.. ENOT reference documentation master file.

############################
ENOT Reference Documentation
############################

You can find ENOT reference documentation here. If you experience lack of
documentation, you probably should contact ENOT team to complete it or to make
it more clear.

**Before proceeding to examples or documentation, we recommend you to read this
page carefully.**

Enot-autodl framework supports
`neural architecture search (NAS) <https://arxiv.org/abs/2005.11074>`_,
`quantization <https://arxiv.org/abs/2106.08295>`_ and
`pruning <https://arxiv.org/abs/2003.03033>`_.

**Neural architecture search** is a procedure which aims to find suitable
architecture from a specific search space - set of architectures to consider.
This procedure is the most resource-intensive, and requires the most
programming and experimenting among methods listed above.

Tutorials:

* :ref:`1. Tutorial - getting started <E1 ref>`
* :ref:`2. Tutorial - search space autogeneration <E2 ref>`
* :ref:`3. Tutorial - custom model <E3 ref>`
* :ref:`4. Tutorial - using latency optimization <E4 ref>`
* :ref:`5. Tutorial - latency calculation <E5 ref>`
* :ref:`6. Tutorial - search with the specified latency <E6 ref>`
* :ref:`7. Tutorial - resolution search for image classification <E7 ref>`
* :ref:`8. Tutorial - search space autogeneration (EfficientNet-V2 S) <E8 ref>`
* :ref:`9. Tutorial - metric learning <E9 ref>`
* :ref:`10. Tutorial - adding custom operations for model builder <E10 ref>`

**Neural network quantization** attempts to make weights of neural network discrete
(instead of using their floating-point representation, usually float32 format).
Quantization decreases model size by a factor of 4 by using int8 data type.
On NVIDIA GPUs and Intel CPUs quantization may also provide noticeable
inference time decrease.

Tutorials:

* :ref:`11. Experimental - Tutorial - automatic quantization for enot-lite <E11 ref>`

**Neural network pruning** removes redundant (or least important) channels (or
features) from neural network. Our framework implements structured pruning
(removing channels or neurons, not separate connections).

Tutorials:

* :ref:`12. Experimental - Tutorial - automatic pruning <E12 ref>`
* :ref:`13. Experimental - Tutorial - automatic pruning (manual) <E13 ref>`

Before reading the documentation, we recommend you to look at the
:ref:`ENOT Tutorials` to clarify basic notions and concepts of the framework.


*****************
Table of contents
*****************

Packages are listed in descending importance order.

.. toctree::
   :maxdepth: 1

   Autogeneration <autogeneration>
   Autogeneration transforms <autogeneration_transformation_parameters>
   Models <models>
   Optimize <optimize>
   Distributed pretrain <distributed_pretrain>
   Latency <latency>
   Distributed <distributed>
   Quantization <quantization>
   Pruning <pruning>
   Utils <utils>
   Logging <logging>
   Visualization <visualization>
   Experimental <experimental>
   Converting dataloader items to PyTorch model inputs <dataloader2model>
   Modules <modules>
   Model builders <model_builders>