squeeze introduces transpose, which is not implemented

[sei-jgwohlbier]

I'm trying to use a squeeze in my PyTorch model. The line in the forward function is simply

z = torch.squeeze(z_latent)

This introduces an object <hls4ml.backends.fpga.fpga_backend.VivadoTranspose object at 0x7f7a4cd3e020> which leads to a compilation error

firmware/myproject.cpp: In function ‘void myproject(hls::stream<nnet::array<ap_fixed<16, 6>, 2> >&, hls::stream<nnet::array<ap_fixed<16, 6>, 10> >&)’:
firmware/myproject.cpp:297:80: error: no matching function for call to ‘transpose_3d<layer41_t, layer44_t, config44>(hls::stream<nnet::array<ap_fixed<16, 6>, 64> >&, hls::stream<nnet::array<ap_fixed<16, 6>, 1> >&)’
 ranspose_3d<layer41_t, layer44_t, config44>(layer41_out, layer44_out); // transpose_input_for_squeeze
                                                                     ^
In file included from firmware/parameters.h:12:0,
                 from firmware/myproject.cpp:4:
firmware/nnet_utils/nnet_array.h:27:6: note: candidate: template<class data_T, class res_T, class CONFIG_T> void nnet::transpose_3d(data_T*, res_T*)
 void transpose_3d(data_T data[CONFIG_T::depth * CONFIG_T::height * CONFIG_T::width],
      ^~~~~~~~~~~~
firmware/nnet_utils/nnet_array.h:27:6: note:   template argument deduction/substitution failed:
firmware/myproject.cpp:297:80: note:   cannot convert ‘layer41_out’ (type ‘hls::stream<nnet::array<ap_fixed<16, 6>, 64> >’) to type ‘nnet::array<ap_fixed<16, 6>, 64>*’
 ranspose_3d<layer41_t, layer44_t, config44>(layer41_out, layer44_out); // transpose_input_for_squeeze

Are there any suggestions for getting past this?