import jax import jax.numpy as jnp import jax.tree_util as jtu import flax.linen as nn from layers import UNFLayer, UNFSequential, nf_relu class MLP(nn.Module): """Two layer MLP""" @nn.compact def __call__(self, x): x = nn.relu(nn.Dense(128)(x)) return nn.Dense(10)(x) # Label the dims of each parameter with a permutation index. We count the permutations # 0, 1, 2, ... but any distinct integers work. perm_spec = { "params": { "Dense_0": { "kernel": (0, 1), "bias": (1,) }, "Dense_1": { "kernel": (1, 2), "bias": (2,) } } } key = jax.random.PRNGKey(0) mlp_params = MLP().init(key, jnp.ones((64))) unf = UNFSequential([ UNFLayer(16, 1), nf_relu, UNFLayer(16, 16), ]) # add channel dim to input mlp_params = jtu.tree_map(lambda x: jnp.expand_dims(x, -1), mlp_params) print(jtu.tree_map(jnp.shape, mlp_params)) unf_params = unf.init(key, mlp_params, perm_spec) # apply unf to input unf_out = unf.apply(unf_params, mlp_params, perm_spec) print(jtu.tree_map(jnp.shape, unf_out))