Error in training a custom classifier using Ubuntu CLI

[AdityaPanigrahy]

So, I made a few local changes to log the epoch history using tensorboard callbacks (mostly to plot the training and evaluation loss). It worked for few models. I then restored the repository to its original form, stashing the local changes. But now on training I recieve this error. analyze.py and embeddings.py work fine though. Only train.py has this issue.

multiprocessing.pool.RemoteTraceback: 
"""
Traceback (most recent call last):
  File "/home/birdlab/miniconda3/envs/birdnet/lib/python3.10/multiprocessing/pool.py", line 125, in worker
    result = (True, func(*args, **kwds))
  File "/home/birdlab/softs/BirdNet_CLI/BirdNET-Analyzer/train.py", line 60, in _loadAudioFile
    embeddings = model.embeddings(batch_sig)
  File "/home/birdlab/softs/BirdNet_CLI/BirdNET-Analyzer/model.py", line 640, in embeddings
    features = INTERPRETER.get_tensor(OUTPUT_LAYER_INDEX)
  File "/home/birdlab/miniconda3/envs/birdnet/lib/python3.10/site-packages/tensorflow/lite/python/interpreter.py", line 888, in get_tensor
    return self._interpreter.GetTensor(tensor_index, subgraph_index)
ValueError: Tensor data is null. Run allocate_tensors() first
"""

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "/home/birdlab/softs/BirdNet_CLI/BirdNET-Analyzer/train.py", line 437, in <module>
    trainModel()
  File "/home/birdlab/softs/BirdNet_CLI/BirdNET-Analyzer/train.py", line 213, in trainModel
    x_train, y_train, labels = _loadTrainingData(cfg.TRAIN_CACHE_MODE, cfg.TRAIN_CACHE_FILE, on_data_load_end)
  File "/home/birdlab/softs/BirdNet_CLI/BirdNET-Analyzer/train.py", line 172, in _loadTrainingData
    result = task.get()
  File "/home/birdlab/miniconda3/envs/birdnet/lib/python3.10/multiprocessing/pool.py", line 774, in get
    raise self._value
ValueError: Tensor data is null. Run allocate_tensors() first

Can someone help me out here to as what exactly is happening?

[Josef-Haupt]

Hard to tell, if embeddings works on its own, this piece of code should work too, in theory atleast. What exactly did you change in your local repo?

[AdityaPanigrahy]

I had just added a tensor-flow callbacks option to save the epoch history logs and plot the custom loss for training vs validation. But I had stashed the changes and it was back to original code when this error popped up.