Fix config structure name in pragma for SeparableConv1D (#884)

* Raise exception if Vivado command fail

* Duplicate sepconv2d test for sepconv1d

* Test that csynth is working for sepconv1d

* Define multiplier_limit in nnet::conv1d_config (for sepconv1d)

* Revert build test

---------

Co-authored-by: Quentin Berthet <[email protected]>
Co-authored-by: Vladimir Loncar <[email protected]>

Author: 3 people

hls4ml/backends/vivado/passes/convolution_templates.py

@@ -41,6 +41,8 @@
     static const unsigned out_width = {out_width};
     static const unsigned reuse_factor = {reuse};
     static const unsigned n_zeros = {nzeros};
+    static const unsigned multiplier_limit =
+        DIV_ROUNDUP(kernel_size * n_chan * n_filt, reuse_factor) - n_zeros / reuse_factor;
     static const bool store_weights_in_bram = false;
     static const unsigned strategy = nnet::{strategy};
     static const nnet::conv_implementation implementation = nnet::conv_implementation::{implementation};

test/pytest/test_sepconv1d.py

@@ -0,0 +1,63 @@
+from pathlib import Path
+
+import numpy as np
+import pytest
+import tensorflow as tf
+from tensorflow.keras.layers import SeparableConv1D
+
+import hls4ml
+
+test_root_path = Path(__file__).parent
+
+keras_conv1d = [SeparableConv1D]
+padds_options = ['same', 'valid']
+chans_options = ['channels_last']
+io_type_options = ['io_stream']
+strides_options = [(1), (2)]
+kernel_options = [(1), (3)]
+bias_options = [False]
+
+
+@pytest.mark.parametrize('conv1d', keras_conv1d)
+@pytest.mark.parametrize('chans', chans_options)
+@pytest.mark.parametrize('padds', padds_options)
+@pytest.mark.parametrize('strides', strides_options)
+@pytest.mark.parametrize('kernels', kernel_options)
+@pytest.mark.parametrize('bias', bias_options)
+@pytest.mark.parametrize('io_type', io_type_options)
+@pytest.mark.parametrize('backend', ['Vivado', 'Vitis'])
+def test_sepconv1d(conv1d, chans, padds, strides, kernels, bias, io_type, backend):
+    model = tf.keras.models.Sequential()
+    input_shape = (28, 3)
+    model.add(
+        conv1d(
+            filters=32,
+            kernel_size=kernels,
+            strides=strides,
+            padding=padds,
+            input_shape=input_shape,
+            kernel_initializer='normal',
+            use_bias=bias,
+            data_format=chans,
+        )
+    )
+
+    model.compile(optimizer='adam', loss='mse')
+    X_input = np.random.rand(100, *input_shape)
+    keras_prediction = model.predict(X_input)
+    config = hls4ml.utils.config_from_keras_model(model, default_precision='ap_fixed<32,16>')
+    stride_cfg = str(strides).replace(', ', '_').replace('(', '').replace(')', '')
+    kernel_cfg = str(kernels).replace(', ', '_').replace('(', '').replace(')', '')
+    output_dir = str(
+        test_root_path
+        / 'hls4mlprj_{}_{}_strides_{}_kernels_{}_{}_padding_{}_{}'.format(
+            conv1d.__name__.lower(), chans, stride_cfg, kernel_cfg, padds, backend, io_type
+        )
+    )
+    hls_model = hls4ml.converters.convert_from_keras_model(
+        model, hls_config=config, output_dir=output_dir, io_type=io_type, backend=backend
+    )
+    hls_model.compile()
+    hls_prediction = hls_model.predict(X_input).reshape(keras_prediction.shape)
+
+    np.testing.assert_allclose(hls_prediction, keras_prediction, rtol=0, atol=0.001)