0

I am trying to run the TFlite audio classification notebook, but swap out the birds dataset and use the ravdess speech emotion dataset instead. I reshaped the ravdess folder structure to match the birds dataset structure which is:

  ravdess
  ├ Train
  | ├ Anger
  | ├ Surprise
  | ├ Calm
  | ├ Sad
  | ├ Fearful
  | ├ Happy
  | ├ Neutral
  | ├ Disgust
  ├ Test
  | ├ Anger
  | ├ Surprise
  | ├ Calm
  | ├ Sad
  | ├ Fearful
  | ├ Happy
  | ├ Neutral
  | ├ Disgust

There are 8 classes instead of the 5 for the bird's dataset, but my dataset and the birds have the same folder structure, just a different number of classes as subfolders.

Here is my code for the model, which is taken straight from the notebook:

!sudo apt -y install libportaudio2
!pip install tflite-model-maker

import tensorflow as tf
import tflite_model_maker as mm
from tflite_model_maker import audio_classifier
import os

import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

import itertools
import glob
import random

from IPython.display import Audio, Image
from scipy.io import wavfile

print(f"TensorFlow Version: {tf.__version__}")
print(f"Model Maker Version: {mm.__version__}")

# I decided to download the zip, and unzip the file, because sometimes attempting to # train models that have dataloaded from google drive is slow. 

!unzip /content/ravdess_data-20221005T170225Z-001.zip

ravdess_dataset_folder = "/content/ravdess_data"

data_dir = ravdess_dataset_folder

spec = audio_classifier.YamNetSpec(
    keep_yamnet_and_custom_heads=True,
    frame_step=3 * audio_classifier.YamNetSpec.EXPECTED_WAVEFORM_LENGTH,
    frame_length=6 * audio_classifier.YamNetSpec.EXPECTED_WAVEFORM_LENGTH)

train_data = audio_classifier.DataLoader.from_folder(
    spec, os.path.join(data_dir, 'train'), cache=True)
train_data, validation_data = train_data.split(0.8)
test_data = audio_classifier.DataLoader.from_folder(
    spec, os.path.join(data_dir, 'test'), cache=True)

batch_size = 5
epochs = 100

print('Training the model')
model = audio_classifier.create(
    train_data,
    spec,
    validation_data,
    batch_size=batch_size,
    epochs=epochs)

https://colab.research.google.com/github/googlecodelabs/odml-pathways/blob/main/audio_classification/colab/model_maker_audio_colab.ipynb#scrollTo=rwUA9u4oWoCR

This is the error I am getting:

InvalidArgumentError                      Traceback (most recent call last)
<ipython-input-46-a3613f172d24> in <module>
      8     validation_data,
      9     batch_size=batch_size,
---> 10     epochs=epochs)

4 frames
/usr/local/lib/python3.7/dist-packages/tensorflow_examples/lite/model_maker/core/task/audio_classifier.py in create(cls, train_data, model_spec, validation_data, batch_size, epochs, model_dir, do_train, train_whole_model)
    137         train_whole_model=train_whole_model)
    138     if do_train:
--> 139       task.train(train_data, validation_data, epochs, batch_size)
    140     return task
    141 

/usr/local/lib/python3.7/dist-packages/tensorflow_examples/lite/model_maker/core/task/audio_classifier.py in train(self, train_data, validation_data, epochs, batch_size)
     55           train_ds,
     56           validation_ds,
---> 57           callbacks=self._keras_callbacks(self.model_spec.model_dir))
     58 
     59   def create_model(self, num_classes, train_whole_model):

/usr/local/lib/python3.7/dist-packages/tensorflow_examples/lite/model_maker/core/task/model_spec/audio_spec.py in run_classifier(self, model, epochs, train_ds, validation_ds, **kwargs)
    490 
    491     hist = model.fit(
--> 492         train_ds, validation_data=validation_ds, epochs=epochs, **kwargs)
    493     return hist
    494 

/usr/local/lib/python3.7/dist-packages/keras/utils/traceback_utils.py in error_handler(*args, **kwargs)
     65     except Exception as e:  # pylint: disable=broad-except
     66       filtered_tb = _process_traceback_frames(e.__traceback__)
---> 67       raise e.with_traceback(filtered_tb) from None
     68     finally:
     69       del filtered_tb

/usr/local/lib/python3.7/dist-packages/tensorflow/python/eager/execute.py in quick_execute(op_name, num_outputs, inputs, attrs, ctx, name)
     53     ctx.ensure_initialized()
     54     tensors = pywrap_tfe.TFE_Py_Execute(ctx._handle, device_name, op_name,
---> 55                                         inputs, attrs, num_outputs)
     56   except core._NotOkStatusException as e:
     57     if name is not None:

InvalidArgumentError: Graph execution error:

2 root error(s) found.
  (0) INVALID_ARGUMENT:  Data too short when trying to read string
     [[{{node DecodeWav}}]]
     [[IteratorGetNext]]
     [[categorical_crossentropy/softmax_cross_entropy_with_logits/Shape_2/_6]]
  (1) INVALID_ARGUMENT:  Data too short when trying to read string
     [[{{node DecodeWav}}]]
     [[IteratorGetNext]]
0 successful operations.
0 derived errors ignored. [Op:__inference_train_function_59314]

Things that I have tried so far to fix this:

Paul Fentress
  • 51
  • 3

0 Answers0