def __call__(self, shape: Shape, dtype: DType) -> np.ndarray:
if len(shape) < 2:
raise ValueError(
"Orthogonal initializer requires at least a 2D shape.")
n_rows = shape[self.axis]
n_cols = np.prod(shape) // n_rows
matrix_shape = (n_rows, n_cols) if n_rows > n_cols else (n_cols,
n_rows)
norm_dst = jax.random.normal(module.next_rng_key(), matrix_shape,
dtype)
q_mat, r_mat = jnp.linalg.qr(norm_dst)
# Enforce Q is uniformly distributed
q_mat *= jnp.sign(jnp.diag(r_mat))
if n_rows < n_cols:
q_mat = q_mat.T
q_mat = jnp.reshape(q_mat,
(n_rows, ) + tuple(np.delete(shape, self.axis)))
q_mat = jnp.moveaxis(q_mat, 0, self.axis)
return jax.lax.convert_element_type(self.scale, dtype) * q_mat
def call(
self,
x: np.ndarray,
training: tp.Optional[bool] = None,
rng: tp.Optional[np.ndarray] = None,
) -> jnp.ndarray:
"""
Arguments:
x: The value to be dropped out.
training: Whether training is currently happening.
rng: Optional RNGKey.
Returns:
x but dropped out and scaled by `1 / (1 - rate)`.
"""
if training is None:
training = module.is_training()
return hk.dropout(
rng=rng if rng is not None else module.next_rng_key(),
rate=self.rate if training else 0.0,
x=x,
)
def __call__(self, shape: Shape, dtype: DType) -> np.ndarray:
m = jax.lax.convert_element_type(self.mean, dtype)
s = jax.lax.convert_element_type(self.stddev, dtype)
unscaled = jax.random.truncated_normal(module.next_rng_key(), -2.0,
2.0, shape, dtype)
return s * unscaled + m
def __call__(self, shape: Shape, dtype: DType) -> np.ndarray:
m = jax.lax.convert_element_type(self.mean, dtype)
s = jax.lax.convert_element_type(self.stddev, dtype)
return m + s * jax.random.normal(module.next_rng_key(), shape, dtype)
def __call__(self, shape: Shape, dtype: DType) -> np.ndarray:
return jax.random.uniform(module.next_rng_key(), shape, dtype,
self.minval, self.maxval)