How the Kendryte Model Compiler work for our own models



  • Re: model compiler

    Can anyone help me with the compiling step? I followed the kendryte-model-compiler example but it always shows:

    Traceback (most recent call last):
    File "main.py", line 145, in <module>
    main()
    File "main.py", line 118, in main
    k210_layers = model_loader_module.load_model(dataset_val, rfb, args)
    File "E:\Python_Coding\kendryte-model-compiler-master\model_loader\pb_init_.py", line 65, in load_model
    converter.convert()
    File "E:\Python_Coding\kendryte-model-compiler-master\model_loader\pb\tensor_head_to_tensor_list.py", line 240, in convert
    while self.convert_step():
    File "E:\Python_Coding\kendryte-model-compiler-master\model_loader\pb\tensor_head_to_tensor_list.py", line 235, in convert_step
    raise ValueError('no converter for', self.output_tensor.op.type, 'name:', self.output_tensor.op.name)
    ValueError: ('no converter for', 'Softmax', 'name:', 'softmax')

    But there is actually 'softmax' as one of the activations in 'tensor_head_to_tensor_list.py'. I am really confused. Can anyone give some advice? Thanks a lot.


  • |  Mod



  • Hi,

    Thanks for your reply. I tried several models but it worked sometimes while sometimes not. I cannot figure out why. Do you have any more examples for those so that maybe I can understand the nncase better.

    Thank you.


  • Global Moderator |  Mod

    Hi,
    Please use nncase to instead of kmodel.
    https://github.com/kendryte/nncase
    Thanks.