soroosh129 / neuos Goto Github PK

Hi,
I use the caffenet-cudnn-HASHTABLE.txt as a parameter to run the classify.bin.

However, I don't know how to get caffe-build/dvfs/med/*.log files to write Latency.txt, TotalEnergy.txt, and so on, when I read the generate_hashtable.py. Could you give me a way how I do it?

Hi, thank you for the excellent work!

I encounter a segmentation fault problem, when I run the classify.bin. If I set the iteration from 15 to 1, It runs correct. How to solve this problem?

++++ shmem__us/sched_energy: 52 0
Concorrency= 0
Latency(us)/Energy(mJ): 0 0
Using lowrank instead.
Layer names:conv1_v and conv1_h
Latency(us)/Energy(mJ): 82483 332.063
Latency(us)/Energy(mJ): 26 0
Latency(us)/Energy(mJ): 179 0.719461
Latency(us)/Energy(mJ): 94 0.237628
Using lowrank instead.
Layer names:conv2_v and conv2_h
Latency(us)/Energy(mJ): 5808 20.0019
Latency(us)/Energy(mJ): 22 0
Latency(us)/Energy(mJ): 143 0.527584
Latency(us)/Energy(mJ): 94 0.260632
Using lowrank instead.
Layer names:conv3_v and conv3_h
Latency(us)/Energy(mJ): 1285 5.00857
Latency(us)/Energy(mJ): 21 0
Using lowrank instead.
Layer names:conv4_v and conv4_h
Latency(us)/Energy(mJ): 1807 7.31882
Latency(us)/Energy(mJ): 20 0
Using lowrank instead.
Layer names:conv5_v and conv5_h
Latency(us)/Energy(mJ): 4245 17.9505
Latency(us)/Energy(mJ): 21 0.313044
Latency(us)/Energy(mJ): 84 0.326688
Latency(us)/Energy(mJ): 3703 17.2541
Latency(us)/Energy(mJ): 23 0
Latency(us)/Energy(mJ): 0 0
Latency(us)/Energy(mJ): 4901 27.0591
Latency(us)/Energy(mJ): 39 0
Latency(us)/Energy(mJ): 0 0
Latency(us)/Energy(mJ): 2076 13.6563
Latency(us)/Energy(mJ): 83 0.576356
_peroid: 25
Progress Cohort={ 20.5,0,0,0,0,0,0,0,0,0,Accuracy Cohort={ 20.5,0,0,0,0,0,0,0,0,0,}
}
++++ sched_us/sched_energy: 52 11.5271
++++ xup_exec_us/xup_energy_mj: 107157 443.274
++++ shmem__us/sched_energy: 73 0
Concorrency= 1101266944
Segmentation fault (core dumped)

This is a great project.
Can you explain what the value in the HASHTABLE represents?
Thank you.

Background

Board: Jetson Xavier NX (with 8G RAM and 128G SSD)
OS: Ubuntu 18.04.6 LTS

• NVIDIA Jetson Xavier NX (Developer Kit Version)
  • Jetpack 4.6.1 [L4T 32.7.1]
  • NV Power Mode: MODE_20W_6CORE - Type: 8
  • jetson_stats.service: active
• Libraries:
  • CUDA: 10.2.300
  • cuDNN: 8.2.1.32
  • TensorRT: 8.2.1.8
  • Visionworks: 1.6.0.501
  • OpenCV: 3.4.3 compiled CUDA: YES
  • VPI: ii libnvvpi1 1.2.3 arm64 NVIDIA Vision Programming Interface library
  • Vulkan: 1.2.70

Problems

When I run run_cudnn.sh command to run vggnet, the program will stop after two warnings.
Here is my output:

[libprotobuf WARNING google/protobuf/io/coded_stream.cc:604] Reading dangerously large protocol message.  If the message turns out to be larger than 2147483647 bytes, parsing will be halted for security reasons.  To increase the limit (or to disable these warnings), see CodedInputStream::SetTotalBytesLimit() in google/protobuf/io/coded_stream.h.
[libprotobuf WARNING google/protobuf/io/coded_stream.cc:81] The total number of bytes read was 574671192
I0512 00:11:41.895918 23203 upgrade_proto.cpp:53] Attempting to upgrade input file specified using deprecated V1LayerParameter: /home/lqz/download/NeuOS/models/vggnet/VGG_ILSVRC_19_layers.caffemodel
./run-cudnn.sh: line 48: 23203 Killed                  LD_PRELOAD="/home/lqz/download/NeuOS/build/lib/libcaffe.so.1.0.0" ./classify.bin $proto $weight /home/lqz/download/NeuOS/data/ilsvrc12/imagenet_mean.binaryproto /home/lqz/download/NeuOS/data/ilsvrc12/synset_words.txt /home/lqz/download/NeuOS/examples/images/cat.jpg 15 25 5 all-AGX.config $uncert $lowrank_proto $lowrank_weight

For two warnings above, I tried to google them but I didn't found any answer that can solve my problems.

Reading dangerously large protocol message

How to change the buffer limit in Google's protobuf?

Attempting to upgrade input file specified using deprecated V1LayerParameter

what-does-attempting-to-upgrade-input-file-specified-using-deprecated-transform

Do you have any idea for my problems? I'd be very grateful if you can help me.

PS: English is not my native language; please excuse typing errors.