Test failure

[mrlucas]

Hi, I'm getting a test failure.
Here's the list of commands I used to install. This is on a fresh Debian Stretch vm (9.11):
apt install curl
curl -sL https://deb.nodesource.com/setup_10.x -o nodesource_setup.sh
bash nodesource_setup.sh
apt update
apt install nodejs
apt-get install -y git ssh tar gzip ca-certificates build-essential sqlite3 ffmpeg
apt-get install python-dev portaudio19-dev
curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py
python get-pip.py
pip install python_speech_features h5py numpy scipy keras tensorflow zerorpc sounddevice psutil
git clone https://github.com/adblockradio/adblockradio.git
cd adblockradio/
npm install
npm test

Below is the full output of "npm test", the most telling of the errors seems to be:
File "/root/adblockradio/predictor-ml/mlpredict.py", line 57, in get_session
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=gpu_fraction)
AttributeError: 'module' object has no attribute 'GPUOptions'

It sounds like it may be a dependency version issue? There is no GPU passed through to the vm, which may also be a factor.

adblockradio@1.0.0 test /root/adblockradio
mocha --delay test/file.js && mocha --delay test/file.js --mljs && mocha --delay test/online.js && mocha --delay test/online.js --mljs && mocha --delay test/offline.js && mocha --delay test/offline.js --mljs

[2019-10-06T12:53:22.667Z] warn post-processing: updating ML model is not possible when analysing files. skip.
[2019-10-06T12:53:22.671Z] warn post-processing: updating hotlist DB is not possible when analysing files. skip.
[2019-10-06T12:53:22.672Z] warn post-processing: updating hotlist DB is not possible when analysing files. skip.
[2019-10-06T12:53:22.673Z] warn post-processing: not possible (yet?) to stop the processing of files. please kill the process instead.
[2019-10-06T12:53:22.674Z] info test-file: refresh attempted. result=false
[2019-10-06T12:53:23.177Z] info checkModelUpdates: /root/adblockradio/model/France_RTL/model.keras.tar.gz is up to date
[2019-10-06T12:53:23.863Z] info checkModelUpdates: /root/adblockradio/model/France_RTL/hotlist.sqlite.tar.gz is up to date
[2019-10-06T12:53:23.864Z] info predictor: run predictor on file /root/adblockradio/test/file.mp3 with config={"predInterval":1,"saveDuration":10,"enablePredictorMl":true,"enablePredictorHotlist":true,"saveMetadata":true,"verbose":false,"modelPath":"/root/adblockradio/model","hotlistFile":"France_RTL/hotlist.sqlite","modelUpdates":true,"modelUpdateInterval":60,"JSPredictorMl":false,"modelFile":"France_RTL/model.keras"}
[2019-10-06T12:53:23.874Z] debug predictor: readAmount=44100 bytes
[2019-10-06T12:53:23.878Z] info pred-hotlist: open hotlist db /root/adblockradio/model/France_RTL/hotlist.sqlite (memory=true)
[2019-10-06T12:53:23.881Z] info pred-ml: France_RTL Python predictor. __dirname=/root/adblockradio/predictor-ml env: PKG=false Electron=false
[2019-10-06T12:53:24.063Z] info pred-hotlist: /root/adblockradio/model/France_RTL/hotlist.sqlite db found
[2019-10-06T12:53:24.311Z] info pred-hotlist: France_RTL: Hotlist ready
[2019-10-06T12:53:25.119Z] info predictor: decoding finished
[2019-10-06T12:53:27.056Z] error pred-ml: France_RTL mlpredict child stderr data: Traceback (most recent call last):

[2019-10-06T12:53:27.057Z] error pred-ml: France_RTL mlpredict child stderr data: File "/root/adblockradio/predictor-ml/mlpredict.py", line 64, in

[2019-10-06T12:53:27.059Z] error pred-ml: France_RTL mlpredict child stderr data: tensorflow_backend.set_session(get_session())
File "/root/adblockradio/predictor-ml/mlpredict.py", line 57, in get_session
gpu_options = tf.GPUOptions(per_process_gpu_memory_fraction=gpu_fraction)
AttributeError: 'module' object has no attribute 'GPUOptions'

[2019-10-06T12:54:22.027Z] error test-file: analysis timed out or was too slow. kill it.

File analysis (Python child process)
1) should have emitted data
2) should have reached the end of the file
✓ should reject attempts to reload ML model, hotlist DB or metadata scraper during analysis.
✓ should not have thrown errors
3) should have exited properly
4) should write results in JSON format

2 passing (46ms)
4 failing

1. File analysis (Python child process)
   should have emitted data:

   AssertionError [ERR_ASSERTION]: The expression evaluated to a falsy value:

assert(gotData)

  + expected - actual

  -false
  +true

  at Context.<anonymous> (test/file.js:93:4)

2. File analysis (Python child process)
   should have reached the end of the file:

   AssertionError [ERR_ASSERTION]: The expression evaluated to a falsy value:

assert(finished)

  + expected - actual

  -false
  +true

  at Context.<anonymous> (test/file.js:98:4)

3. File analysis (Python child process)
   should have exited properly:
   AssertionError [ERR_ASSERTION]: null == 0
   at Context. (test/file.js:112:11)

4. File analysis (Python child process)
   should write results in JSON format:

   AssertionError [ERR_ASSERTION]: The expression evaluated to a falsy value:

assert(fileOutputIsSane)

  + expected - actual

  -false
  +true

  at Context.<anonymous> (test/file.js:117:4)

npm ERR! Test failed. See above for more details.

[dest4]

Thanks for the nice report. It's difficult to update Tensorflow across the whole backend, but I'll have to do it and better manage version pinning.