EVK-MIMXRT685 dsp_lenet_demo example input output format

cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

EVK-MIMXRT685 dsp_lenet_demo example input output format

Jump to solution
686 Views
tugrulyatagan
Contributor I

Hi,

I am trying to port (convert) some CMSIS-NN code to Hi-Fi DSP calls. However input-output format of Hi-Fi DSP calls are not clear. What is the data format (protobuf, endianness, order) of weights, bias, input parameters of Hi-Fi DSP functions (xa_nn_*) ?

I need to use following functions:

signed int xa_nn_conv2d_std_8x8(signed char *p_out,
                                signed char *p_inp,
                                signed char *p_kernel,
                                signed char *p_bias,
                                signed int input_height,
                                signed int input_width,
                                signed int input_channels,
                                signed int kernel_height,
                                signed int kernel_width,
                                signed int out_channels,
                                signed int x_stride,
                                signed int y_stride,
                                signed int x_padding,
                                signed int y_padding,
                                signed int out_height,
                                signed int out_width,
                                signed int bias_shift,
                                signed int acc_shift,
                                signed int out_data_format,
                                void *p_handle)
                                
signed int xa_nn_vec_relu_std_8_8(signed char *p_out, const signed char *p_vec, signed int vec_length)

signed int xa_nn_vec_relu_std_32_32(signed int *p_out, const signed int *p_vec, signed int vec_length)

signed int xa_nn_maxpool_8(signed char *p_out,
                           signed char *p_inp,
                           signed int input_height,
                           signed int input_width,
                           signed int input_channels,
                           signed int kernel_height,
                           signed int kernel_width,
                           signed int x_stride,
                           signed int y_stride,
                           signed int x_padding,
                           signed int y_padding,
                           signed int out_height,
                           signed int out_width,
                           signed int inp_data_format,
                           signed int out_data_format,
                           void *p_scratch)

signed int xa_nn_avgpool_8(signed char *p_out,
                           signed char *p_inp,
                           signed int input_height,
                           signed int input_width,
                           signed int input_channels,
                           signed int kernel_height,
                           signed int kernel_width,
                           signed int x_stride,
                           signed int y_stride,
                           signed int x_padding,
                           signed int y_padding,
                           signed int out_height,
                           signed int out_width,
                           signed int inp_data_format,
                           signed int out_data_format,
                           void *p_scratch)

signed int xa_nn_fully_connected_8x8_8(signed char *p_out,
                                       signed char *p_weight,
                                       signed char *p_inp,
                                       signed char *p_bias,
                                       signed int weight_depth,
                                       signed int out_depth,
                                       signed int acc_shift,
                                       signed int bias_shift)

Thanks,

Tugrul

Labels (1)
0 Kudos
1 Solution
671 Views
jeremyzhou
NXP Employee
NXP Employee

Hi,
Thank you for your interest in NXP Semiconductor products and for the opportunity to serve you.
The above function is from the Cadence HiFi 5 Neural Network Library, please learn the information of these functions via the below link.
https://github.com/foss-xtensa/nnlib-hifi5/tree/master/xa_nnlib/doc
Have a great day,
TIC

-------------------------------------------------------------------------------
Note:
- If this post answers your question, please click the "Mark Correct" button. Thank you!

 

- We are following threads for 7 weeks after the last post, later replies are ignored
Please open a new thread and refer to the closed one, if you have a related question at a later point in time.
-------------------------------------------------------------------------------

View solution in original post

0 Kudos
1 Reply
672 Views
jeremyzhou
NXP Employee
NXP Employee

Hi,
Thank you for your interest in NXP Semiconductor products and for the opportunity to serve you.
The above function is from the Cadence HiFi 5 Neural Network Library, please learn the information of these functions via the below link.
https://github.com/foss-xtensa/nnlib-hifi5/tree/master/xa_nnlib/doc
Have a great day,
TIC

-------------------------------------------------------------------------------
Note:
- If this post answers your question, please click the "Mark Correct" button. Thank you!

 

- We are following threads for 7 weeks after the last post, later replies are ignored
Please open a new thread and refer to the closed one, if you have a related question at a later point in time.
-------------------------------------------------------------------------------

0 Kudos