feat: Add missing operators for AIDGE model benchmarking

Context

I am working on benchmarking AIDGE models by fully creating them in AIDGE, setting their weights randomly, and exporting them to C++. The goal is to verify if the outputs in C++ match the forward pass results in AIDGE. You can find it here

To achieve this, I have added several missing operators necessary for running models such as:

• SqueezeNet
• ResNet18 / ResNet50
• MLP for MNIST
• MobileNet

These operators are now available in the export pipeline to ensure proper benchmarking.

Modified Files

• Operators & Kernels

  • aidge_export_cpp/kernels/batchnorm.hpp: Added missing batch normalization operator.
  • aidge_export_cpp/kernels/concat.hpp: Added concatenation operator.
  • aidge_export_cpp/kernels/pad.hpp: Added padding operator.
  • aidge_export_cpp/kernels/pooling.hpp: Added AvgPooling2D.
  • aidge_export_cpp/kernels/softmax.hpp: Added softmax operator.

• Python Script

  • aidge_export_cpp/operators.py: Updated operator handling to include new additions.

• Configuration Templates

  • aidge_export_cpp/templates/configuration/*: Added configurations for batchnorm, concat, pad, and softmax.

• Kernel Forward Templates

  • aidge_export_cpp/templates/kernel_forward/*: Added missing forward implementations for batchnorm, concat, pad, and softmax.

• Unit Tests

  • aidge_export_cpp/unit_tests/test_export.py: Updated tests to validate newly added operators.

---

Detailed Major Modifications

• Implemented missing operators to support model export and validation in C++.
• Ensured each new operator follows the existing API.
• Updated test cases to validate correctness of the exported model execution.
• Added the possibility of setting the weights inside the model randomly or with a default value to stay deterministic for the unit tests

TODO

• Add some of the missing args of Pad2D. Do add assertion for now that attribute for padding is constant fill (i.e.do not support other attributes such as mirror etc...).

changed title from fix: Better control over random number generation to feat: Add missing operators for AIDGE model benchmarking

requested review from @cmoineau and @cguillon

@axelfarr FYI

assigned to @gallasko

added Feature 🚀 Unit-test 🧪 labels

See review for softmax commit

changed milestone to %aidge v0.6.0

added StatusChanges required label

Ref review for concat

Some issues with batch on HW / WH indexing.

Please for the unit tests, add a non-1 batch dimension and use non square HW, i.e. set H != W.

added 10 commits

• 7d5b8ce3 - feat: Added Batchnorm2d export op
• 3136f3c2 - feat: Added Concat export op
• a298e118 - feat: Added Pad2d export op
• 4a0dbb6f - feat: Added more tests
• 2c410ef5 - fix: Better control over random number generation
• b2dbf4da - fix: Some unit test failing
• a4c4b922 - fix: Only allow possitive value of variances in batchnorm
• c0763c37 - feat: Added AvgPooling2D export op
• 1bac314b - Pad works for with size mismatch
• 0b984f82 - Pad works now for all size of batch and rectangular aspects

Compare with previous version

@mnewson FYI. Btw it may be worth adding your ConvNeXt model to the aidge_benchmark at a later date...

added 3 commits

• c9893d2b - Renamed correctly pad forward
• 158b1569 - Fix average pooling
• 4285dbc9 - Added more test cases to the test suite

Compare with previous version

added 5 commits

• d0d6193c - Added more ut for batchnorm
• 6b00b5cc - fixed indentation in pad
• 116bb0c1 - Fix batchnorm with denser input
• 2131467a - Assert if concat is done on anorther axis than 1
• 937b8fe6 - Pass on softmax op

Compare with previous version

mentioned in merge request !33 (closed)

changed milestone to %aidge v0.7.0

mentioned in merge request aidge_core!403 (merged)

added 6 commits

• 12e07914 - feat: Added AvgPooling2D export op
• ef1ccd8b - fix: Better pad op implementation
• 8291303d - feat: Added more test cases to the test suite
• 3eacad44 - fix: Batch support for batchnorm
• 5dfeac42 - feat: Softmax works with any number of dimensions
• efa5e9bb - feat: Concat works for any axis

Compare with previous version

resolved all threads

added 21 commits

• efa5e9bb...da88eb0f - 11 earlier commits
• 3d6707b0 - feat: Added more tests
• 939fdfc4 - fix: Better control over random number generation
• 6612538a - fix: Some unit test failing
• 9c5d3b0b - fix: Only allow possitive value of variances in batchnorm
• eefac137 - feat: Added AvgPooling2D export op
• 649a27e8 - fix: Better pad op implementation
• 52475cd1 - feat: Added more test cases to the test suite
• fb5ebdb2 - fix: Batch support for batchnorm
• 0cb91a4e - feat: Softmax works with any number of dimensions
• 520cabd6 - feat: Concat works for any axis

Compare with previous version

added 4 commits

• 5ec88a78 - feat: Added more test cases to the test suite
• c436eeb7 - fix: Batch support for batchnorm
• 2f808fe9 - feat: Softmax works with any number of dimensions
• e63fdee8 - feat: Concat works for any axis

Compare with previous version

added 26 commits

• e63fdee8...6a5441df - 12 commits from branch eclipse/aidge:dev
• 6a5441df...1044a6b6 - 4 earlier commits
• 8f3170bf - feat: Added more tests
• 1aecd6c3 - fix: Better control over random number generation
• 0ee3b761 - fix: Some unit test failing
• 90f53c3d - fix: Only allow possitive value of variances in batchnorm
• f4815323 - feat: Added AvgPooling2D export op
• 8c84b470 - fix: Better pad op implementation
• d521dbed - feat: Added more test cases to the test suite
• d6064e31 - fix: Batch support for batchnorm
• 141e7c3e - feat: Softmax works with any number of dimensions
• 0f32cc6d - feat: Concat works for any axis

Compare with previous version

resolved all threads

Fine for me.

@cmoineau, I let you merge optionally as you may arbitrate better than me currently opened MR for this repo.

We can merge :)