Test after train.. problem...

[daniele777]

TRAIN :

(.venv) dani@NB-Daniele:/mnt/c/Users/daniele.angelini/mycroft-precise-dev/mycroft-precise-dev$ precise-train -e 50 tars-
start.net tars-start/
Using TensorFlow backend.
WARNING:tensorflow:From /mnt/c/users/daniele.angelini/mycroft-precise-dev/mycroft-precise-dev/.venv/lib/python3.7/site-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.
Instructions for updating:
Colocations handled automatically by placer.
WARNING:tensorflow:From /mnt/c/users/daniele.angelini/mycroft-precise-dev/mycroft-precise-dev/.venv/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:3138: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.
Instructions for updating:
Please use rate instead of keep_prob. Rate should be set to rate = 1 - keep_prob.
Data:
Loading wake-word...
Loading not-wake-word...
Loading wake-word...
Loading not-wake-word...
Inputs shape: (10, 29, 13)
Outputs shape: (10, 1)
Test inputs shape: (5, 29, 13)
Test outputs shape: (5, 1)

---

Layer (type) Output Shape Param #

net (GRU) (None, 20) 2040

---

dense_1 (Dense) (None, 1) 21

Total params: 2,061
Trainable params: 2,061
Non-trainable params: 0

---

Train on 10 samples, validate on 5 samples
2021-05-20 11:51:48.868410: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
2021-05-20 11:51:48.933481: I tensorflow/core/platform/profile_utils/cpu_utils.cc:94] CPU Frequency: 2904000000 Hz
2021-05-20 11:51:48.938174: I tensorflow/compiler/xla/service/service.cc:150] XLA service 0x349df40 executing computations on platform Host. Devices:
2021-05-20 11:51:48.939065: I tensorflow/compiler/xla/service/service.cc:158] StreamExecutor device (0): ,
Epoch 1/50
10/10 [==============================] - 0s 40ms/step - loss: 0.1834 - acc: 0.6000 - val_loss: 0.1735 - val_acc: 0.8000
Epoch 2/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0476 - acc: 0.9000 - val_loss: 0.1760 - val_acc: 0.8000
Epoch 3/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0932 - acc: 0.9000 - val_loss: 0.1251 - val_acc: 0.8000
Epoch 4/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0419 - acc: 1.0000 - val_loss: 0.1146 - val_acc: 1.0000
Epoch 5/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0231 - acc: 1.0000 - val_loss: 0.1090 - val_acc: 1.0000
Epoch 6/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0531 - acc: 1.0000 - val_loss: 0.0883 - val_acc: 1.0000
Epoch 7/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0837 - acc: 0.9000 - val_loss: 0.0742 - val_acc: 1.0000
Epoch 8/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0532 - acc: 0.9000 - val_loss: 0.0794 - val_acc: 1.0000
Epoch 9/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0093 - acc: 1.0000 - val_loss: 0.0780 - val_acc: 1.0000
Epoch 10/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0587 - acc: 1.0000 - val_loss: 0.0596 - val_acc: 1.0000
Epoch 11/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0259 - acc: 1.0000 - val_loss: 0.0517 - val_acc: 1.0000
Epoch 12/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0685 - acc: 1.0000 - val_loss: 0.0441 - val_acc: 1.0000
Epoch 13/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0641 - acc: 0.9000 - val_loss: 0.0345 - val_acc: 1.0000
Epoch 14/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0512 - acc: 0.9000 - val_loss: 0.0317 - val_acc: 1.0000
Epoch 15/50
10/10 [==============================] - 0s 3ms/step - loss: 0.1730 - acc: 0.9000 - val_loss: 0.0256 - val_acc: 1.0000
Epoch 16/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0216 - acc: 1.0000 - val_loss: 0.0237 - val_acc: 1.0000
Epoch 17/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0129 - acc: 1.0000 - val_loss: 0.0229 - val_acc: 1.0000
Epoch 18/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0090 - acc: 1.0000 - val_loss: 0.0222 - val_acc: 1.0000
Epoch 19/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0567 - acc: 0.8000 - val_loss: 0.0219 - val_acc: 1.0000
Epoch 20/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0516 - acc: 1.0000 - val_loss: 0.0192 - val_acc: 1.0000
Epoch 21/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0307 - acc: 0.9000 - val_loss: 0.0187 - val_acc: 1.0000
Epoch 22/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0372 - acc: 0.9000 - val_loss: 0.0186 - val_acc: 1.0000
Epoch 23/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0240 - acc: 1.0000 - val_loss: 0.0182 - val_acc: 1.0000
Epoch 24/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0027 - acc: 1.0000 - val_loss: 0.0181 - val_acc: 1.0000
Epoch 25/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0155 - acc: 1.0000 - val_loss: 0.0158 - val_acc: 1.0000
Epoch 26/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0240 - acc: 1.0000 - val_loss: 0.0147 - val_acc: 1.0000
Epoch 27/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0085 - acc: 1.0000 - val_loss: 0.0142 - val_acc: 1.0000
Epoch 28/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0147 - acc: 1.0000 - val_loss: 0.0141 - val_acc: 1.0000
Epoch 29/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0087 - acc: 1.0000 - val_loss: 0.0142 - val_acc: 1.0000
Epoch 30/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0301 - acc: 0.9000 - val_loss: 0.0152 - val_acc: 1.0000
Epoch 31/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0170 - acc: 1.0000 - val_loss: 0.0144 - val_acc: 1.0000
Epoch 32/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0180 - acc: 1.0000 - val_loss: 0.0136 - val_acc: 1.0000
Epoch 33/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0093 - acc: 1.0000 - val_loss: 0.0129 - val_acc: 1.0000
Epoch 34/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0286 - acc: 1.0000 - val_loss: 0.0097 - val_acc: 1.0000
Epoch 35/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0058 - acc: 1.0000 - val_loss: 0.0093 - val_acc: 1.0000
Epoch 36/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0264 - acc: 1.0000 - val_loss: 0.0070 - val_acc: 1.0000
Epoch 37/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0016 - acc: 1.0000 - val_loss: 0.0070 - val_acc: 1.0000
Epoch 38/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0035 - acc: 1.0000 - val_loss: 0.0068 - val_acc: 1.0000
Epoch 39/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0049 - acc: 1.0000 - val_loss: 0.0065 - val_acc: 1.0000
Epoch 40/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0139 - acc: 1.0000 - val_loss: 0.0056 - val_acc: 1.0000
Epoch 41/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0062 - acc: 1.0000 - val_loss: 0.0055 - val_acc: 1.0000
Epoch 42/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0028 - acc: 1.0000 - val_loss: 0.0052 - val_acc: 1.0000
Epoch 43/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0156 - acc: 1.0000 - val_loss: 0.0052 - val_acc: 1.0000
Epoch 44/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0195 - acc: 1.0000 - val_loss: 0.0047 - val_acc: 1.0000
Epoch 45/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0027 - acc: 1.0000 - val_loss: 0.0043 - val_acc: 1.0000
Epoch 46/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0030 - acc: 1.0000 - val_loss: 0.0041 - val_acc: 1.0000
Epoch 47/50
10/10 [==============================] - 0s 2ms/step - loss: 7.9729e-04 - acc: 1.0000 - val_loss: 0.0041 - val_acc: 1.0000
Epoch 48/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0061 - acc: 1.0000 - val_loss: 0.0041 - val_acc: 1.0000
Epoch 49/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0025 - acc: 1.0000 - val_loss: 0.0038 - val_acc: 1.0000
Epoch 50/50
10/10 [==============================] - 0s 2ms/step - loss: 0.0021 - acc: 1.0000 - val_loss: 0.0039 - val_acc: 1.0000

TEST
(.venv) dani@NB-Daniele:/mnt/c/Users/daniele.angelini/mycroft-precise-dev/mycroft-precise-dev$ precise-test tars-start.net tars-start/
Loading wake-word...
Loading not-wake-word...
2021-05-20 11:53:05.242040: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
2021-05-20 11:53:05.248528: I tensorflow/core/platform/profile_utils/cpu_utils.cc:94] CPU Frequency: 2904000000 Hz
2021-05-20 11:53:05.252104: I tensorflow/compiler/xla/service/service.cc:150] XLA service 0x43832f0 executing computations on platform Host. Devices:
2021-05-20 11:53:05.252860: I tensorflow/compiler/xla/service/service.cc:158] StreamExecutor device (0): ,
Using TensorFlow backend.
Traceback (most recent call last):
File "/mnt/c/users/daniele.angelini/mycroft-precise-dev/mycroft-precise-dev/.venv/bin/precise-test", line 33, in
sys.exit(load_entry_point('mycroft-precise', 'console_scripts', 'precise-test')())
File "/mnt/c/users/daniele.angelini/mycroft-precise-dev/mycroft-precise-dev/precise/scripts/base_script.py", line 49, in run_main
script.run()
File "/mnt/c/users/daniele.angelini/mycroft-precise-dev/mycroft-precise-dev/precise/scripts/test.py", line 52, in run
predictions = Listener.find_runner(args.model)(args.model).predict(inputs)
File "/mnt/c/users/daniele.angelini/mycroft-precise-dev/mycroft-precise-dev/precise/network_runner.py", line 85, in init
self.model = load_precise_model(model_name)
File "/mnt/c/users/daniele.angelini/mycroft-precise-dev/mycroft-precise-dev/precise/model.py", line 54, in load_precise_model
return load_keras().models.load_model(model_name)
File "/mnt/c/users/daniele.angelini/mycroft-precise-dev/mycroft-precise-dev/.venv/lib/python3.7/site-packages/keras/models.py", line 242, in load_model
model_config = json.loads(model_config.decode('utf-8'))
AttributeError: 'str' object has no attribute 'decode'

Result of test sholuld be ok .???

[JoySkipper]

So I know this is a bit old but I'm having the same issue. Looks to be for me (as far as my amateur mind can tell) because precise only works for python versions <=3.6 and my raspberry pi 4 came with python 3.7 installed and thus built the venv on 3.7.

Thus, str.decode() which is a python2 skeleton that was finally dropped in 3.7, so it no longer is necessary and breaks the code. Removing the '.decode('utf-8') fixes that line, but then I run into the same problem again...and again...etc.

So instead of updating my entire local repo to be compatible with python 3.7 which could be an endless rabbit hole that'd break as soon as there's another update, I tried to re-clone the repo and run setup.sh using a build of python3.6 that I'd put onto my pi instead. I'll update here if this is successful.

Looks like the dev is working on uploading the tensorflow backend at the moment (the source of all the warnings in the top). Hopefully their next step would be to make precise compatible with 3.7? I'm tempted to try and do so myself but it's probably a lot hairier to do than it appears at first glance and I'm afraid my hunches are way off base. @krisgesling am I on the right track or do I have no idea what I'm doing? I don't mind working on updating compatibility to 3.7 to help out if I'm on the right track here, but I'm not confident enough that I am at the moment and don't want to put all that work in if it doesn't make sense or isn't actually the source of the problem.

[JoySkipper]

Update on building with Python 3.6. It worked, but it did take quite a bit of fiddling to get working. Mainly the issue revolved around installing the dependencies such as scipy which was weirdly difficult. I had to install a lot of scipy's dependencies by hand even though pip should have taken care of that for me.

Anyways, after installing precise I had to upgrade tensorflow to 1.14.0 even though it warns that precise only works for tensorflow as 1.13. When I ran it with 1.13 I got this error and didn't have the know-how to fix it :

tensorflow/tensorflow#23386 (comment)

I basically upgraded to 1.14 as a blind guess and it fixed the issue I was having. Point is precise training, testing, and listening is working now for me so perhaps I had the right idea with the python version issue.

[el-tocino]

You can use the latest tf1.x version for precise (.3, at any rate)