Mdt

===================

Changed
-------
- Small fix to the buffer allocation flags. The post-processing failed sometimes.

===================
Upgrade to match the latest version of MOT, which contains a vast number of changes regarding buffer and kernel allocation.

===================

Added
-----
- Adds CL load balancing option to the model processing functions.

Changed
-------
- Updates the credits with sponsors information.
- Renamed extra_prior to prior in the compartment and composite model definitions.

Other
-----
- Following changes in MOT to allow easier cl environment initialization.

===================

Added
-----
- Adds some introduction text to the mdt.conf file.

Other
-----
- Small bugfix in the noddi-dti computation. MD was changed in place.

===================
- Upped the voxel batch size again as the previous change did not provide a solution.

===================

Changed
-------
- Changed the default optimization batch size from 100k to 10k voxels per run. Due to changes in MOT, the current batch size fails for some users.

Changed
-------
- Removed the NODDI-DTI maps from the Tensor and KurtosisTensor post-processing. They do not seem to be widely used.
For users that do desire these maps there are two ways to go about it:

1) Reinstate the post-processing. This can be done by going to the
Tensor compartment model, and add ``noddi_dti_maps`` to the
extra_optimization_maps list. This function can be imported from:
``from mdt.lib.post_processing import noddi_dti_maps``.

2) Use the function mdt.compute_noddi_dti() to compute the maps manually
based on optimization results. Typically used as:

.. code-block:: python

results = mdt.fit_model(model, input_data, ...)
noddi_dti = mdt.compute_noddi_dti(model, input_data, results)