Error when concatenating data in gather_recording()

[ma-sauter]

The following error is produced when the epoch number is an integer multiple of the chunk size in the recorder. It tries to concatenate the data from the database to the currently computed data but can't since the non existing current data has a different dimensionality.

Might be fixed by using dump_records() and dropping the concatenation completely.

---

ValueError Traceback (most recent call last)
Cell In[12], line 1
----> 1 train_report = train_recorder.gather_recording()
2 test_report = test_recorder.gather_recording()

File ~/software/ZnNL/znnl/training_recording/jax_recording.py:599, in JaxRecorder.gather_recording(self, selected_properties)
597 for item, data in selected_data.items():
598 print(item)
--> 599 selected_data[item] = onp.concatenate((db_data[item], data), axis=0)
601 except FileNotFoundError: # There is no database.
602 pass

File <array_function internals>:180, in concatenate(*args, **kwargs)

ValueError: all the input arrays must have same number of dimensions, but the array at index 0 has 5 dimension(s) and the array at index 1 has 1 dimension(s)

[SamTov]

This arises because of the following:

db_data = self._data_storage.fetch_data(selected_properties)
# Add db data to the selected data dict.
for item, data in selected_data.items():
    selected_data[item] = onp.concatenate((db_data[item], data), axis=0)

However, if the finalising of recorders is done after training, as outlined in the examples, this should not be necessary, so the concatenation can simply be removed and only the DB data returned.