Hello! I’m still working on Simple audio recognition: Recognizing keywords | TensorFlow Core and I was unable to use one of my own audio files in place of the training files. When I try to use it, I get the following error: InvalidArgumentError: Dimension -6522 must be >= 0
[[{{node zeros}}]] [Op:IteratorGetNext]
What am I doing wrong?
https://drive.google.com/drive/folders/1QRgtnXqaT9Eu29BENzDiVceBb62KVx9k?usp=sharing
Hi @Henry_Sommer , The drive link shared by you is not an open link and it asks for access.
@Divvya_Saxena Sorry! got that fixed.
Hello @Henry_Sommer
When you check the shape of in the original WAV file in your dataset, what does your output show?
For example, when using the original tutorial’s mini Speech Commands dataset:
DATASET_PATH = 'data/mini_speech_commands'
test_file = tf.io.read_file(DATASET_PATH+'/down/0a9f9af7_nohash_0.wav')
test_audio, _ = tf.audio.decode_wav(contents=test_file)
test_audio.shape
Output:
TensorShape([13654, 1])
where 1 is a channel dimension (mono). So, the waveform on its own is kind of like a 1D array.
If you can share some code, that might help us help you with debugging.