Vladimir

Thanks, looks good. Do all reports say the timing is met? (No scheduling warnings etc, the clock uncertainty is met etc)

We're using pyDigitalWaveTools' `VcdParser ` to parse the generated waveform. Does that still work with the new change? It should, given the change between 1.1 and 1.2 is small and...

> Why co-simulation not found? Because you didn't run co-sim. Check the paramters of the `hls_model.build()`. As for the second error, your model is simply too large.

Unfortunately, if you use channels_first, we need to do internal transpose and this was not implemented for streaming io in Catapult backend

It can still be channels first during training and conversion, just be sure to transpose your inputs before passing them to the model predict and later the firmware

Somewhere in the console output there will be a message from the compiler what went wrong, the OS error is just what happens after the compilation failed.

Vitis accelerator backend is under development, see #1376. The bitfile generation will be similar to the existing Vivado accelerator backend. See the tutorial, part [7a](https://github.com/fastmachinelearning/hls4ml-tutorial/blob/main/part7a_bitstream.ipynb) in particular. We'll update the...

This is now ready for review/merge. The latest feature is the support for `compile()` but this had to be implemented in a hacky way. The `ap_fixpt` types require the SmartHLS...