 , 44 , x , x , x , 47 , 1 , 1 , 1 , 80640 , 1 , 1 , 1 , 121212 , 1 , , detection_out 2, 2 , 1 , 46 , 47 , x , x , x , x , x , x , 49 , 1 , 1 , 1 , 23088 , 1 , 1 , 1 , 560 , 1 , = j('#fragment-1a84591e31034fac832d29ed8584666c1431701329_code-editor-fs'); = j('#fragment-1a84591e31034fac832d29ed8584666c1431701329_code-editor'); function(){ .evolutionCodeEditor('fullscreen')) { .evolutionCodeEditor('fullscreen', false); .evolutionCodeEditor('fullscreen', true);

Fullscreen Log_Importtool_JDetNet_768x320_ini_B-dense.txt Download

Caffe Network File : ..\..\test\testvecs\config\caffe-jacinto\set\deploy_v2.prototxt  
Caffe Model File   : ..\..\test\testvecs\config\caffe-jacinto\set\voc0712_ssdJacintoNetV2_iter_120000.caffemodel  
TIDL Network File  : ..\..\test\testvecs\config\tidl_models\jdetnet768_test\tidl_net_jdetNet_ssd_768_ini_B-dense.bin  
TIDL Model File    : ..\..\test\testvecs\config\tidl_models\jdetnet768_test\tidl_param_jdetNet_ssd_768_ini_B-dense.bin  
Name of the Network : ssdJacintoNetV2_deploy 
Num Inputs :               1 
Could not find detection_out Params
 Num of Layer Detected :  50 
  0, TIDL_DataLayer                , data                                      0,  -1 ,  1 ,   x ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  0 ,       0 ,       0 ,       0 ,       0 ,       1 ,       3 ,     320 ,     768 ,         0 ,
  1, TIDL_BatchNormLayer           , data/bias                                 1,   1 ,  1 ,   0 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  1 ,       1 ,       3 ,     320 ,     768 ,       1 ,       3 ,     320 ,     768 ,    737280 ,
  2, TIDL_ConvolutionLayer         , conv1a                                    1,   1 ,  1 ,   1 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  2 ,       1 ,       3 ,     320 ,     768 ,       1 ,      32 ,     160 ,     384 , 147456000 ,
  3, TIDL_ConvolutionLayer         , conv1b                                    1,   1 ,  1 ,   2 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  3 ,       1 ,      32 ,     160 ,     384 ,       1 ,      32 ,      80 ,     192 , 141557760 ,
  4, TIDL_ConvolutionLayer         , res2a_branch2a                            1,   1 ,  1 ,   3 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  4 ,       1 ,      32 ,      80 ,     192 ,       1 ,      64 ,      80 ,     192 , 283115520 ,
  5, TIDL_ConvolutionLayer         , res2a_branch2b                            1,   1 ,  1 ,   4 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  5 ,       1 ,      64 ,      80 ,     192 ,       1 ,      64 ,      40 ,      96 , 141557760 ,
  6, TIDL_ConvolutionLayer         , res3a_branch2a                            1,   1 ,  1 ,   5 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  6 ,       1 ,      64 ,      40 ,      96 ,       1 ,     128 ,      40 ,      96 , 283115520 ,
  7, TIDL_ConvolutionLayer         , res3a_branch2b                            1,   1 ,  1 ,   6 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  7 ,       1 ,     128 ,      40 ,      96 ,       1 ,     128 ,      20 ,      48 , 141557760 ,
  8, TIDL_ConvolutionLayer         , res4a_branch2a                            1,   1 ,  1 ,   7 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  8 ,       1 ,     128 ,      20 ,      48 ,       1 ,     256 ,      20 ,      48 , 283115520 ,
  9, TIDL_ConvolutionLayer         , res4a_branch2b                            1,   1 ,  1 ,   8 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  9 ,       1 ,     256 ,      20 ,      48 ,       1 ,     256 ,      20 ,      48 , 141557760 ,
 10, TIDL_PoolingLayer             , pool4                                     1,   1 ,  1 ,   9 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 10 ,       1 ,     256 ,      20 ,      48 ,       1 ,     256 ,      10 ,      24 ,    245760 ,
 11, TIDL_ConvolutionLayer         , res5a_branch2a                            1,   1 ,  1 ,  10 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 11 ,       1 ,     256 ,      10 ,      24 ,       1 ,     512 ,      10 ,      24 , 283115520 ,
 12, TIDL_ConvolutionLayer         , res5a_branch2b                            1,   1 ,  1 ,  11 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 12 ,       1 ,     512 ,      10 ,      24 ,       1 ,     512 ,      10 ,      24 , 141557760 ,
 13, TIDL_PoolingLayer             , pool6                                     1,   1 ,  1 ,  12 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 13 ,       1 ,     512 ,      10 ,      24 ,       1 ,     512 ,       5 ,      12 ,    122880 ,
 14, TIDL_PoolingLayer             , pool7                                     1,   1 ,  1 ,  13 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 14 ,       1 ,     512 ,       5 ,      12 ,       1 ,     512 ,       3 ,       6 ,     36864 ,
 15, TIDL_PoolingLayer             , pool8                                     1,   1 ,  1 ,  14 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 15 ,       1 ,     512 ,       3 ,       6 ,       1 ,     512 ,       2 ,       3 ,     12288 ,
 16, TIDL_ConvolutionLayer         , ctx_output1                               1,   1 ,  1 ,   9 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 16 ,       1 ,     256 ,      20 ,      48 ,       1 ,     256 ,      20 ,      48 ,  62914560 ,
 17, TIDL_ConvolutionLayer         , ctx_output2                               1,   1 ,  1 ,  12 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 17 ,       1 ,     512 ,      10 ,      24 ,       1 ,     256 ,      10 ,      24 ,  31457280 ,
 18, TIDL_ConvolutionLayer         , ctx_output3                               1,   1 ,  1 ,  13 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 18 ,       1 ,     512 ,       5 ,      12 ,       1 ,     256 ,       5 ,      12 ,   7864320 ,
 19, TIDL_ConvolutionLayer         , ctx_output4                               1,   1 ,  1 ,  14 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 19 ,       1 ,     512 ,       3 ,       6 ,       1 ,     256 ,       3 ,       6 ,   2359296 ,
 20, TIDL_ConvolutionLayer         , ctx_output5                               1,   1 ,  1 ,  15 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 20 ,       1 ,     512 ,       2 ,       3 ,       1 ,     256 ,       2 ,       3 ,    786432 ,
 21, TIDL_ConvolutionLayer         , ctx_output1/relu_mbox_loc                 1,   1 ,  1 ,  16 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 21 ,       1 ,     256 ,      20 ,      48 ,       1 ,      16 ,      20 ,      48 ,   3932160 ,
 22, TIDL_FlattenLayer             , ctx_output1/relu_mbox_loc_perm            1,   1 ,  1 ,  21 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 22 ,       1 ,      16 ,      20 ,      48 ,       1 ,       1 ,       1 ,   15360 ,         1 ,
 23, TIDL_ConvolutionLayer         , ctx_output1/relu_mbox_conf                1,   1 ,  1 ,  16 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 23 ,       1 ,     256 ,      20 ,      48 ,       1 ,      84 ,      20 ,      48 ,  20643840 ,
 24, TIDL_FlattenLayer             , ctx_output1/relu_mbox_conf_perm           1,   1 ,  1 ,  23 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 24 ,       1 ,      84 ,      20 ,      48 ,       1 ,       1 ,       1 ,   80640 ,         1 ,
 26, TIDL_ConvolutionLayer         , ctx_output2/relu_mbox_loc                 1,   1 ,  1 ,  17 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 26 ,       1 ,     256 ,      10 ,      24 ,       1 ,      24 ,      10 ,      24 ,   1474560 ,
 27, TIDL_FlattenLayer             , ctx_output2/relu_mbox_loc_perm            1,   1 ,  1 ,  26 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 27 ,       1 ,      24 ,      10 ,      24 ,       1 ,       1 ,       1 ,    5760 ,         1 ,
 28, TIDL_ConvolutionLayer         , ctx_output2/relu_mbox_conf                1,   1 ,  1 ,  17 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 28 ,       1 ,     256 ,      10 ,      24 ,       1 ,     126 ,      10 ,      24 ,   7741440 ,
 29, TIDL_FlattenLayer             , ctx_output2/relu_mbox_conf_perm           1,   1 ,  1 ,  28 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 29 ,       1 ,     126 ,      10 ,      24 ,       1 ,       1 ,       1 ,   30240 ,         1 ,
 31, TIDL_ConvolutionLayer         , ctx_output3/relu_mbox_loc                 1,   1 ,  1 ,  18 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 31 ,       1 ,     256 ,       5 ,      12 ,       1 ,      24 ,       5 ,      12 ,    368640 ,
 32, TIDL_FlattenLayer             , ctx_output3/relu_mbox_loc_perm            1,   1 ,  1 ,  31 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 32 ,       1 ,      24 ,       5 ,      12 ,       1 ,       1 ,       1 ,    1440 ,         1 ,
 33, TIDL_ConvolutionLayer         , ctx_output3/relu_mbox_conf                1,   1 ,  1 ,  18 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 33 ,       1 ,     256 ,       5 ,      12 ,       1 ,     126 ,       5 ,      12 ,   1935360 ,
 34, TIDL_FlattenLayer             , ctx_output3/relu_mbox_conf_perm           1,   1 ,  1 ,  33 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 34 ,       1 ,     126 ,       5 ,      12 ,       1 ,       1 ,       1 ,    7560        1 �̃t�@�C����R�s�[���܂����B

Processing config file .\tempDir\qunat_stats_config.txt !
  0, TIDL_DataLayer                ,  0,  -1 ,  1 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  0 ,    0 ,    0 ,    0 ,    0 ,    1 ,    3 ,  320 ,  768 ,
  1, TIDL_BatchNormLayer           ,  1,   1 ,  1 ,  0 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  1 ,    1 ,    3 ,  320 ,  768 ,    1 ,    3 ,  320 ,  768 ,
  2, TIDL_ConvolutionLayer         ,  1,   1 ,  1 ,  1 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  2 ,    1 ,    3 ,  320 ,  768 ,    1 ,   32 ,  160 ,  384 ,
  3, TIDL_ConvolutionLayer         ,  1,   1 ,  1 ,  2 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  3 ,    1 ,   32 ,  160 ,  384 ,    1 ,   32 ,   80 ,  192 ,
  4, TIDL_ConvolutionLayer         ,  1,   1 ,  1 ,  3 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  4 ,    1 ,   32 ,   80 ,  192 ,    1 ,   64 ,   80 ,  192 ,
  5, TIDL_ConvolutionLayer         ,  1,   1 ,  1 ,  4 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  5 ,    1 ,   64 ,   80 ,  192 ,    1 ,   64 ,   40 ,   96 ,
  6, TIDL_ConvolutionLayer         ,  1,   1 ,  1 ,  5 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  6 ,    1 ,   64 ,   40 ,   96 ,    1 ,  128 ,   40 ,   96 ,
  7, TIDL_ConvolutionLayer         ,  1,   1 ,  1 ,  6 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  7 ,    1 ,  128 ,   40 ,   96 ,    1 ,  128 ,   20 ,   48 ,
  8, TIDL_ConvolutionLayer         ,  1,   1 ,  1 ,  7 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  8 ,    1 ,  128 ,   20 ,   48 ,    1 ,  256 ,   20 ,   48 ,
  9, TIDL_ConvolutionLayer         ,  1,   1 ,  1 ,  8 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  9 ,    1 ,  256 ,   20 ,   48 ,    1 ,  256 ,   20 ,   48 ,
 10, TIDL_PoolingLayer             ,  1,   1 ,  1 ,  9 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 10 ,    1 ,  256 ,   20 ,   48 ,    1 ,  256 ,   10 ,   24 ,
 11, TIDL_ConvolutionLayer         ,  1,   1 ,  1 , 10 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 11 ,    1 ,  256 ,   10 ,   24 ,    1 ,  512 ,   10 ,   24 ,
 12, TIDL_ConvolutionLayer         ,  1,   1 ,  1 , 11 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 12 ,    1 ,  512 ,   10 ,   24 ,    1 ,  512 ,   10 ,   24 ,
 13, TIDL_PoolingLayer             ,  1,   1 ,  1 , 12 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 13 ,    1 ,  512 ,   10 ,   24 ,    1 ,  512 ,    5 ,   12 ,
 14, TIDL_PoolingLayer             ,  1,   1 ,  1 , 13 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 14 ,    1 ,  512 ,    5 ,   12 ,    1 ,  512 ,    3 ,    6 ,
 15, TIDL_PoolingLayer             ,  1,   1 ,  1 , 14 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 15 ,    1 ,  512 ,    3 ,    6 ,    1 ,  512 ,    2 ,    3 ,
 16, TIDL_ConvolutionLayer         ,  1,   1 ,  1 ,  9 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 16 ,    1 ,  256 ,   20 ,   48 ,    1 ,  256 ,   20 ,   48 ,
 17, TIDL_ConvolutionLayer         ,  1,   1 ,  1 , 12 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 17 ,    1 ,  512 ,   10 ,   24 ,    1 ,  256 ,   10 ,   24 ,
 18, TIDL_ConvolutionLayer         ,  1,   1 ,  1 , 13 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 18 ,    1 ,  512 ,    5 ,   12 ,    1 ,  256 ,    5 ,   12 ,
 19, TIDL_ConvolutionLayer         ,  1,   1 ,  1 , 14 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 19 ,    1 ,  512 ,    3 ,    6 ,    1 ,  256 ,    3 ,    6 ,
 20, TIDL_ConvolutionLayer         ,  1,   1 ,  1 , 15 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 20 ,    1 ,  512 ,    2 ,    3 ,    1 ,  256 ,    2 ,    3 ,
 21, TIDL_ConvolutionLayer         ,  1,   1 ,  1 , 16 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 21 ,    1 ,  256 ,   20 ,   48 ,    1 ,   16 ,   20 ,   48 ,
 22, TIDL_FlattenLayer             ,  1,   1 ,  1 , 21 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 22 ,    1 ,   16 ,   20 ,   48 ,    1 ,    1 ,    1 ,15360 ,
 23, TIDL_ConvolutionLayer         ,  1,   1 ,  1 , 16 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 23 ,    1 ,  256 ,   20 ,   48 ,    1 ,   84 ,   20 ,   48 ,
 24, TIDL_FlattenLayer             ,  1,   1 ,  1 , 23 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 24 ,    1 ,   84 ,   20 ,   48 ,    1 ,    1 ,    1 ,80640 ,
 25, TIDL_ConvolutionLayer         ,  1,   1 ,  1 , 17 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 26 ,    1 ,  256 ,   10 ,   24 ,    1 ,   24 ,   10 ,   24 ,
 26, TIDL_FlattenLayer             ,  1,   1 ,  1 , 26 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 27 ,    1 ,   24 ,   10 ,   24 ,    1 ,    1 ,    1 , 5760 ,
 27, TIDL_ConvolutionLayer         ,  1,   1 ,  1 , 17 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 28 ,    1 ,  256 ,   10 ,   24 ,    1 ,  126 ,   10 ,   24 ,
 28, TIDL_FlattenLayer             ,  1,   1 ,  1 , 28 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 29 ,    1 ,  126 ,   10 ,   24 ,    1 ,    1 ,    1 ,30240 ,
 29, TIDL_ConvolutionLayer         ,  1,   1 ,  1 , 18 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 31 ,    1 ,  256 ,    5 ,   12 ,    1 ,   24 ,    5 ,   12 ,
 30, TIDL_FlattenLayer             ,  1,   1 ,  1 , 31 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 32 ,    1 ,   24 ,    5 ,   12 ,    1 ,    1 ,    1 , 1440 ,
 31, TIDL_ConvolutionLayer         ,  1,   1 ,  1 , 18 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 33 ,    1 ,  256 ,    5 ,   12 ,    1 ,  126 ,    5 ,   12 ,
 32, TIDL_FlattenLayer             ,  1,   1 ,  1 , 33 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 34 ,    1 ,  126 ,    5 ,   12 ,    1 ,    1 ,    1 , 7560 ,
 33, TIDL_ConvolutionLayer         ,  1,   1 ,  1 , 19 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 36 ,    1 ,  256 ,    3 ,    6 ,    1 ,   24 ,    3 ,    6 ,
 34, TIDL_FlattenLayer             ,  1,   1 ,  1 , 36 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 37 ,    1 ,   24 ,    3 ,    6 ,    1 ,    1 ,    1 ,  432 ,
 35, TIDL_ConvolutionLayer         ,  1,   1 ,  1 , 19 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 38 ,    1 ,  256 ,    3 ,    6 ,    1 ,  126 ,    3 ,    6 ,
 36, TIDL_FlattenLayer             ,  1,   1 ,  1 , 38 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 39 ,    1 ,  126 ,    3 ,    6 ,    1 ,    1 ,    1 , 2268 ,
 37, TIDL_ConvolutionLayer         ,  1,   1 ,  1 , 20 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 41 ,    1 ,  256 ,    2 ,    3 ,    1 ,   16 ,    2 ,    3 ,
 38, TIDL_FlattenLayer             ,  1,   1 ,  1 , 41 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 42 ,    1 ,   16 ,    2 ,    3 ,    1 ,    1 ,    1 ,   96 ,
 39, TIDL_ConvolutionLayer         ,  1,   1 ,  1 , 20 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 43 ,    1 ,  256 ,    2 ,    3 ,    1 ,   84 ,    2 ,    3 ,
 40, TIDL_FlattenLayer             ,  1,   1 ,  1 , 43 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 44 ,    1 ,   84 ,    2 ,    3 ,    1 ,    1 ,    1 ,  504 ,
 41, TIDL_ConcatLayer              ,  1,   5 ,  1 , 22 , 27 , 32 , 37 , 42 ,  x ,  x ,  x , 46 ,    1 ,    1 ,    1 ,15360 ,    1 ,    1 ,    1 ,23088 ,
 42, TIDL_ConcatLayer              ,  1,   5 ,  1 , 24 , 29 , 34 , 39 , 44 ,  x ,  x ,  x , 47 ,    1 ,    1 ,    1 ,80640 ,    1 ,    1 ,    1 ,121212 ,
 43, TIDL_DetectionOutputLayer     ,  1,   2 ,  1 , 46 , 47 ,  x ,  x ,  x ,  x ,  x ,  x , 49 ,    1 ,    1 ,    1 ,23088 ,    1 ,    1 ,    1 ,  560 ,
 44, TIDL_DataLayer                ,  0,   1 , -1 , 49 ,  x ,  x ,  x ,  x ,  x ,  x ,  x ,  0 ,    1 ,    1 ,    1 ,  560 ,    0 ,    0 ,    0 ,    0 ,
Layer ID    ,inBlkWidth  ,inBlkHeight ,inBlkPitch  ,outBlkWidth ,outBlkHeight,outBlkPitch ,numInChs    ,numOutChs   ,numProcInChs,numLclInChs ,numLclOutChs,numProcItrs ,numAccItrs  ,numHorBlock ,numVerBlock ,inBlkChPitch,outBlkChPitc,alignOrNot 
      2           72           72           72           32           32           32            3           32            3            1            8            1            3           12            5         5184         1024            1    
      3          194            4          194          192            2          192            8            8            8            4            8            1            2            2           80          776          384            1    
      4           50            7           50           48            5           48           32           64           32           16            8            1            2            4           16          350          240            1    
      5           98            7           98           96            5           96           16           16           16            8            8            1            2            2           16          686          480            1    
      6           50            6           50           48            4           48           64          128           64           32            8            1            2            2           10          300          192            1    
      7           50            6           50           48            4           48           32           32           32           16            8            1            2            2           10          300          192            1    
      8           50            6           50           48            4           48          128          256          128           32            8            1            4            1            5          300          192            1    
      9           50            6           50           48            4           48           64           64           64           32            8            1            2            1            5          300          192            1    
     11           26            4           26           24            2           24          256          512          256           16           32            1           16            1            5          104           48            1    
     12           26            4           26           24            2           24          128          128          128           16           32            1            8            1            5          104           48            1    
     16           48            4           48           48            4           48          256          256          256           32            8            1            8            1            5          192          192            1    
     17           24           10           24           24           10           24          512          256          512           32           32            1           16            1            1          240          240            1    
     18           12            5           12           12            5           12          512          256          512           32           32            1           16            1            1           60           60            1    
     19            6            3            6            6            3            6          512          256          512           32           32            1           16            1            1           18           18            1    
     20            3            2            3            3            2            3          512          256          512           32           32            1           16            1            1            6            6            1    
     21           48            4           48           48            4           48          256           16          256           32            8            1            8            1            5          192          192            1    
     23           48            4           48           48            4           48          256           88          256           32            8            1            8            1            5          192          192            1    
     25           24           10           24           24           10           24          256           24          256           32           24            1            8            1            1          240          240            1    
     27           24           10           24           24           10           24          256          128          256           32           32            1            8            1            1          240          240            1    
     29           12            5           12           12            5           12          256           24          256           32           24            1            8            1            1           60           60            1    
     31           12            5           12           12            5           12          256          128          256           32           32            1            8            1            1           60           60            1    
     33            6            3            6            6            3            6          256           24          256           32           24            1            8            1            1           18           18            1    
     35            6            3            6            6            3            6          256          128          256           32           32            1            8            1            1           18           18            1    
     37            3            2            3            3            2            3          256           16          256           32           16            1            8            1            1            6            6            1    
     39            3            2            3            3            2            3          256           96          256           32           32            1            8            1            1            6            6            1    

Processing Frame Number : 0 

 Layer    1 : Out Q :      254 , TIDL_BatchNormLayer  , PASSED  #MMACs =     0.74,     0.74, Sparsity :   0.00
 Layer    2 : Out Q :     5662 , TIDL_ConvolutionLayer, PASSED  #MMACs =   147.46,   127.80, Sparsity :  13.33
 Layer    3 : Out Q :     6568 , TIDL_ConvolutionLayer, PASSED  #MMACs =   141.56,   141.56, Sparsity :   0.00
 Layer    4 : Out Q :    11145 , TIDL_ConvolutionLayer, PASSED  #MMACs =   283.12,   283.12, Sparsity :   0.00
 Layer    5 : Out Q :    11360 , TIDL_ConvolutionLayer, PASSED  #MMACs =   141.56,   141.56, Sparsity :   0.00
 Layer    6 : Out Q :    13400 , TIDL_ConvolutionLayer, PASSED  #MMACs =   283.12,   283.12, Sparsity :   0.00
 Layer    7 : Out Q :    15675 , TIDL_ConvolutionLayer, PASSED  #MMACs =   141.56,   141.56, Sparsity :   0.00
 Layer    8 : Out Q :    16548 , TIDL_ConvolutionLayer, PASSED  #MMACs =   283.12,   283.12, Sparsity :   0.00
 Layer    9 : Out Q :    14680 , TIDL_ConvolutionLayer, PASSED  #MMACs =   141.56,   141.56, Sparsity :   0.00
 Layer   10 :TIDL_PoolingLayer,     PASSED  #MMACs =     0.06,     0.06, Sparsity :   0.00
 Layer   11 : Out Q :    21005 , TIDL_ConvolutionLayer, PASSED  #MMACs =   283.12,   283.12, Sparsity :   0.00
 Layer   12 : Out Q :     5819 , TIDL_ConvolutionLayer, PASSED  #MMACs =   141.56,   141.56, Sparsity :   0.00
 Layer   13 :TIDL_PoolingLayer,     PASSED  #MMACs =     0.03,     0.03, Sparsity :   0.00
 Layer   14 :TIDL_PoolingLayer,     PASSED  #MMACs =     0.01,     0.01, Sparsity :   0.00
 Layer   15 :TIDL_PoolingLayer,     PASSED  #MMACs =     0.00,     0.00, Sparsity :   0.00
 Layer   16 : Out Q :    21250 , TIDL_ConvolutionLayer, PASSED  #MMACs =    62.91,    62.91, Sparsity :   0.00
 Layer   17 : Out Q :    15487 , TIDL_ConvolutionLayer, PASSED  #MMACs =    31.46,    31.46, Sparsity :   0.00
 Layer   18 : Out Q :    20361 , TIDL_ConvolutionLayer, PASSED  #MMACs =     7.86,     7.86, Sparsity :   0.00
 Layer   19 : Out Q :    22575 , TIDL_ConvolutionLayer, PASSED  #MMACs =     2.36,     2.36, Sparsity :   0.00
 Layer   20 : Out Q :    27565 , TIDL_ConvolutionLayer, PASSED  #MMACs =     0.79,     0.79, Sparsity :   0.00
 Layer   21 : Out Q :     4644 , TIDL_ConvolutionLayer, PASSED  #MMACs =     3.93,     3.93, Sparsity :   0.00
 Layer   22 :TIDL_FlattenLayer, PASSED  #MMACs =     0.02,     0.02, Sparsity :   0.00
 Layer   23 : Out Q :     3760 , TIDL_ConvolutionLayer, PASSED  #MMACs =    21.63,    21.63, Sparsity :   0.00
 Layer   24 :TIDL_FlattenLayer, PASSED  #MMACs =     0.08,     0.08, Sparsity :   0.00
 Layer   25 : Out Q :     8166 , TIDL_ConvolutionLayer, PASSED  #MMACs =     1.47,     1.47, Sparsity :   0.00
 Layer   26 :TIDL_FlattenLayer, PASSED  #MMACs =     0.01,     0.01, Sparsity :   0.00
 Layer   27 : Out Q :     3123 , TIDL_ConvolutionLayer, PASSED  #MMACs =     7.86,     7.86, Sparsity :   0.00
 Layer   28 :TIDL_FlattenLayer, PASSED  #MMACs =     0.03,     0.03, Sparsity :   0.00
 Layer   29 : Out Q :    10031 , TIDL_ConvolutionLayer, PASSED  #MMACs =     0.37,     0.37, Sparsity :   0.00
 Layer   30 :TIDL_FlattenLayer, PASSED  #MMACs =     0.00,     0.00, Sparsity :   0.00
 Layer   31 : Out Q :     2884 , TIDL_ConvolutionLayer, PASSED  #MMACs =     1.97,     1.97, Sparsity :   0.00
 Layer   32 :TIDL_FlattenLayer, PASSED  #MMACs =     0.01,     0.01, Sparsity :   0.00
 Layer   33 : Out Q :     8777 , TIDL_ConvolutionLayer, PASSED  #MMACs =     0.11,     0.11, Sparsity :   0.00
 Layer   34 :TIDL_FlattenLayer, PASSED  #MMACs =     0.00,     0.00, Sparsity :   0.00
 Layer   35 : Out Q :     3649 , TIDL_ConvolutionLayer, PASSED  #MMACs =     0.59,     0.59, Sparsity :   0.00
 Layer   36 :TIDL_FlattenLayer, PASSED  #MMACs =     0.00,     0.00, Sparsity :   0.00
 Layer   37 : Out Q :     9392 , TIDL_ConvolutionLayer, PASSED  #MMACs =     0.02,     0.02, Sparsity :   0.00
 Layer   38 :TIDL_FlattenLayer, PASSED  #MMACs =     0.00,     0.00, Sparsity :   0.00
 Layer   39 : Out Q :     4266 , TIDL_ConvolutionLayer, PASSED  #MMACs =     0.15,     0.15, Sparsity :   0.00
 Layer   40 :TIDL_FlattenLayer, PASSED  #MMACs =     0.00,     0.00, Sparsity :   0.00
 Layer   41 : Out Q :     4662 , TIDL_ConcatLayer, PASSED  #MMACs =     0.00,     0.00, Sparsity :  -1.#J
 Layer   42 : Out Q :     2895 , TIDL_ConcatLayer, PASSED  #MMACs =     0.00,     0.00, Sparsity :  -1.#J
 Layer   43 : #MMACs =     0.00,     0.00, Sparsity :   0.00
End of config list found !
 ,         1 ,
 36, TIDL_ConvolutionLayer         , ctx_output4/relu_mbox_loc                 1,   1 ,  1 ,  19 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 36 ,       1 ,     256 ,       3 ,       6 ,       1 ,      24 ,       3 ,       6 ,    110592 ,
 37, TIDL_FlattenLayer             , ctx_output4/relu_mbox_loc_perm            1,   1 ,  1 ,  36 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 37 ,       1 ,      24 ,       3 ,       6 ,       1 ,       1 ,       1 ,     432 ,         1 ,
 38, TIDL_ConvolutionLayer         , ctx_output4/relu_mbox_conf                1,   1 ,  1 ,  19 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 38 ,       1 ,     256 ,       3 ,       6 ,       1 ,     126 ,       3 ,       6 ,    580608 ,
 39, TIDL_FlattenLayer             , ctx_output4/relu_mbox_conf_perm           1,   1 ,  1 ,  38 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 39 ,       1 ,     126 ,       3 ,       6 ,       1 ,       1 ,       1 ,    2268 ,         1 ,
 41, TIDL_ConvolutionLayer         , ctx_output5/relu_mbox_loc                 1,   1 ,  1 ,  20 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 41 ,       1 ,     256 ,       2 ,       3 ,       1 ,      16 ,       2 ,       3 ,     24576 ,
 42, TIDL_FlattenLayer             , ctx_output5/relu_mbox_loc_perm            1,   1 ,  1 ,  41 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 42 ,       1 ,      16 ,       2 ,       3 ,       1 ,       1 ,       1 ,      96 ,         1 ,
 43, TIDL_ConvolutionLayer         , ctx_output5/relu_mbox_conf                1,   1 ,  1 ,  20 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 43 ,       1 ,     256 ,       2 ,       3 ,       1 ,      84 ,       2 ,       3 ,    129024 ,
 44, TIDL_FlattenLayer             , ctx_output5/relu_mbox_conf_perm           1,   1 ,  1 ,  43 ,  x ,  x ,  x ,  x ,  x ,  x ,  x , 44 ,       1 ,      84 ,       2 ,       3 ,       1 ,       1 ,       1 ,     504 ,         1 ,
 46, TIDL_ConcatLayer              , mbox_loc                                  1,   5 ,  1 ,  22 , 27 , 32 , 37 , 42 ,  x ,  x ,  x , 46 ,       1 ,       1 ,       1 ,   15360 ,       1 ,       1 ,       1 ,   23088 ,         1 ,
 47, TIDL_ConcatLayer              , mbox_conf                                 1,   5 ,  1 ,  24 , 29 , 34 , 39 , 44 ,  x ,  x ,  x , 47 ,       1 ,       1 ,       1 ,   80640 ,       1 ,       1 ,       1 ,  121212 ,         1 ,
 49, TIDL_DetectionOutputLayer     , detection_out                             2,   2 ,  1 ,  46 , 47 ,  x ,  x ,  x ,  x ,  x ,  x , 49 ,       1 ,       1 ,       1 ,   23088 ,       1 ,       1 ,       1 ,     560 ,         1 ,
Total Giga Macs : 2.1312

Unexpectedly, in both (A) and (B) “Total Giga Macs” is 2.1312.

Any insights?

Regards,

-Shibata

0 Praveen Eppa1 over 5 years ago in reply to Taro Shibata

TI__Genius 17580 points

Hi Shibata,

I thought customer had two models, one with Sparsity (A) and other is Dense (B) and was doing these experiments. But from the logs I see that customer using same model for both Sparse and Dense profiling by just changing the "conv2dKernelType" parameter.

This "conv2dKernelType" parameter will not change the sparisty in the model, it will just select different convolution kernel based on "conv2dKernelType" parameter. TIDL support two flavors of convolution layer namely Sparse and Dense flow. Functionally both the flows will generate same results. The sparse flow takes advantage for zero kernels co-efficient and improves the execution speed by not computing them. So default sparse convolution can be used, however the sparse flow has relatively high overheads when processing small ROIs (Input/Feature maps smaller than 32x32). For these cases we recommend to use dense convolution flow. So while importing models each layer can be set to either sparse or dense using “conv2dKernelType” parameter in import config file.

Hence the customer need to train different models by inducing sparsity to really see the effect of the sparsity on the performance.

Hope this clarifies.

Thanks,

Praveen

0 Taro Shibata over 5 years ago in reply to Praveen Eppa1

TI__Expert 8110 points

Hi Praveen,

Thanks for the explanation.

Yes, as you pointed out customer used the same model for both Sparse and Dense profiling.

They understood your explanation.

I have one question about “Total Giga Macs”.

This value shows total sum of Macs in original (before conversion) model, not calculating the Macs after conversion.

Is my understanding correct?

Thanks,

-Shibata

0 Praveen Eppa1 over 5 years ago in reply to Taro Shibata

TI__Genius 17580 points

Hi Shibata,

Yes, your understanding is correct.

Thanks,

Praveen

0 Taro Shibata over 5 years ago in reply to Praveen Eppa1

TI__Expert 8110 points

Hi Praveen,

Thanks for the clarification.

Regards,

-Shibata

Processors

Processors forum

TDA2EXEVM: [TIDL] Dense model is faster than Sparse model