When using the web training, it's possible that the training axistags might
not match those of the data when running in headless mode, since the
training might happen on precomputed chunks, for example, and the headless
run might happen on local .n5 files. This commit makes ilastik check
the dataset itself for the axis tags rather than trusting the training data
When using the web training, it's possible that the training axistags might not match those of the data when running in headless mode, since the training might happen on precomputed chunks, for example, and the headless run might happen on local .n5 files. This commit makes ilastik check the dataset itself for the axis tags rather than trusting the training data
This assumes that the ilastik installation in the HPC has https://github.com/ilastik/ilastik/pull/2241