Cannot Create Training Dataset DLCma 2.2b7

I am having trouble creating the training dataset in DLCma 2.2b7

here is the error message:

Traceback (most recent call last):
File “C:\Users\Fish_Behavior.conda\envs\dlc-WindowsGPU\lib\site-packages\deeplabcut\gui\”, line 375, in create_training_dataset
self.config, num_shuffles, Shuffles=[self.shuffle.GetValue()]
File “C:\Users\Fish_Behavior.conda\envs\dlc-WindowsGPU\lib\site-packages\deeplabcut\generate_training_dataset\”, line 97, in create_multianimaltraining_dataset
cfg, full_training_path, windows2linux
File “C:\Users\Fish_Behavior.conda\envs\dlc-WindowsGPU\lib\site-packages\deeplabcut\generate_training_dataset\”, line 598, in merge_annotateddatasets
data = pd.read_hdf(file_path, “df_with_missing”)
File “C:\Users\Fish_Behavior.conda\envs\dlc-WindowsGPU\lib\site-packages\pandas\io\”, line 428, in read_hdf
File “C:\Users\Fish_Behavior.conda\envs\dlc-WindowsGPU\lib\site-packages\pandas\io\”, line 814, in select
return it.get_result()
File “C:\Users\Fish_Behavior.conda\envs\dlc-WindowsGPU\lib\site-packages\pandas\io\”, line 1829, in get_result
results = self.func(self.start, self.stop, where)
File “C:\Users\Fish_Behavior.conda\envs\dlc-WindowsGPU\lib\site-packages\pandas\io\”, line 798, in func
return, stop=_stop, where=_where, columns=columns)
File “C:\Users\Fish_Behavior.conda\envs\dlc-WindowsGPU\lib\site-packages\pandas\io\”, line 3061, in read
ax = self.read_index(f"axis{i}", start=_start, stop=_stop)
File “C:\Users\Fish_Behavior.conda\envs\dlc-WindowsGPU\lib\site-packages\pandas\io\”, line 2759, in read_index
index = self.read_index_node(node, start=start, stop=stop)
File “C:\Users\Fish_Behavior.conda\envs\dlc-WindowsGPU\lib\site-packages\pandas\io\”, line 2838, in read_index_node
data = node[start:stop]
File “C:\Users\Fish_Behavior.conda\envs\dlc-WindowsGPU\lib\site-packages\tables\”, line 685, in getitem
return, stop, step)
File “C:\Users\Fish_Behavior.conda\envs\dlc-WindowsGPU\lib\site-packages\tables\”, line 821, in read
listarr = self._read_array(start, stop, step)
File “tables\hdf5extension.pyx”, line 2155, in tables.hdf5extension.VLArray._read_array
ValueError: cannot set WRITEABLE flag to True of this array

@Kyle_Newton, what version of pytables, numpy and pandas do you have installed? Could it be that some of your h5 data files were created with versions anterior to the current ones?

i just updated them all and have numpy 1.19.1, pandas 1.0.5, and pytables 3.4.3.

I retried creating the training dataset and I get the same error message

Should I recreate the project and move the labeled data into the new folders?

The issue is from the h5 files, rather than the creation of the dataset per se.
If you use the labeling GUI, can you load and visualize all your labeled data?

yes I can load and visualize the labels.

Oddly enough, I decided to start over from scratch, create a new environment,reinstall DLC2.2b7 and now it magically works. Weird.

1 Like