Inconsistent number of parameters loading encoding with tiny-cuda-nn pytorch bindings and instant-ngp

[fedeceola]

Hi, I am creating an encoding with the pytorch bindings from the following .json config file.

smaller_no_base.json

If i create the encoding with this line of code
tiny_cuda_nn_encoding = tcnn.Encoding(n_input_dims=3, encoding_config=config["encoding"]) I have a number of parameters which is different from the one that I have when loading this configuration file in instant-ngp. With other configuration files such as the original base.json or bigger models the problem doesn't occur.

Is it a bug or am I missing something?

Thanks in advance for the support.