Depthwise convolution for oneAPI

[laurilaatu]

Description

:memo: This pull request adds Depthwise convolution to oneAPI backend.

Type of change

Adding Depthwise convolution in a parallel implementation. Supports depth multiplier. When a separable convolution layer is used, it is split to depthwise and pointwise layers.

Note: Please delete options that are not relevant.

• [ ] Bug fix (non-breaking change that fixes an issue)
• [ ] Documentation update
• [x] New feature (non-breaking change which adds functionality)
• [ ] A new research paper code implementation
• [ ] Other (Specify)

Tests

:memo: Pytests included and models compiled with depthwise and separable convolution.

Test Configuration:

model = Sequential()
model.add(DepthwiseConv2D(input_shape = (10,10,3), kernel_size=(3,3), padding="same", depth_multiplier=2, activation='relu', strides=(2,2)))
model.compile(loss='mse', optimizer=Adam())
model.summary()

config = hls4ml.utils.config_from_keras_model(model, granularity='name', default_precision='ac_fixed<16, 6>')
hls_model = hls4ml.converters.convert_from_keras_model(model=model, output_dir="model_dconv_out", backend="oneAPI", part="Agilex7", hls_config=config)
hls_model.compile()

Checklist

• [x] I have read the guidelines for contributing.
• [x] I have commented my code, particularly in hard-to-understand areas.
• [ ] I have made corresponding changes to the documentation.
• [x] My changes generate no new warnings.
• [x] I have installed and run pre-commit on the files I edited or added.
• [x] I have added tests that prove my fix is effective or that my feature works.

[jmitrevs]

Is it worth adding 1D versions, too?

[jmitrevs]

Why so many commits? Is it worth rebasing on a new branch from main?