# Copyright 2024 The Flax Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import annotations
from collections.abc import Sequence
import jax
import jax.numpy as jnp
from jax import lax, random
from flax.nnx import rnglib
from flax.nnx.module import Module, first_from
from flax import nnx
[docs]class Dropout(Module):
"""Create a dropout layer.
To use dropout, call the :func:`train` method (or pass in
``deterministic=False`` in the constructor or during call time).
To disable dropout, call the :func:`eval` method (or pass in
``deterministic=True`` in the constructor or during call time).
Example usage::
>>> from flax import nnx
>>> import jax.numpy as jnp
>>> class MLP(nnx.Module):
... def __init__(self, rngs):
... self.linear = nnx.Linear(in_features=3, out_features=4, rngs=rngs)
... self.dropout = nnx.Dropout(0.5, rngs=rngs)
... def __call__(self, x):
... x = self.linear(x)
... x = self.dropout(x)
... return x
>>> model = MLP(rngs=nnx.Rngs(0))
>>> x = jnp.ones((1, 3))
>>> model.train() # use dropout
>>> model(x)
Array([[ 2.1067007, -2.5359864, -1.592019 , -2.5238838]], dtype=float32)
>>> model.eval() # don't use dropout
>>> model(x)
Array([[ 1.0533503, -1.2679932, -0.7960095, -1.2619419]], dtype=float32)
Args:
rate: the dropout probability. (_not_ the keep rate!)
broadcast_dims: dimensions that will share the same dropout mask
deterministic: if false the inputs are scaled by ``1 / (1 - rate)`` and
masked, whereas if true, no mask is applied and the inputs are returned
as is.
rng_collection: the rng collection name to use when requesting an rng key.
rngs: rng key.
"""
def __init__(
self,
rate: float,
*,
broadcast_dims: Sequence[int] = (),
deterministic: bool = False,
rng_collection: str = 'dropout',
rngs: rnglib.Rngs | rnglib.RngStream | None = None,
):
self.rate = rate
self.broadcast_dims = broadcast_dims
self.deterministic = deterministic
self.rng_collection = rng_collection
if isinstance(rngs, rnglib.Rngs):
self.rngs = rngs[self.rng_collection].fork()
elif isinstance(rngs, rnglib.RngStream):
self.rngs = rngs.fork()
elif rngs is None:
self.rngs = nnx.data(None)
else:
raise TypeError(
f'rngs must be a Rngs, RngStream or None, but got {type(rngs)}.'
)
def __call__(
self,
inputs,
*,
deterministic: bool | None = None,
rngs: rnglib.Rngs | rnglib.RngStream | jax.Array | None = None,
) -> jax.Array:
"""Applies a random dropout mask to the input.
Args:
inputs: the inputs that should be randomly masked.
deterministic: if false the inputs are scaled by ``1 / (1 - rate)`` and
masked, whereas if true, no mask is applied and the inputs are returned
as is. The ``deterministic`` flag passed into the call method will take
precedence over the ``deterministic`` flag passed into the constructor.
rngs: an optional key, RngStream, or Rngs object used to generate the dropout mask.
If given it will take precedence over the rngs passed into the constructor.
Returns:
The masked inputs reweighted to preserve mean.
"""
deterministic = first_from(
deterministic,
self.deterministic,
error_msg="""No `deterministic` argument was provided to Dropout
as either a __call__ argument or class attribute""",
)
if (self.rate == 0.0) or deterministic:
return inputs
# Prevent gradient NaNs in 1.0 edge-case.
if self.rate == 1.0:
return jnp.zeros_like(inputs)
rngs = first_from( # type: ignore[assignment]
rngs,
self.rngs,
error_msg="""`deterministic` is False, but no `rngs` argument was provided to Dropout
as either a __call__ argument or class attribute.""",
)
if isinstance(rngs, rnglib.Rngs):
key = rngs[self.rng_collection]()
elif isinstance(rngs, rnglib.RngStream):
key = rngs()
elif isinstance(rngs, jax.Array):
key = rngs
else:
raise TypeError(
f'rngs must be a Rngs, RngStream or jax.Array, but got {type(rngs)}.'
)
keep_prob = 1.0 - self.rate
broadcast_shape = list(inputs.shape)
for dim in self.broadcast_dims:
broadcast_shape[dim] = 1
mask = random.bernoulli(key, p=keep_prob, shape=broadcast_shape)
mask = jnp.broadcast_to(mask, inputs.shape)
return lax.select(mask, inputs / keep_prob, jnp.zeros_like(inputs))
