Hi,
I previously had a custom trigger phrase (“computer”) using precise-engine back when I had Mycroft. I am attempting to get it working on OVOS now using precise-lite-trainer. I have an error I’m not sure how to solve. To quickly walk through how I solved prior errors (in case I still did something wrong there), I had the following error:
ValueError: The filepath provided must end in .keras (Keras model format). Received: filepath=computer/model/computer
I attempted to solve the error by creating an empty zip file ending in .keras instead of .zip. I had also gotten a different error, which I didn’t record, but I solved that by removing the use_multiprocessing=True line in train.py. I’m sure there’s a better way to handle that error. I’m open to suggestions on that, but it isn’t my primary error.
The main error I need help on is KeyError: "There is no item named 'config.json' in the archive". I don’t see any information on how to create a .json file anywhere. I don’t have much knowledge on these machine learning libaries, at least for this type of usage with keras, etc. I attempted to glean information I could about how the machine learning model was structured and had AI attempt to create its best guess for a JSON file. Below is the error I get from the attempted JSON file guestimated by AI:
Loading from computer/model/computer.keras...
Traceback (most recent call last):
File "/home/pi/precise-lite-trainer/training.py", line 9, in <module>
trainer = PreciseTrainer(model_path, folder, epochs=100, log_dir=log_dir)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/pi/precise-lite-trainer/precise_trainer/train.py", line 75, in __init__
self.model = get_model(model, model_params)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/pi/precise-lite-trainer/precise_trainer/model.py", line 56, in get_model
model = load_model(model_name)
^^^^^^^^^^^^^^^^^^^^^^
File "/home/pi/.venvs/training/.venv/lib/python3.11/site-packages/keras/src/saving/saving_api.py", line 189, in load_model
return saving_lib.load_model(
^^^^^^^^^^^^^^^^^^^^^^
File "/home/pi/.venvs/training/.venv/lib/python3.11/site-packages/keras/src/saving/saving_lib.py", line 367, in load_model
return _load_model_from_fileobj(
^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/pi/.venvs/training/.venv/lib/python3.11/site-packages/keras/src/saving/saving_lib.py", line 444, in _load_model_from_fileobj
model = _model_from_config(
^^^^^^^^^^^^^^^^^^^
File "/home/pi/.venvs/training/.venv/lib/python3.11/site-packages/keras/src/saving/saving_lib.py", line 433, in _model_from_config
model = deserialize_keras_object(
^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/pi/.venvs/training/.venv/lib/python3.11/site-packages/keras/src/saving/serialization_lib.py", line 694, in deserialize_keras_object
cls = _retrieve_class_or_fn(
^^^^^^^^^^^^^^^^^^^^^^
File "/home/pi/.venvs/training/.venv/lib/python3.11/site-packages/keras/src/saving/serialization_lib.py", line 803, in _retrieve_class_or_fn
raise TypeError(
TypeError: Could not locate class 'Sequential'. Make sure custom classes are decorated with `@keras.saving.register_keras_serializable()`. Full object config: {'class_name': 'Sequential', 'config': {'name': 'sequential', 'layers': [{'class_name': 'GRU', 'config': {'name': 'net', 'trainable': True, 'batch_input_shape': [None, 29, 13], 'dtype': 'float32', 'units': 20, 'activation': 'tanh', 'recurrent_activation': 'sigmoid', 'use_bias': True, 'kernel_initializer': {'class_name': 'GlorotUniform', 'config': {'seed': None}}, 'recurrent_initializer': {'class_name': 'Orthogonal', 'config': {'gain': 1.0, 'seed': None}}, 'bias_initializer': {'class_name': 'Zeros', 'config': {}}, 'kernel_regularizer': None, 'recurrent_regularizer': None, 'bias_regularizer': None, 'activity_regularizer': None, 'kernel_constraint': None, 'recurrent_constraint': None, 'bias_constraint': None, 'dropout': 0.0, 'recurrent_dropout': 0.0, 'return_sequences': False, 'return_state': False, 'go_backwards': False, 'stateful': False, 'unroll': False, 'time_major': False, 'reset_after': False, 'zero_output_for_mask': False}}, {'class_name': 'Dense', 'config': {'name': 'dense', 'trainable': True, 'dtype': 'float32', 'units': 1, 'activation': 'linear', 'use_bias': True, 'kernel_initializer': {'class_name': 'GlorotUniform', 'config': {'seed': None}}, 'bias_initializer': {'class_name': 'Zeros', 'config': {}}, 'kernel_regularizer': None, 'bias_regularizer': None, 'activity_regularizer': None, 'kernel_constraint': None, 'bias_constraint': None}}]}, 'keras_version': '3.8.0', 'backend': 'tensorflow'}
Here’s the actual JSON file I attempted with precise-lite:
{
"class_name": "Sequential",
"config": {
"name": "sequential",
"layers": [
{
"class_name": "GRU",
"config": {
"name": "net",
"trainable": true,
"batch_input_shape": [null, 29, 13],
"dtype": "float32",
"units": 20,
"activation": "tanh",
"recurrent_activation": "sigmoid",
"use_bias": true,
"kernel_initializer": {
"class_name": "GlorotUniform",
"config": {
"seed": null
}
},
"recurrent_initializer": {
"class_name": "Orthogonal",
"config": {
"gain": 1.0,
"seed": null
}
},
"bias_initializer": {
"class_name": "Zeros",
"config": {}
},
"kernel_regularizer": null,
"recurrent_regularizer": null,
"bias_regularizer": null,
"activity_regularizer": null,
"kernel_constraint": null,
"recurrent_constraint": null,
"bias_constraint": null,
"dropout": 0.0,
"recurrent_dropout": 0.0,
"return_sequences": false,
"return_state": false,
"go_backwards": false,
"stateful": false,
"unroll": false,
"time_major": false,
"reset_after": false,
"zero_output_for_mask": false
}
},
{
"class_name": "Dense",
"config": {
"name": "dense",
"trainable": true,
"dtype": "float32",
"units": 1,
"activation": "linear",
"use_bias": true,
"kernel_initializer": {
"class_name": "GlorotUniform",
"config": {
"seed": null
}
},
"bias_initializer": {
"class_name": "Zeros",
"config": {}
},
"kernel_regularizer": null,
"bias_regularizer": null,
"activity_regularizer": null,
"kernel_constraint": null,
"bias_constraint": null
}
}
]
},
"keras_version": "3.8.0",
"backend": "tensorflow"
}
I likely have some stuff wrong because I gave the AI as much information as I knew about what I was doing, but some stuff was inferred and likely incorrect.
Here’s the script I tried using to actually run precise-trainer:
from precise_trainer import PreciseTrainer
model_name = "computer"
folder = f"computer/{model_name}" # dataset here
model_path = f"computer/model/{model_name}.keras" # save here
log_dir = f"./logs/fit/{model_name}" # for tensorboard
# train a model
trainer = PreciseTrainer(model_path, folder, epochs=100, log_dir=log_dir)
model_file = trainer.train()
# Data: <TrainData wake_words=155 not_wake_words=89356 test_wake_words=39 test_not_wake_words=22339>
# Loading wake-word...
# Loading not-wake-word...
# Loading wake-word...
# Loading not-wake-word...
# Inputs shape: (81602, 29, 13)
# Outputs shape: (81602, 1)
# Test inputs shape: (20486, 29, 13)
# Test outputs shape: (20486, 1)
# Model: "sequential"
# _________________________________________________________________
# Layer (type) Output Shape Param #
# =================================================================
# net (GRU) (None, 20) 2100
#
# dense (Dense) (None, 1) 21
#
# =================================================================
# Total params: 2,121
# Trainable params: 2,121
# Non-trainable params: 0
# .....
# _________________________________________________________________
# Epoch 1280/1379
# 157/160 [============================>.] - ETA: 0s - loss: 0.0308 - accuracy: 0.9868
# ....
# Wrote to /home/miro/PycharmProjects/ovos-audio-classifiers/trained/hey_computer/model.tflite
trainer.test()
# === Counts ===
# False Positives: 2
# True Negatives: 20445
# False Negatives: 2
# True Positives: 37
#
# === Summary ===
# 20482 out of 20486
# 99.98%
#
# 0.01% false positives
# 5.13% false negatives
Please advise on what I need to do to get things working. I honestly don’t know what I’m doing and I’m aimlessly guessing at this point. Thanks!