Open
Description
Prerequisites
Please make sure to check off these prerequisites before submitting a bug report.
- Test that the bug appears on the current version of the master branch. Make sure to include the commit hash of the commit you checked out.
- Check that the issue hasn't already been reported, by checking the currently open issues.
- If there are steps to reproduce the problem, make sure to write them down below.
- If relevant, please include the hls4ml project files, which were created directly before and/or after the bug.
Quick summary
I'm trying to implement a CNN which implements a PReLU activation function. The parameters of the PReLU are not being propagated correctly to the HLS files.
Details
Steps to Reproduce
- This is the sample code to reproduce the issue:
import hls4ml
import keras
import numpy as np
keras_model = keras.Sequential([
keras.layers.Input(shape=(9, 1)),
keras.layers.Conv1D(filters=6, kernel_size=3, strides=1, padding='same'),
keras.layers.PReLU(shared_axes=[1]),
])
keras_model.summary()
for layer in keras_model.layers:
weights = layer.get_weights()
if weights:
random_weights = [np.random.randn(*w.shape) for w in weights]
layer.set_weights(random_weights)
for i, layer in enumerate(keras_model.layers):
weights = layer.get_weights()
if weights:
print(f"Layer {i} ({layer.name}) weights:")
for w in weights:
print(w)
hls4ml_config = hls4ml.utils.config_from_keras_model(keras_model, granularity='name', default_reuse_factor=1, default_precision='fixed<16,6>')
hls4ml_model = hls4ml.converters.convert_from_keras_model(keras_model, hls_config=hls4ml_config, output_dir='out/', backend='Vitis', part='xcku115-flva1517-2-e', clock_period=2.5, project_name='test_project')
hls4ml.utils.plot_model(hls4ml_model, show_shapes=True, show_precision=True, to_file='hls4ml_model.png')
hls4ml_model.write()
hls4ml_model.compile()
hls4ml_model.build()- Run it using
python3 main.py - See the generated .h file for the weights (parameters) of the PReLU. It should look like this. The file is ./out/firmware/weights/a4.h.
//Numpy array shape [1, 6]
//Min -2.232540130615
//Max 1.219962954521
//Number of zeros 0
#ifndef A4_H_
#define A4_H_
#ifndef __SYNTHESIS__
p_re_lu_param_t a4[6];
#else
p_re_lu_param_t a4[6] = {, , , , , };
#endif
#endif- Also, the terminal log looks like this:
ERROR: [HLS 207-1186] expected expression (firmware/weights/a4.h:12:26)
ERROR: [HLS 207-1186] expected expression (firmware/weights/a4.h:12:28)
ERROR: [HLS 207-1186] expected expression (firmware/weights/a4.h:12:30)
ERROR: [HLS 207-1186] expected expression (firmware/weights/a4.h:12:32)
ERROR: [HLS 207-1186] expected expression (firmware/weights/a4.h:12:34)Expected behavior
It should propagate the parameters from the model to the header files for using them in HLS.
Actual behavior
It doesn't propagate the parameters.