Sioni Summers

Hi, Instead of: `config['LayerName']['fc1_input'][Precision']['result']='ap_int'` it sounds like you need to do: `config['LayerName']['fc1_input'][Precision']['result']='ap_uint'` Notice the extra '`u`' for 'unsigned' in `ap_uint`

The test failure is due to the new import from `profiling` that is introduced in utils. Specifically [this line](https://github.com/maksgraczyk/hls4ml/blob/auto-type-setup/hls4ml/model/profiling.py#L3) that imports matplotlib (but the pandas and seaborn import would fail...

Okay, I get it now. Yeah I agree neither using the default precision nor the input precision is necessarily best. I would say setting the accumulator precision to the same...

> I'm currently looking into making set_data_types_from_keras_model() set accum_t of a layer to the same type as the layer output one That sounds great 👍 > Is there motivation for...

Thanks for the quick turnaround. I gave `set_accum_from_keras_model` a quick try on the hls4ml tutorial part 4 QKeras model, and the previous accuracy loss is recovered 👍 and further, the...

Do you see this when trying to `hls_model.compile()` or `hls_model.build()`?

Could you try `hls_model.build()` ? I think it's because the `hls_math.h` header file is not copied [here](https://github.com/yiiyama/hls4ml/tree/topic-config-garnet/hls4ml/templates/vivado/ap_types). But it should work when you use `build`, i.e. compile it with Vivado....

``` +-----------------+---------+-------+--------+--------+-----+ | Name | BRAM_18K| DSP48E| FF | LUT | URAM| +-----------------+---------+-------+--------+--------+-----+ ... +-----------------+---------+-------+--------+--------+-----+ |Utilization (%) | 20| 22| 11| 125| 0| +-----------------+---------+-------+--------+--------+-----+ ``` I think you could try...

Hi, thanks for the query. Something that you can try quickly: `io_type='io_parallel'` instead of `io_type='io_stream'`. That may help to remove those FIFOs, and since the NN doesn't contain any Conv...

It looks like this does indeed touch the same issue that #509 addresses. There is still some more work needed before that PR gets merged, it doesn't fit perfectly right...