problem reading in QKeras model

[jmitrevs]

I haven't gotten a chance to debug much, but reading in a QKeras model fails for me at:

hls4ml/hls4ml/converters/keras_to_hls.py

Line 331 in 881efea

act_layer['name'] = layer['name'] + '_' + layer['activation']

when the layer name is "q_conv1d_18" and the activation is {'class_name': 'quantized_relu', 'config': {'bits': 8, 'integer': 0, 'use_sigmoid': 0, 'negative_slope': 0.0, 'use_stochastic_rounding': False, 'relu_upper_bound': None, 'qnoise_factor': 1.0}}, and you can't concatenate a string and dictionary. Any recommendations on how to fix this problem? Or does one not use keras_to_hls for qkeras models (called from convert_from_keras_model)?

[jmitrevs]

The issue is that hls4ml doesn't seem to handle the case when the QKeras model doesn't explicitly use QActivation but instead passes, for example, activation=quantized_relu(..) to, for example, QConv1D. I will try to add support for this.

[thesps]

Hey, that's actually recently been addressed in #553 (not merged at time of writing)

[thesps]

That PR is merged now. Could you try again on master branch?

[jmitrevs]

The problem is fixed.