How do I obtain the parameters required by RealMappedFourierEntropyModelBase?

[litianyi926]

The RealMappedFourierEntropyModelBase class in the code requires the following parameters:

coef = jax.lax.complex(self.real, self.imag)
scale = nn.softplus(self.scale)
offset = self.offset

How should I determine these parameters? From codex-main/codex/ems/fourier_test.py and equinox.py, it can be seen that in your test methods, these parameters are obtained by passing rng, num_pdfs, num_freqs: int = 10, init_scale: float = 1e-3:

em = equinox.RealMappedFourierEntropyModel(
jax.random.PRNGKey(0), num_freqs=num_freq, num_pdfs=num_dims, init_scale=1
)

How are these parameters determined? Do they need to be trained, or should I set them to the same default values as you have?