TDA4x EVM: Inception based model takes 20 minutes per frame

Ying Cai36

Hi E2E'ers

I'm trying to run a 30 layer model based on inception net for performing image segmentation. During the conversion process from caffe a performance analysis showed that the model requires about 1.6 GMACs, which does not seem like an infeasible load for the TDA4x to handle. Nonetheless, when trying to run the model on our EVM, I am finding that the TIDL openVX node is taking more than 20 minutes to push out a frame.

Our OpenVX app is based loosely on the provided AVP TIDL demo.

I'm stuck here - any ideas why this could be happening?

-Ying

over 6 years ago

0 Rishabh Garg over 6 years ago

TI__Guru 55685 points

Hi Ying,

Can you first try to run the model standalone with TIDL and see the performance.

Please refer to TIDL user guide: psdk_rtos_auto_j7_06_01_00_15/tidl_j7_01_00_00_00/ti_dl/docs/user_guide_html/md_tidl_fsg_steps_to_debug_mismatch.html

Regards,

Rishabh

0 Ying Cai36 over 6 years ago in reply to Rishabh Garg

Prodigy 225 points

Thanks for the quick response Rishabh!

So I've tried running "./PC_dsp_test_dl_algo.out configList.txt"

Contents of "configList.txt":

test.txt

Contents of "test.txt"

netBinFile = "model.bin"

ioConfigFile = "ioConfig.bin"
inData = "rawImg.bgr"
outData = "outputTensor.bin"
inFileFormat = 1
padInBuffInTB = 0
numFrames = 1
postProcType = 0
debugTraceLevel = 2
enableLayerPerfTraces = 1

rawImg.bgr is a raw 512*288*3 binary test image datafile, in BGR planar format. This file was generated from a real image from a driving log and is strictly pixel values with no padding (e.g. the file's size is exactly 512*288*3)
outputTensor.bin is expected to have a pixel segmented output with dimensions 512*288.

Running the demo produces the following output:

Processing config file #0 : test.txt 
Alg Alloc for Layer # -    0
Alg Alloc for Layer # -    1
Alg Alloc for Layer # -    2
Alg Alloc for Layer # -    3
Alg Alloc for Layer # -    4
Alg Alloc for Layer # -    5
Alg Alloc for Layer # -    6
Alg Alloc for Layer # -    7
Alg Alloc for Layer # -    8
Alg Alloc for Layer # -    9
Alg Alloc for Layer # -   10
Alg Alloc for Layer # -   11
Alg Alloc for Layer # -   12
Alg Alloc for Layer # -   13
Alg Alloc for Layer # -   14
Alg Alloc for Layer # -   15
Alg Alloc for Layer # -   16
Alg Alloc for Layer # -   17
Alg Alloc for Layer # -   18
Alg Alloc for Layer # -   19
Alg Alloc for Layer # -   20
Alg Alloc for Layer # -   21
Alg Alloc for Layer # -   22
Alg Alloc for Layer # -   23
Alg Alloc for Layer # -   24
Alg Alloc for Layer # -   25
Alg Alloc for Layer # -   26
Alg Alloc for Layer # -   27
Alg Alloc for Layer # -   28
Alg Alloc for Layer # -   29
Num,    Space,     SizeinBytes,   SineInMB
   0,    17,         2816,      0.003 0x2c5b430
   1,    17,          120,      0.000 0x2c5bf40
   2,    17,        16384,      0.016 0x7fef24f0d080
   3,    17,       458752,      0.438 0x7fef24f11080
   4,    17,      8257536,      7.875 0x7fef24f81080
   5,    17,      2766484,      2.638 0x7fef249be010
   6,    17,       717568,      0.684 0x7fef25761080
   7,    17,      5787136,      5.519 0x7fef25810380
   8,    17,     28915712,     27.576 0x7fef25d95180
   9,    17,      9636864,      9.190 0x7fef27928980
  10,    17,       692480,      0.660 0x7fef672b6010
  11,    17,       381248,      0.364 0x7fef24960010
  12,    17,          256,      0.000 0x7fef28259580
Total External Memory (DDR) Size =     57633356,     54.963 
Alg Init for Layer # -    0
Alg Init for Layer # -    1
Alg Init for Layer # -    2
Alg Init for Layer # -    3
Alg Init for Layer # -    4
Alg Init for Layer # -    5
Alg Init for Layer # -    6
Alg Init for Layer # -    7
Alg Init for Layer # -    8
Alg Init for Layer # -    9
Alg Init for Layer # -   10
Alg Init for Layer # -   11
Alg Init for Layer # -   12
Alg Init for Layer # -   13
Alg Init for Layer # -   14
Alg Init for Layer # -   15
Alg Init for Layer # -   16
Alg Init for Layer # -   17
Alg Init for Layer # -   18
Alg Init for Layer # -   19
Alg Init for Layer # -   20
Alg Init for Layer # -   21
Alg Init for Layer # -   22
Alg Init for Layer # -   23
Alg Init for Layer # -   24
Alg Init for Layer # -   25
Alg Init for Layer # -   26
Alg Init for Layer # -   27
Alg Init for Layer # -   28
Alg Init for Layer # -   29

 Instance created for  test.txt

Processing Cnt :    0, InstCnt :    0 model.bin!
 ----------------------- TIDL Process with REF_ONLY FLOW ------------------------

#    0 . ..Starting Layer # -    0
Starting Layer # -    1
Processing Layer # -    1
End of Layer # -    1
Starting Layer # -    2
Processing Layer # -    2
End of Layer # -    2
Starting Layer # -    3
Processing Layer # -    3
End of Layer # -    3
Starting Layer # -    4
Processing Layer # -    4
End of Layer # -    4
Starting Layer # -    5
Processing Layer # -    5
End of Layer # -    5
Starting Layer # -    6
Processing Layer # -    6
End of Layer # -    6
Starting Layer # -    7
Processing Layer # -    7
End of Layer # -    7
Starting Layer # -    8
Processing Layer # -    8
End of Layer # -    8
Starting Layer # -    9
Processing Layer # -    9
End of Layer # -    9
Starting Layer # -   10
Processing Layer # -   10
End of Layer # -   10
Starting Layer # -   11
Processing Layer # -   11
End of Layer # -   11
Starting Layer # -   12
Processing Layer # -   12
End of Layer # -   12
Starting Layer # -   13
Processing Layer # -   13
End of Layer # -   13
Starting Layer # -   14
Processing Layer # -   14
End of Layer # -   14
Starting Layer # -   15
Processing Layer # -   15
End of Layer # -   15
Starting Layer # -   16
Processing Layer # -   16
End of Layer # -   16
Starting Layer # -   17
Processing Layer # -   17
End of Layer # -   17
Starting Layer # -   18
Processing Layer # -   18
End of Layer # -   18
Starting Layer # -   19
Processing Layer # -   19
End of Layer # -   19
Starting Layer # -   20
Processing Layer # -   20
End of Layer # -   20
Starting Layer # -   21
Processing Layer # -   21
End of Layer # -   21
Starting Layer # -   22
Processing Layer # -   22
End of Layer # -   22
Starting Layer # -   23
Processing Layer # -   23
End of Layer # -   23
Starting Layer # -   24
Processing Layer # -   24
End of Layer # -   24
Starting Layer # -   25
Processing Layer # -   25
End of Layer # -   25
Starting Layer # -   26
Processing Layer # -   26
End of Layer # -   26
Starting Layer # -   27
Processing Layer # -   27
End of Layer # -   27
Starting Layer # -   28
Processing Layer # -   28
End of Layer # -   28
Starting Layer # -   29

Network Cycles 0 
           Layer,   Layer Cycles,kernelOnlyCycles, coreLoopCycles,LayerSetupCycles,dmaPipeupCycles, dmaPipeDownCycles, PrefetchCycles,copyKerCoeffCycles,LayerDeinitCycles,LastBlockCycles, paddingTrigger,    paddingWait,LayerWithoutPad,LayerHandleCopy,
     1,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
     2,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
     3,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
     4,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
     5,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
     6,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
     7,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
     8,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
     9,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    10,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    11,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    12,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    13,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    14,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    15,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    16,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    17,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    18,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    19,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    20,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    21,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    22,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    23,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    24,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    25,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    26,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    27,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,
    28,              0,              0,              0,              0,              0,                 0,              0,                 0,              0,              0,              0,              0,              0,              0,

Additionally, the outputTensor.bin file is generated as a 147456B file (i.e. 512*288 bytes) containing all zeros. Toggling the "padInBuffInTB" parameter does not change the results.

...I'm assuming I haven't specified some input correctly.

0 Ying Cai36 over 6 years ago in reply to Ying Cai36

Prodigy 225 points

Also, here's an analysis that I think was output by the perfSim tool, showing the layers of our toy model and some configuration details. The original file is also attached.

Fullscreen 6862.tidl_net_tinetv3_seg_argmax.bin_netLog.txt Download

Num of Layer Detected :  30 
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
  Num|TIDL Layer Name               |Out Data Name                                     |Group |#Ins  |#Outs |Inbuf Ids                       |Outbuf Id |In NCHW                             |Out NCHW                            |MACS       |
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
    0|TIDL_DataLayer                |img_data                                          |     0|    -1|     1|  x   x   x   x   x   x   x   x |  0       |       0        0        0        0 |       1        3      288      512 |         0 |
    1|TIDL_ConvolutionLayer         |conv1a                                            |     0|     1|     1|  0   x   x   x   x   x   x   x |  1       |       1        3      288      512 |       1       32      144      256 |  90832896 |
    2|TIDL_ConvolutionLayer         |conv1b                                            |     0|     1|     1|  1   x   x   x   x   x   x   x |  2       |       1       32      144      256 |       1       32      144      256 |  87293952 |
    3|TIDL_PoolingLayer             |pool1                                             |     0|     1|     1|  2   x   x   x   x   x   x   x |  3       |       1       32      144      256 |       1       32       72      128 |   1179648 |
    4|TIDL_ConvolutionLayer         |res2a_branch2a                                    |     0|     1|     1|  3   x   x   x   x   x   x   x |  4       |       1       32       72      128 |       1       64       72      128 | 171048960 |
    5|TIDL_ConvolutionLayer         |res2a_branch2b                                    |     0|     1|     1|  4   x   x   x   x   x   x   x |  5       |       1       64       72      128 |       1       64       72      128 |  86114304 |
    6|TIDL_PoolingLayer             |pool2                                             |     0|     1|     1|  5   x   x   x   x   x   x   x |  6       |       1       64       72      128 |       1       64       36       64 |    589824 |
    7|TIDL_ConvolutionLayer         |res3a_branch2a                                    |     0|     1|     1|  6   x   x   x   x   x   x   x |  7       |       1       64       36       64 |       1      128       36       64 | 170459136 |
    8|TIDL_ConvolutionLayer         |res3a_branch2b                                    |     0|     1|     1|  7   x   x   x   x   x   x   x |  8       |       1      128       36       64 |       1      128       36       64 |  85524480 |
    9|TIDL_PoolingLayer             |pool3                                             |     0|     1|     1|  8   x   x   x   x   x   x   x |  9       |       1      128       36       64 |       1      128       18       32 |    294912 |
   10|TIDL_ConvolutionLayer         |out3a                                             |     0|     1|     1|  8   x   x   x   x   x   x   x | 10       |       1      128       36       64 |       1       64       36       64 |  85229568 |
   11|TIDL_ConvolutionLayer         |res4a_branch2a                                    |     0|     1|     1|  9   x   x   x   x   x   x   x | 11       |       1      128       18       32 |       1      256       18       32 | 170164224 |
   12|TIDL_ConvolutionLayer         |res4a_branch2b                                    |     0|     1|     1| 11   x   x   x   x   x   x   x | 12       |       1      256       18       32 |       1      256       18       32 |  85229568 |
   13|TIDL_PoolingLayer             |pool4                                             |     0|     1|     1| 12   x   x   x   x   x   x   x | 13       |       1      256       18       32 |       1      256        9       16 |    147456 |
   14|TIDL_ConvolutionLayer         |res5a_branch2a                                    |     0|     1|     1| 13   x   x   x   x   x   x   x | 14       |       1      256        9       16 |       1      512        9       16 | 170016768 |
   15|TIDL_ConvolutionLayer         |res5a_branch2b                                    |     0|     1|     1| 14   x   x   x   x   x   x   x | 15       |       1      512        9       16 |       1      512        9       16 |  85082112 |
   16|TIDL_ConvolutionLayer         |out5a                                             |     0|     1|     1| 15   x   x   x   x   x   x   x | 16       |       1      512        9       16 |       1       64        9       16 |  21252096 |
   17|TIDL_Deconv2DLayer            |out5a_up2                                         |     0|     1|     1| 16   x   x   x   x   x   x   x | 17       |       1       64        9       16 |       1       64       18       32 |    663552 |
   18|TIDL_Deconv2DLayer            |out5a_up4                                         |     0|     1|     1| 17   x   x   x   x   x   x   x | 18       |       1       64       18       32 |       1       64       36       64 |   2654208 |
   19|TIDL_EltWiseLayer             |out3_out5_combined                                |     0|     2|     1| 18  10   x   x   x   x   x   x | 19       |       1       64       36       64 |       1       64       36       64 |    147456 |
   20|TIDL_ConvolutionLayer         |ctx_conv1                                         |     0|     1|     1| 19   x   x   x   x   x   x   x | 20       |       1       64       36       64 |       1       64       36       64 |  85229568 |
   21|TIDL_ConvolutionLayer         |ctx_conv2                                         |     0|     1|     1| 20   x   x   x   x   x   x   x | 21       |       1       64       36       64 |       1       64       36       64 |  85229568 |
   22|TIDL_ConvolutionLayer         |ctx_conv3                                         |     0|     1|     1| 21   x   x   x   x   x   x   x | 22       |       1       64       36       64 |       1       64       36       64 |  85229568 |
   23|TIDL_ConvolutionLayer         |ctx_conv4                                         |     0|     1|     1| 22   x   x   x   x   x   x   x | 23       |       1       64       36       64 |       1       64       36       64 |  85229568 |
   24|TIDL_ConvolutionLayer         |ctx_final                                         |     0|     1|     1| 23   x   x   x   x   x   x   x | 24       |       1       64       36       64 |       1        5       36       64 |   6658560 |
   25|TIDL_Deconv2DLayer            |out_deconv_final_up2                              |     0|     1|     1| 24   x   x   x   x   x   x   x | 25       |       1        5       36       64 |       1        5       72      128 |    829440 |
   26|TIDL_Deconv2DLayer            |out_deconv_final_up4                              |     0|     1|     1| 25   x   x   x   x   x   x   x | 26       |       1        5       72      128 |       1        5      144      256 |   3317760 |
   27|TIDL_Deconv2DLayer            |out_deconv_final_up8                              |     0|     1|     1| 26   x   x   x   x   x   x   x | 27       |       1        5      144      256 |       1        5      288      512 |  11796480 |
   28|TIDL_ArgMaxLayer              |seg_out                                           |     0|     1|     1| 27   x   x   x   x   x   x   x | 28       |       1        5      288      512 |       1        1      288      512 |    147456 |
   29|TIDL_DataLayer                |seg_out                                           |     0|     1|    -1| 28   x   x   x   x   x   x   x |  0       |       1        1      288      512 |       0        0        0        0 |         0 |
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Total Giga Macs : 1.6776
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

Num of Layer Detected :  30 
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
  Num|TIDL Layer Name               |Out Data Name                                     |Group |#Ins  |#Outs |Inbuf Ids                       |Outbuf Id |In NCHW                             |Out NCHW                            |MACS       |
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
    0|TIDL_DataLayer                |img_data                                          |     0|    -1|     1|  x   x   x   x   x   x   x   x |  0       |       0        0        0        0 |       1        3      288      512 |         0 |
    1|TIDL_ConvolutionLayer         |conv1a                                            |     0|     1|     1|  0   x   x   x   x   x   x   x |  1       |       1        3      288      512 |       1       32      144      256 |  90832896 |
    2|TIDL_ConvolutionLayer         |conv1b                                            |     0|     1|     1|  1   x   x   x   x   x   x   x |  2       |       1       32      144      256 |       1       32      144      256 |  87293952 |
    3|TIDL_PoolingLayer             |pool1                                             |     0|     1|     1|  2   x   x   x   x   x   x   x |  3       |       1       32      144      256 |       1       32       72      128 |   1179648 |
    4|TIDL_ConvolutionLayer         |res2a_branch2a                                    |     0|     1|     1|  3   x   x   x   x   x   x   x |  4       |       1       32       72      128 |       1       64       72      128 | 171048960 |
    5|TIDL_ConvolutionLayer         |res2a_branch2b                                    |     0|     1|     1|  4   x   x   x   x   x   x   x |  5       |       1       64       72      128 |       1       64       72      128 |  86114304 |
    6|TIDL_PoolingLayer             |pool2                                             |     0|     1|     1|  5   x   x   x   x   x   x   x |  6       |       1       64       72      128 |       1       64       36       64 |    589824 |
    7|TIDL_ConvolutionLayer         |res3a_branch2a                                    |     0|     1|     1|  6   x   x   x   x   x   x   x |  7       |       1       64       36       64 |       1      128       36       64 | 170459136 |
    8|TIDL_ConvolutionLayer         |res3a_branch2b                                    |     0|     1|     1|  7   x   x   x   x   x   x   x |  8       |       1      128       36       64 |       1      128       36       64 |  85524480 |
    9|TIDL_PoolingLayer             |pool3                                             |     0|     1|     1|  8   x   x   x   x   x   x   x |  9       |       1      128       36       64 |       1      128       18       32 |    294912 |
   10|TIDL_ConvolutionLayer         |out3a                                             |     0|     1|     1|  8   x   x   x   x   x   x   x | 10       |       1      128       36       64 |       1       64       36       64 |  85229568 |
   11|TIDL_ConvolutionLayer         |res4a_branch2a                                    |     0|     1|     1|  9   x   x   x   x   x   x   x | 11       |       1      128       18       32 |       1      256       18       32 | 170164224 |
   12|TIDL_ConvolutionLayer         |res4a_branch2b                                    |     0|     1|     1| 11   x   x   x   x   x   x   x | 12       |       1      256       18       32 |       1      256       18       32 |  85229568 |
   13|TIDL_PoolingLayer             |pool4                                             |     0|     1|     1| 12   x   x   x   x   x   x   x | 13       |       1      256       18       32 |       1      256        9       16 |    147456 |
   14|TIDL_ConvolutionLayer         |res5a_branch2a                                    |     0|     1|     1| 13   x   x   x   x   x   x   x | 14       |       1      256        9       16 |       1      512        9       16 | 170016768 |
   15|TIDL_ConvolutionLayer         |res5a_branch2b                                    |     0|     1|     1| 14   x   x   x   x   x   x   x | 15       |       1      512        9       16 |       1      512        9       16 |  85082112 |
   16|TIDL_ConvolutionLayer         |out5a                                             |     0|     1|     1| 15   x   x   x   x   x   x   x | 16       |       1      512        9       16 |       1       64        9       16 |  21252096 |
   17|TIDL_Deconv2DLayer            |out5a_up2                                         |     0|     1|     1| 16   x   x   x   x   x   x   x | 17       |       1       64        9       16 |       1       64       18       32 |    663552 |
   18|TIDL_Deconv2DLayer            |out5a_up4                                         |     0|     1|     1| 17   x   x   x   x   x   x   x | 18       |       1       64       18       32 |       1       64       36       64 |   2654208 |
   19|TIDL_EltWiseLayer             |out3_out5_combined                                |     0|     2|     1| 18  10   x   x   x   x   x   x | 19       |       1       64       36       64 |       1       64       36       64 |    147456 |
   20|TIDL_ConvolutionLayer         |ctx_conv1                                         |     0|     1|     1| 19   x   x   x   x   x   x   x | 20       |       1       64       36       64 |       1       64       36       64 |  85229568 |
   21|TIDL_ConvolutionLayer         |ctx_conv2                                         |     0|     1|     1| 20   x   x   x   x   x   x   x | 21       |       1       64       36       64 |       1       64       36       64 |  85229568 |
   22|TIDL_ConvolutionLayer         |ctx_conv3                                         |     0|     1|     1| 21   x   x   x   x   x   x   x | 22       |       1       64       36       64 |       1       64       36       64 |  85229568 |
   23|TIDL_ConvolutionLayer         |ctx_conv4                                         |     0|     1|     1| 22   x   x   x   x   x   x   x | 23       |       1       64       36       64 |       1       64       36       64 |  85229568 |
   24|TIDL_ConvolutionLayer         |ctx_final                                         |     0|     1|     1| 23   x   x   x   x   x   x   x | 24       |       1       64       36       64 |       1        5       36       64 |   6658560 |
   25|TIDL_Deconv2DLayer            |out_deconv_final_up2                              |     0|     1|     1| 24   x   x   x   x   x   x   x | 25       |       1        5       36       64 |       1        5       72      128 |    829440 |
   26|TIDL_Deconv2DLayer            |out_deconv_final_up4                              |     0|     1|     1| 25   x   x   x   x   x   x   x | 26       |       1        5       72      128 |       1        5      144      256 |   3317760 |
   27|TIDL_Deconv2DLayer            |out_deconv_final_up8                              |     0|     1|     1| 26   x   x   x   x   x   x   x | 27       |       1        5      144      256 |       1        5      288      512 |  11796480 |
   28|TIDL_ArgMaxLayer              |seg_out                                           |     0|     1|     1| 27   x   x   x   x   x   x   x | 28       |       1        5      288      512 |       1        1      288      512 |    147456 |
   29|TIDL_DataLayer                |seg_out                                           |     0|     1|    -1| 28   x   x   x   x   x   x   x |  0       |       1        1      288      512 |       0        0        0        0 |         0 |
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Total Giga Macs : 1.6776
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

0 Rishabh Garg over 6 years ago in reply to Ying Cai36

TI__Guru 55685 points

Hi,

This network looks similar to jsegNet21v2. Please see https://e2e.ti.com/support/processors/f/791/t/848577

Also can you run the standalone network on target to see the performance numbers.

Regards,

Rishabh

0 Rishabh Garg over 6 years ago in reply to Rishabh Garg

TI__Guru 55685 points

Hi,

Have you been able to make progress on this issue?

Regards,

Rishabh

0 Ying Cai36 over 6 years ago in reply to Rishabh Garg

Prodigy 225 points

Hi Rishabh,

I'm getting the following issue when trying to initialize the DSP app via JTAG:

C71X_0: Trouble Reading Memory Block at 0x0 on Page 0 of Length 0x1: (Error -1344 - (0:0:0)) Register or Memory request has failed due to a Transaction Error. This may be the result of reading from a write only register, writing to a read only register, or accessing a memory location that is not assigned. No user action required. (Emulation package 8.3.1.00007)

C71X_0: Trouble Reading Memory Block at 0x707e8e42 on Page 0 of Length 0x1: (Error -1350 - (0:0:0)) Register or Memory request has timed out. The CPU is executing a High Priority Interrupt, this is typically due to a Double Page Fault. Choose 'Force' to force the CPU into a state that allows debug of why the HPI is active. (Emulation package 8.3.1.00007)

I'm using a Spectrum Digital XDS560v2 STM.

0 Ying Cai36 over 6 years ago in reply to Ying Cai36

Prodigy 225 points

Regarding the jsegNet model you linked to, you mentioned in that thread that there were some changes made to deploy.protoxt. Would you be able to describe those changes? Do they address an issue that could be causing the symptoms I'm seeing here?

0 Ying Cai36 over 6 years ago in reply to Ying Cai36

Prodigy 225 points

I also got clearance from my team to share with you our model, prototxt, and tidl_import.txt files.

I've sent them to John Smrstk and David Copeland and asked them to forward to you. Hopefully this will help us more directly approach a root cause. Please let me know if you are able to get meaningful test bench results and/or coherent tensor outputs from these files.

Meanwhile I've attached the prototxt file:

Fullscreen deploy_seg.prototxt.txt Download

name: "tinet_v3"

input: "img_data"
input_shape {
  dim: 1
  dim: 3
  dim: 288
  dim: 512
}
layer {
  name: "conv1a"
  type: "Convolution"
  bottom: "img_data"
  top: "conv1a"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    bias_term: true
    pad: 2
    kernel_size: 5
    group: 1
    stride: 2
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 1
  }
}
layer {
  name: "conv1a/bn"
  type: "BatchNorm"
  bottom: "conv1a"
  top: "conv1a"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "conv1a/relu"
  type: "ReLU"
  bottom: "conv1a"
  top: "conv1a"
}
layer {
  name: "conv1b"
  type: "Convolution"
  bottom: "conv1a"
  top: "conv1b"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 32
    bias_term: true
    pad: 1
    kernel_size: 3
    group: 4
    stride: 1
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 1
  }
}
layer {
  name: "conv1b/bn"
  type: "BatchNorm"
  bottom: "conv1b"
  top: "conv1b"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "conv1b/relu"
  type: "ReLU"
  bottom: "conv1b"
  top: "conv1b"
}
layer {
  name: "pool1"
  type: "Pooling"
  bottom: "conv1b"
  top: "pool1"
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 2
  }
}
layer {
  name: "res2a_branch2a"
  type: "Convolution"
  bottom: "pool1"
  top: "res2a_branch2a"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 64
    bias_term: true
    pad: 1
    kernel_size: 3
    group: 1
    stride: 1
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 1
  }
}
layer {
  name: "res2a_branch2a/bn"
  type: "BatchNorm"
  bottom: "res2a_branch2a"
  top: "res2a_branch2a"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "res2a_branch2a/relu"
  type: "ReLU"
  bottom: "res2a_branch2a"
  top: "res2a_branch2a"
}
layer {
  name: "res2a_branch2b"
  type: "Convolution"
  bottom: "res2a_branch2a"
  top: "res2a_branch2b"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 64
    bias_term: true
    pad: 1
    kernel_size: 3
    group: 4
    stride: 1
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 1
  }
}
layer {
  name: "res2a_branch2b/bn"
  type: "BatchNorm"
  bottom: "res2a_branch2b"
  top: "res2a_branch2b"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "res2a_branch2b/relu"
  type: "ReLU"
  bottom: "res2a_branch2b"
  top: "res2a_branch2b"
}
layer {
  name: "pool2"
  type: "Pooling"
  bottom: "res2a_branch2b"
  top: "pool2"
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 2
  }
}
layer {
  name: "res3a_branch2a"
  type: "Convolution"
  bottom: "pool2"
  top: "res3a_branch2a"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 128
    bias_term: true
    pad: 1
    kernel_size: 3
    group: 1
    stride: 1
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 1
  }
}
layer {
  name: "res3a_branch2a/bn"
  type: "BatchNorm"
  bottom: "res3a_branch2a"
  top: "res3a_branch2a"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "res3a_branch2a/relu"
  type: "ReLU"
  bottom: "res3a_branch2a"
  top: "res3a_branch2a"
}
layer {
  name: "res3a_branch2b"
  type: "Convolution"
  bottom: "res3a_branch2a"
  top: "res3a_branch2b"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 128
    bias_term: true
    pad: 1
    kernel_size: 3
    group: 4
    stride: 1
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 1
  }
}
layer {
  name: "res3a_branch2b/bn"
  type: "BatchNorm"
  bottom: "res3a_branch2b"
  top: "res3a_branch2b"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "res3a_branch2b/relu"
  type: "ReLU"
  bottom: "res3a_branch2b"
  top: "res3a_branch2b"
}
layer {
  name: "pool3"
  type: "Pooling"
  bottom: "res3a_branch2b"
  top: "pool3"
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 2
  }
}
layer {
  name: "res4a_branch2a"
  type: "Convolution"
  bottom: "pool3"
  top: "res4a_branch2a"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 256
    bias_term: true
    pad: 1
    kernel_size: 3
    group: 1
    stride: 1
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 1
  }
}
layer {
  name: "res4a_branch2a/bn"
  type: "BatchNorm"
  bottom: "res4a_branch2a"
  top: "res4a_branch2a"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "res4a_branch2a/relu"
  type: "ReLU"
  bottom: "res4a_branch2a"
  top: "res4a_branch2a"
}
layer {
  name: "res4a_branch2b"
  type: "Convolution"
  bottom: "res4a_branch2a"
  top: "res4a_branch2b"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 256
    bias_term: true
    pad: 1
    kernel_size: 3
    group: 4
    stride: 1
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 1
  }
}
layer {
  name: "res4a_branch2b/bn"
  type: "BatchNorm"
  bottom: "res4a_branch2b"
  top: "res4a_branch2b"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "res4a_branch2b/relu"
  type: "ReLU"
  bottom: "res4a_branch2b"
  top: "res4a_branch2b"
}
layer {
  name: "pool4"
  type: "Pooling"
  bottom: "res4a_branch2b"
  top: "pool4"
  pooling_param {
    pool: MAX
    kernel_size: 2
    stride: 2
  }
}
layer {
  name: "res5a_branch2a"
  type: "Convolution"
  bottom: "pool4"
  top: "res5a_branch2a"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 512
    bias_term: true
    pad: 2
    kernel_size: 3
    group: 1
    stride: 1
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 2
  }
}
layer {
  name: "res5a_branch2a/bn"
  type: "BatchNorm"
  bottom: "res5a_branch2a"
  top: "res5a_branch2a"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "res5a_branch2a/relu"
  type: "ReLU"
  bottom: "res5a_branch2a"
  top: "res5a_branch2a"
}
layer {
  name: "res5a_branch2b"
  type: "Convolution"
  bottom: "res5a_branch2a"
  top: "res5a_branch2b"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 512
    bias_term: true
    pad: 2
    kernel_size: 3
    group: 4
    stride: 1
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 2
  }
}
layer {
  name: "res5a_branch2b/bn"
  type: "BatchNorm"
  bottom: "res5a_branch2b"
  top: "res5a_branch2b"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "res5a_branch2b/relu"
  type: "ReLU"
  bottom: "res5a_branch2b"
  top: "res5a_branch2b"
}
################
# segmentation #
################
layer {
  name: "out5a"
  type: "Convolution"
  bottom: "res5a_branch2b"
  top: "out5a"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 64
    bias_term: true
    pad: 4
    kernel_size: 3
    group: 2
    stride: 1
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 4
  }
}
layer {
  name: "out5a/bn"
  type: "BatchNorm"
  bottom: "out5a"
  top: "out5a"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "out5a/relu"
  type: "ReLU"
  bottom: "out5a"
  top: "out5a"
}
layer {
  name: "out5a_up2"
  type: "Deconvolution"
  bottom: "out5a"
  top: "out5a_up2"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  convolution_param {
    num_output: 64
    bias_term: false
    pad: 1
    kernel_size: 4
    group: 64
    stride: 2
    weight_filler {
      type: "bilinear"
    }
  }
}
layer {
  name: "out5a_up2/bn"
  type: "BatchNorm"
  bottom: "out5a_up2"
  top: "out5a_up2"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "out5a_up2/relu"
  type: "ReLU"
  bottom: "out5a_up2"
  top: "out5a_up2"
}
layer {
  name: "out5a_up4"
  type: "Deconvolution"
  bottom: "out5a_up2"
  top: "out5a_up4"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  convolution_param {
    num_output: 64
    bias_term: false
    pad: 1
    kernel_size: 4
    group: 64
    stride: 2
    weight_filler {
      type: "bilinear"
    }
  }
}
layer {
  name: "out5a_up4/bn"
  type: "BatchNorm"
  bottom: "out5a_up4"
  top: "out5a_up4"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "out5a_up4/relu"
  type: "ReLU"
  bottom: "out5a_up4"
  top: "out5a_up4"
}
layer {
  name: "out3a"
  type: "Convolution"
  bottom: "res3a_branch2b"
  top: "out3a"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 64
    bias_term: true
    pad: 1
    kernel_size: 3
    group: 2
    stride: 1
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 1
  }
}
layer {
  name: "out3a/bn"
  type: "BatchNorm"
  bottom: "out3a"
  top: "out3a"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "out3a/relu"
  type: "ReLU"
  bottom: "out3a"
  top: "out3a"
}
layer {
  name: "out3_out5_combined"
  type: "Eltwise"
  bottom: "out5a_up4"
  bottom: "out3a"
  top: "out3_out5_combined"
}
layer {
  name: "ctx_conv1"
  type: "Convolution"
  bottom: "out3_out5_combined"
  top: "ctx_conv1"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 64
    bias_term: true
    pad: 1
    kernel_size: 3
    group: 1
    stride: 1
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 1
  }
}
layer {
  name: "ctx_conv1/bn"
  type: "BatchNorm"
  bottom: "ctx_conv1"
  top: "ctx_conv1"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "ctx_conv1/relu"
  type: "ReLU"
  bottom: "ctx_conv1"
  top: "ctx_conv1"
}
layer {
  name: "ctx_conv2"
  type: "Convolution"
  bottom: "ctx_conv1"
  top: "ctx_conv2"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 64
    bias_term: true
    pad: 4
    kernel_size: 3
    group: 1
    stride: 1
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 4
  }
}
layer {
  name: "ctx_conv2/bn"
  type: "BatchNorm"
  bottom: "ctx_conv2"
  top: "ctx_conv2"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "ctx_conv2/relu"
  type: "ReLU"
  bottom: "ctx_conv2"
  top: "ctx_conv2"
}
layer {
  name: "ctx_conv3"
  type: "Convolution"
  bottom: "ctx_conv2"
  top: "ctx_conv3"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 64
    bias_term: true
    pad: 4
    kernel_size: 3
    group: 1
    stride: 1
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 4
  }
}
layer {
  name: "ctx_conv3/bn"
  type: "BatchNorm"
  bottom: "ctx_conv3"
  top: "ctx_conv3"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "ctx_conv3/relu"
  type: "ReLU"
  bottom: "ctx_conv3"
  top: "ctx_conv3"
}
layer {
  name: "ctx_conv4"
  type: "Convolution"
  bottom: "ctx_conv3"
  top: "ctx_conv4"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 64
    bias_term: true
    pad: 4
    kernel_size: 3
    group: 1
    stride: 1
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 4
  }
}
layer {
  name: "ctx_conv4/bn"
  type: "BatchNorm"
  bottom: "ctx_conv4"
  top: "ctx_conv4"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "ctx_conv4/relu"
  type: "ReLU"
  bottom: "ctx_conv4"
  top: "ctx_conv4"
}
layer {
  name: "ctx_final"
  type: "Convolution"
  bottom: "ctx_conv4"
  top: "ctx_final"
  param {
    lr_mult: 1
    decay_mult: 1
  }
  param {
    lr_mult: 2
    decay_mult: 0
  }
  convolution_param {
    num_output: 5
    bias_term: true
    pad: 1
    kernel_size: 3
    group: 1
    stride: 1
    weight_filler {
      type: "msra"
    }
    bias_filler {
      type: "constant"
      value: 0
    }
    dilation: 1
  }
}
layer {
  name: "ctx_final/bn"
  type: "BatchNorm"
  bottom: "ctx_final"
  top: "ctx_final"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "ctx_final/relu"
  type: "ReLU"
  bottom: "ctx_final"
  top: "ctx_final"
}
layer {
  name: "out_deconv_final_up2"
  type: "Deconvolution"
  bottom: "ctx_final"
  top: "out_deconv_final_up2"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  convolution_param {
    num_output: 5
    bias_term: false
    pad: 1
    kernel_size: 4
    group: 5
    stride: 2
    weight_filler {
      type: "bilinear"
    }
  }
}
layer {
  name: "out_deconv_final_up2/bn"
  type: "BatchNorm"
  bottom: "out_deconv_final_up2"
  top: "out_deconv_final_up2"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "out_deconv_final_up2/relu"
  type: "ReLU"
  bottom: "out_deconv_final_up2"
  top: "out_deconv_final_up2"
}
layer {
  name: "out_deconv_final_up4"
  type: "Deconvolution"
  bottom: "out_deconv_final_up2"
  top: "out_deconv_final_up4"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  convolution_param {
    num_output: 5
    bias_term: false
    pad: 1
    kernel_size: 4
    group: 5
    stride: 2
    weight_filler {
      type: "bilinear"
    }
  }
}
layer {
  name: "out_deconv_final_up4/bn"
  type: "BatchNorm"
  bottom: "out_deconv_final_up4"
  top: "out_deconv_final_up4"
  batch_norm_param {
    moving_average_fraction: 0.99
    eps: 0.0001
    scale_bias: true
  }
}
layer {
  name: "out_deconv_final_up4/relu"
  type: "ReLU"
  bottom: "out_deconv_final_up4"
  top: "out_deconv_final_up4"
}
layer {
  name: "out_deconv_final_up8"
  type: "Deconvolution"
  bottom: "out_deconv_final_up4"
  top: "out_deconv_final_up8"
  param {
    lr_mult: 0
    decay_mult: 0
  }
  convolution_param {
    num_output: 5
    bias_term: false
    pad: 1
    kernel_size: 4
    group: 5
    stride: 2
    weight_filler {
      type: "bilinear"
    }
  }
}
layer {
  name: "seg_out"
  type: "ArgMax"
  bottom: "out_deconv_final_up8"
  top: "seg_out"
  argmax_param {
    axis: 1
  }
}
##############
## detection #
##############
#layer {
#  name: "det_conv1"
#  type: "Convolution"
#  bottom: "ctx_conv4"
#  top: "det_conv1"
#  param {
#    lr_mult: 1
#    decay_mult: 1
#  }
#  param {
#    lr_mult: 2
#    decay_mult: 0
#  }
#  convolution_param {
#    num_output: 128
#    bias_term: true
#    pad: 1
#    kernel_size: 3
#    group: 2
#    stride: 2
#    weight_filler {
#      type: "msra"
#    }
#    bias_filler {
#      type: "constant"
#      value: 0
#    }
#    dilation: 1
#  }
#}
#layer {
#  name: "det_conv1/bn"
#  type: "BatchNorm"
#  bottom: "det_conv1"
#  top: "det_conv1"
#  batch_norm_param {
#    moving_average_fraction: 0.99
#    eps: 0.0001
#    scale_bias: true
#  }
#}
#layer {
#  name: "det_conv1/relu"
#  type: "ReLU"
#  bottom: "det_conv1"
#  top: "det_conv1"
#}
#layer {
#  name: "det_conv2"
#  type: "Convolution"
#  bottom: "det_conv1"
#  top: "det_conv2"
#  param {
#    lr_mult: 1
#    decay_mult: 1
#  }
#  param {
#    lr_mult: 2
#    decay_mult: 0
#  }
#  convolution_param {
#    num_output: 512
#    bias_term: true
#    pad: 2
#    kernel_size: 5
#    group: 8
#    stride: 2
#    weight_filler {
#      type: "msra"
#    }
#    bias_filler {
#      type: "constant"
#      value: 0
#    }
#    dilation: 1
#  }
#}
#layer {
#  name: "det_conv2/bn"
#  type: "BatchNorm"
#  bottom: "det_conv2"
#  top: "det_conv2"
#  batch_norm_param {
#    moving_average_fraction: 0.99
#    eps: 0.0001
#    scale_bias: true
#  }
#}
#layer {
#  name: "det_conv2/relu"
#  type: "ReLU"
#  bottom: "det_conv2"
#  top: "det_conv2"
#}
#layer {
#  name: "det_concat"
#  type: "Concat"
#  bottom: "res5a_branch2b"
#  bottom: "det_conv2"
#  top: "det_concat"
#}
#layer {
#  name: "det_conv_final"
#  type: "Convolution"
#  bottom: "det_concat"
#  top: "det_conv_final"
#  param {
#    lr_mult: 1
#    decay_mult: 1
#  }
#  param {
#    lr_mult: 2
#    decay_mult: 0
#  }
#  convolution_param {
#    num_output: 40
## num_anchors * (xywho + num_classes)
#    bias_term: true
#    pad: 0
#    kernel_size: 1
#    group: 1
#    stride: 1
#    weight_filler {
#      type: "msra"
#    }
#    bias_filler {
#      type: "constant"
#      value: 0
#    }
#    dilation: 1
#  }
#}
#layer {
#  name: "det_out"
#  type: "Yolo2"
#  bottom: "det_conv_final"
#  top: "det_out"
#  yolo_param {
#    anchor { w: 1.0 h: 1.0 }
#    anchor { w: 2.0 h: 2.0 }
#    anchor { w: 4.0 h: 4.0 }
#    anchor { w: 8.0 h: 8.0 }
#    anchor { w: 12.0 h: 12.0 }
#    label_offset: 0
#    num_classes: 3
#    width_scale: 512
#    height_scale: 288
#  }
#}

0 Ying Cai36 over 6 years ago in reply to Ying Cai36

Prodigy 225 points

Also, some images showing what we're seeing. This is the first frame, which took 20 minutes to render.

This is an image of the provided avp_tidl demo, swapping out the provided onnx model with our own. The config file was modified to point to our own converted model and the variable pc_size modified to 512 288.

0 Rishabh Garg over 6 years ago in reply to Ying Cai36

TI__Guru 55685 points

Hi,

Instructions for running TIDL standalone using CCS are given here: tidl_j7_01_00_00_00/ti_dl/docs/user_guide_html/md_tidl_sample_test.html.

I had attached the modified deploy.prototxt in thread https://e2e.ti.com/support/j721e/f/1026/t/848577 (please see reply on 6th Nov)

Regards,

Rishabh

0 Ying Cai36 over 6 years ago in reply to Rishabh Garg

Prodigy 225 points

Yes... those are the instructions I was following when I got the error I shared yesterday:

C71X_0: Trouble Reading Memory Block at 0x0 on Page 0 of Length 0x1: (Error -1344 - (0:0:0)) Register or Memory request has failed due to a Transaction Error. This may be the result of reading from a write only register, writing to a read only register, or accessing a memory location that is not assigned. No user action required. (Emulation package 8.3.1.00007)

Steps to run Sample Application on EVM

Copy the "testvecs" folder in to the memory card. Create a empty "trace" folder in memory card if you are enabling writeTraceLevel.
Follow the steps mentioned in SDK Vision Apps documentation to setup the CCS + GELS + CCXML files to bringup/connect EVM.
Insert the memory car in to EVM
Start the CCS and Launch target configuration for the EVM
Run the script for running the sciclient available in PDK "packages\ti\drv\sciclient\tools\ccsLoadDmsc\j721e\launch.js"
Run the TIDL Launch script"ti_dl\utils\targetTest\launch_tidl.js" to run the Sample application. This will stop the C7x core on a break-point. Press the run to continue the execution.
At the end of execution take out the memory card from EVM and compare the results with reference. One can also set writeOutput = 0 to compare the results against a reference instead of file write.

0 Rishabh Garg over 6 years ago in reply to Ying Cai36

TI__Guru 55685 points

Hi Ying,

Is your board set to NO BOOT mode?

Regards,

Rishabh

0 Ying Cai36 over 6 years ago in reply to Rishabh Garg

Prodigy 225 points

I believe so, here's a picture of the DIP switches:

0 Rishabh Garg over 6 years ago in reply to Ying Cai36

TI__Guru 55685 points

Hi Ying,

Dip switches are set correctly.

I had installed CSP and emulator package to match the one you are using after looking at yesterday's response and it worked fine for me.

Can you try after restarting CCS and emulator?

Regards,

Rishabh

0 Ying Cai36 over 6 years ago in reply to Ying Cai36

Prodigy 225 points

In any case, it sounds like the model we are trying to run is not quite within the realm that the SDK was tested for. I will discuss with our model experts about training a model from the list here: tidl_j7_01_00_00_00/ti_dl/docs/user_guide_html/md_tidl_performance.html

0 Ying Cai36 over 6 years ago in reply to Ying Cai36

Prodigy 225 points

OK, so I've tried importing the model you attached here:

https://github.com/tidsp/caffe-jacinto-models/tree/caffe-0.17/trained/image_segmentation/cityscapes5_jsegnet21v2/sparse

Running this model on the tidl_avp demo gives.... better results, but is still quite slow. The attached image took about 7 minutes to render. I will try the JTAG profiler again tomorrow to see if I can get you some performance numbers.

I've also made sure to use the deploy.prototxt that you shared, and not the one in the github folder.

Here are the contents of my tidl_import.txt:

modelType          = 0
inputNetFile       = "cityscapes5_jsegnet21v2/deploy.prototxt"
inputParamsFile    = "cityscapes5_jsegnet21v2/cityscapes5_jsegnet21v2_iter_60000.caffemodel"
outputNetFile      = "cityscapes5_jsegnet21v2/output/tidlNetwork.bin"
outputParamsFile   = "cityscapes5_jsegnet21v2/output/tidlIO_"
numParamBits = 8
numFeatureBits = 8
inElementType = 0
inWidth = 768
inHeight = 384
inNumChannels = 3
inDataFormat = 0

#Quantization
inFileFormat = 2
inData = cityscapes5_jsegnet21v2/calib_list.txt
tidlStatsTool = ../../test/tidl_quant_stats_tool.out

# Performance Simulation
perfSimConfig = ../../test/testvecs/config/import/perfsim_base.cfg
perfSimTool = ../../utils/perfsim/ti_cnnperfsim.out

0 Rishabh Garg over 6 years ago in reply to Ying Cai36

TI__Guru 55685 points

Hi Ying,

I am attaching the import and infer files that I had used

Fullscreen 4863.tidl_import_jSegNet21v2.txt Download

modelType          = 0
inputNetFile       = "../../test/testvecs/models/public/caffe/jsegNet21v2/deploy.prototxt"
inputParamsFile    = "../../test/testvecs/models/public/caffe/jsegNet21v2/deploy.caffemodel"
outputNetFile      = "../../test/testvecs/config/tidl_models/caffe/tidl_net_jSegNet21v2_1024x512.bin"
outputParamsFile   = "../../test/testvecs/config/tidl_models/caffe/tidl_io_jSegNet21v2_1024x512_"
inDataFormat = 0
perfSimConfig = ../../test/testvecs/config/import/perfsim_base.cfg
inData  =   "../../test/testvecs/config/segmentation_list.txt"
postProcType = 3
foldMaxPoolInConv2D = 0

Fullscreen 2502.tidl_infer_jSegNet21v2.txt Download

inFileFormat    = 2
postProcType = 3
padInBuffInTB = 1
netBinFile      = "testvecs/config/tidl_models/caffe/tidl_net_jSegNet21v2_1024x512.bin"
ioConfigFile    = "testvecs/config/tidl_models/caffe/tidl_io_jSegNet21v2_1024x512_1.bin"
outData =   "testvecs/output/jsegNet21v2_1024x512.bin"
inData  =   "testvecs/config/segmentation_list.txt"
totNumClasses = 5
debugTraceLevel = 0
writeTraceLevel = 0

Regards,

Rishabh

0 Rishabh Garg over 6 years ago in reply to Rishabh Garg

TI__Guru 55685 points

The output that I got is also given below:

Regards,

Rishabh

0 Ying Cai36 over 6 years ago in reply to Rishabh Garg

Prodigy 225 points

Hi Rishabh,

Excellent progress!! I've imported with your settings and was able to confirm real-time performance.

Experimenting with the differences between our import.txt files, it looks like the "inWidth" and "inHeight" parameters are causing the severe slow-down.

Not sure what you guys will want to do with this information, but we can probably work with this for now by using resize nodes to preprocess images into "native" inference resolution.

Thanks so much for your patience Rishabh - much appreciated.

-Ying

0 Rishabh Garg over 6 years ago in reply to Ying Cai36

TI__Guru 55685 points

Hi Ying,

Glad that your issue is resolved.

We have multiple publicly available networks running at different resolutions with real time performance.

You can look at the TIDL data sheet (part of user guide) to see those numbers.

We will be further optimizing TIDL performance in subsequent releases.

Regards,

Rishabh

0 Rishabh Garg over 6 years ago in reply to Rishabh Garg

TI__Guru 55685 points

Hi Ying,

Can you please share the final performance numbers that you were able to achieve with this network.

Regards,

Rishabh

0 Ying Cai36 over 6 years ago in reply to Rishabh Garg

Prodigy 225 points

I have a lot of debugging left to do on my custom graph, but the TIDL node seems to be running at about 45 ms per frame (~22 fps)

0 Rishabh Garg over 6 years ago in reply to Ying Cai36

TI__Guru 55685 points

Hi Ying,

Ok , thanks for sharing this information.

Regards,

Rishabh

Processors

Processors forum

TDA4x EVM: Inception based model takes 20 minutes per frame

Steps to run Sample Application on EVM