convert_models error

[idata]

when I use the original convert_models.sh script, then run python3 stream_infer.py is OK, but the inference time is a bit slow, googlenet about 566ms, squeeznet 301ms

when I add the "-s 12" after "python3 $NCS_TOOLKIT_ROOT/mvNCCompile.pyc ../networks/SqueezeNet/NetworkConfig.prototxt -w ../networks/SqueezeNet/squeezenet_v1.0.caffemodel -o ../networks/SqueezeNet/graph" in the script, there is an error as below:

~/Work/movidius/ncapi/py_examples/stream_infer$ python3 stream_infer.py

Number of categories: 2Device 0 Address: 13 - VID/PID 03e7:2150Starting wait for connect with 2000ms timeoutFound Address: 13 - VID/PID 03e7:2150Found EP 0x81 : max packet size is 512 bytesFound EP 0x01 : max packet size is 512 bytesFound and opened devicePerforming bulk write of 825136 bytes…Successfully sent 825136 bytes of data in 48.770596 ms (16.134948 MB/s)Boot successful, device address 13Found Address: 13 - VID/PID 040e:f63bdoneBooted 13 -> VSC1Exception in thread Thread-1:Traceback (most recent call last): File "/usr/lib/python3.5/threading.py", line 914, in _bootstrap_inner self.run() File "/usr/lib/python3.5/threading.py", line 862, in run self._target(*self._args, **self._kwargs) File "stream_infer.py", line 149, in input_thread gGraph.LoadTensor(preprocessed_image_buf ,"frame %s" % frame_number) File "/usr/local/lib/python3.5/dist-packages/mvnc/mvncapi.py", line 194, in LoadTensor raise Exception(Status(status))Exception: Status.ERROR

what can I do?

btw, Ubuntu 16.04, new movidius SDK MvNC_SDK_1.07.07

[idata]

@arlose Hi arlose, it seems like the problem you are experiencing might not be related to using the -s 12 option for compiling the graph file. I have been unable to reproduce the problem so far. I suggest unplugging any connected NCS devices and rebooting your machine and then try running the latest version of stream_infer and let me know if you are still experiencing the same problem. The latest version of stream_infer can be found at: https://ncs-forum-uploads.s3.amazonaws.com/neal/stream_infer.py. Another suggestion is to try running stream_infer again with the graph file that was compiled without using the -s 12 option.

[idata]

hi @Tome_at_Intel , I use the latest version of stream_infer, but have the same problem. when I use the graph file without using the -s 12 option, is all OK.

when I change the option -s 12 to -s 10, I can run the stream_infer, but after a few seconds, it crashed again.

[idata]

@arlose Can you try using a powered USB hub? Also, this may or may not be an issue with using -s 10, but the number of active SHAVEs is configurable in options of 1, 2, 4, 8, or 12 for use with network layers. Try using one of those SHAVE options instead of -s 10.