Model with Concatenation Layer

[williamliao28]

Hello,

I have a model containing a concatenation layer. I found that the output quantization of the concatenation layer always has finn_datatype: FLOAT32. This is causing problem for the layer after concatenation because the input of that layer will have finn_datatype: FLOAT32 and will not be converted to HLS layers. Is there a way to make the output quantization to have finn_datatype: INT?

[MFJI]

Is it possible to use a QuantIdentity layer after the concat layer?

[auphelia]

This problem is solved by PR #857 and was discussed in more detail in Issue #853.