def MultiHeadedAttentionQKV(
feature_depth, num_heads=8, dropout=0.0, mode='train'):
"""Transformer-style multi-headed attention.
Accepts inputs of the form (q, k, v), mask.
Args:
feature_depth: int: depth of embedding
num_heads: int: number of attention heads
dropout: float: dropout rate
mode: str: 'train' or 'eval'
Returns:
Multi-headed self-attention result and the mask.
"""
return combinators.Serial(
combinators.Parallel(
core.Dense(feature_depth),
core.Dense(feature_depth),
core.Dense(feature_depth),
combinators.NoOp()
),
PureMultiHeadedAttention( # pylint: disable=no-value-for-parameter
feature_depth=feature_depth, num_heads=num_heads,
dropout=dropout, mode=mode),
combinators.Parallel(core.Dense(feature_depth), combinators.NoOp())
)
def CausalAttention(d_feature,
n_heads=1,
d_attention_key=None,
d_attention_value=None,
attention_type=DotProductCausalAttention,
share_kv=False,
mode='train'):
"""Transformer-style multi-headed causal attention.
Args:
d_feature: int: dimensionality of feature embedding
n_heads: int: number of attention heads
d_attention_key: int: depth of key vector for each attention head
(default is d_feature // n_heads)
d_attention_value: int: depth of value vector for each attention head
(default is d_feature // n_heads)
attention_type: subclass of BaseCausalAttention: attention class to use
share_kv: bool, whether to share keys and values
mode: str: 'train' or 'eval'
Returns:
Multi-headed self-attention result.
"""
if d_attention_key is None:
assert d_feature % n_heads == 0
d_attention_key = d_feature // n_heads
if d_attention_value is None:
assert d_feature % n_heads == 0
d_attention_value = d_feature // n_heads
if share_kv:
pre_attention = [
cb.Dup(),
cb.Parallel(
ComputeAttentionHeads(n_heads=n_heads, d_head=d_attention_key),
ComputeAttentionHeads(n_heads=n_heads,
d_head=d_attention_value),
),
cb.Dup(),
]
else:
pre_attention = [
cb.Dup(),
cb.Dup(),
cb.Parallel(
ComputeAttentionHeads(n_heads=n_heads, d_head=d_attention_key),
ComputeAttentionHeads(n_heads=n_heads, d_head=d_attention_key),
ComputeAttentionHeads(n_heads=n_heads,
d_head=d_attention_value),
),
]
return pre_attention + [
attention_type(mode=mode),
ComputeAttentionOutput(n_heads=n_heads, d_model=d_feature),
]
def MultiHeadedAttention(feature_depth,
num_heads=8,
dropout=0.0,
mode='train'):
"""Transformer-style multi-headed attention.
Accepts inputs of the form (x, mask) and constructs (q, k, v) from x.
Args:
feature_depth: int: depth of embedding
num_heads: int: number of attention heads
dropout: float: dropout rate
mode: str: 'train' or 'eval'
Returns:
Multi-headed self-attention layer.
"""
return combinators.Serial(
combinators.Parallel(
# q = k = v = first input
combinators.Branch(combinators.Copy(), combinators.Copy(),
combinators.Copy()),
combinators.Copy() # pass the mask
),
MultiHeadedAttentionQKV( # pylint: disable=no-value-for-parameter
feature_depth,
num_heads=num_heads,
dropout=dropout,
mode=mode),
)
def ChunkedCausalMultiHeadedAttention(feature_depth,
num_heads=8,
dropout=0.0,
chunk_selector=None,
mode='train'):
"""Transformer-style causal multi-headed attention operating on chunks.
Accepts inputs that are a list of chunks and applies causal attention.
Args:
feature_depth: int: depth of embedding
num_heads: int: number of attention heads
dropout: float: dropout rate
chunk_selector: a function from chunk number to list of chunks to attend.
mode: str: 'train' or 'eval'
Returns:
Multi-headed self-attention layer.
"""
prepare_attention_input = combinators.Serial(
combinators.Branch(
combinators.Branch( # q = k = v = first input
combinators.Copy(), combinators.Copy(), combinators.Copy()),
CausalMask(axis=-2), # pylint: disable=no-value-for-parameter
),
combinators.Parallel(
combinators.Parallel(
core.Dense(feature_depth),
core.Dense(feature_depth),
core.Dense(feature_depth),
), combinators.Copy()))
return combinators.Serial(
combinators.Map(prepare_attention_input),
ChunkedAttentionSelector(selector=chunk_selector), # pylint: disable=no-value-for-parameter
combinators.Map(
PureMultiHeadedAttention( # pylint: disable=no-value-for-parameter
feature_depth=feature_depth,
num_heads=num_heads,
dropout=dropout,
mode=mode),
check_shapes=False),
combinators.Map(combinators.Select(0),
check_shapes=False), # drop masks
combinators.Map(core.Dense(feature_depth)))
def CausalAttention(d_feature, n_heads=1, dropout=0.0, mode='train'):
"""Transformer-style multi-headed causal attention.
# TODO(jonni,lukaszkaiser): standardize and improve layer comments.
Accepts inputs of the form x and constructs (q, k, v) and causal mask from x.
Args:
d_feature: int: dimensionality of feature embedding
n_heads: int: number of attention heads
dropout: float: dropout rate
mode: str: 'train' or 'eval'
Returns:
Multi-headed self-attention result.
"""
return [
cb.Dup(),
cb.Parallel([], CausalMask(axis=-2)), # pylint: disable=no-value-for-parameter
Attention(d_feature, n_heads=n_heads, dropout=dropout, mode=mode),
cb.Parallel([], cb.Drop()), # x
]
def MaskedScalar(metric_layer, mask_id=None, has_weights=False):
"""Metric as scalar compatible with Trax masking."""
# Stack of (inputs, targets) --> (metric, weight-mask).
metric_and_mask = [
cb.Parallel(
[],
cb.Dup() # Duplicate targets
),
cb.Parallel(
metric_layer, # Metric: (inputs, targets) --> metric
WeightMask(mask_id=mask_id) # pylint: disable=no-value-for-parameter
)
]
if not has_weights:
# Take (metric, weight-mask) and return the weighted mean.
return cb.Serial([metric_and_mask, WeightedMean()]) # pylint: disable=no-value-for-parameter
return cb.Serial([
metric_and_mask,
cb.Parallel(
[],
cb.Multiply() # Multiply given weights by mask_id weights
),
WeightedMean() # pylint: disable=no-value-for-parameter
])
def GeneralGRUCell(candidate_transform,
memory_transform=combinators.Identity,
gate_nonlinearity=core.Sigmoid,
candidate_nonlinearity=core.Tanh,
dropout_rate_c=0.1,
sigmoid_bias=0.5):
r"""Parametrized Gated Recurrent Unit (GRU) cell construction.
GRU update equations:
$$ Update gate: u_t = \sigmoid(U' * s_{t-1} + B') $$
$$ Reset gate: r_t = \sigmoid(U'' * s_{t-1} + B'') $$
$$ Candidate memory: c_t = \tanh(U * (r_t \odot s_{t-1}) + B) $$
$$ New State: s_t = u_t \odot s_{t-1} + (1 - u_t) \odot c_t $$
See combinators.GateBranches for details on the gating function.
Args:
candidate_transform: Transform to apply inside the Candidate branch. Applied
before nonlinearities.
memory_transform: Optional transformation on the memory before gating.
gate_nonlinearity: Function to use as gate activation. Allows trying
alternatives to Sigmoid, such as HardSigmoid.
candidate_nonlinearity: Nonlinearity to apply after candidate branch. Allows
trying alternatives to traditional Tanh, such as HardTanh
dropout_rate_c: Amount of dropout on the transform (c) gate. Dropout works
best in a GRU when applied exclusively to this branch.
sigmoid_bias: Constant to add before sigmoid gates. Generally want to start
off with a positive bias.
Returns:
A model representing a GRU cell with specified transforms.
"""
return combinators.Serial(
combinators.Branch(num_branches=3),
combinators.Parallel(
# s_{t-1} branch - optionally transform
# Typically is an identity.
memory_transform(),
# u_t (Update gate) branch
combinators.Serial(
candidate_transform(),
# Want bias to start out positive before sigmoids.
core.AddConstant(constant=sigmoid_bias),
gate_nonlinearity()),
# c_t (Candidate) branch
combinators.Serial(
combinators.Branch(num_branches=2),
combinators.Parallel(
combinators.Identity(),
# r_t (Reset) Branch
combinators.Serial(
candidate_transform(),
# Want bias to start out positive before sigmoids.
core.AddConstant(constant=sigmoid_bias),
gate_nonlinearity())),
## Gate S{t-1} with sigmoid(candidate_transform(S{t-1}))
combinators.MultiplyBranches(),
# Final projection + tanh to get Ct
candidate_transform(),
candidate_nonlinearity()), # Candidate gate
# Only apply dropout on the C gate.
# Paper reports that 0.1 is a good default.
core.Dropout(rate=dropout_rate_c)),
# Gate memory and candidate
combinators.GateBranches())
def GeneralGRUCell(candidate_transform,
memory_transform_fn=None,
gate_nonlinearity=core.Sigmoid,
candidate_nonlinearity=core.Tanh,
dropout_rate_c=0.1,
sigmoid_bias=0.5):
r"""Parametrized Gated Recurrent Unit (GRU) cell construction.
GRU update equations:
$$ Update gate: u_t = \sigmoid(U' * s_{t-1} + B') $$
$$ Reset gate: r_t = \sigmoid(U'' * s_{t-1} + B'') $$
$$ Candidate memory: c_t = \tanh(U * (r_t \odot s_{t-1}) + B) $$
$$ New State: s_t = u_t \odot s_{t-1} + (1 - u_t) \odot c_t $$
See combinators.Gate for details on the gating function.
Args:
candidate_transform: Transform to apply inside the Candidate branch. Applied
before nonlinearities.
memory_transform_fn: Optional transformation on the memory before gating.
gate_nonlinearity: Function to use as gate activation. Allows trying
alternatives to Sigmoid, such as HardSigmoid.
candidate_nonlinearity: Nonlinearity to apply after candidate branch. Allows
trying alternatives to traditional Tanh, such as HardTanh
dropout_rate_c: Amount of dropout on the transform (c) gate. Dropout works
best in a GRU when applied exclusively to this branch.
sigmoid_bias: Constant to add before sigmoid gates. Generally want to start
off with a positive bias.
Returns:
A model representing a GRU cell with specified transforms.
"""
gate_block = [ # u_t
candidate_transform(),
core.AddConstant(constant=sigmoid_bias),
gate_nonlinearity(),
]
reset_block = [ # r_t
candidate_transform(),
core.AddConstant(
constant=sigmoid_bias), # Want bias to start positive.
gate_nonlinearity(),
]
candidate_block = [
cb.Dup(),
reset_block,
cb.Multiply(), # Gate S{t-1} with sigmoid(candidate_transform(S{t-1}))
candidate_transform(), # Final projection + tanh to get Ct
candidate_nonlinearity(), # Candidate gate
# Only apply dropout on the C gate. Paper reports 0.1 as a good default.
core.Dropout(rate=dropout_rate_c)
]
memory_transform = memory_transform_fn() if memory_transform_fn else []
return cb.Model(
cb.Dup(),
cb.Dup(),
cb.Parallel(memory_transform, gate_block, candidate_block),
cb.Gate(),
)
def test_parallel_no_ops(self):
layer = cb.Parallel([], None)
input_shape = ((3, 2), (4, 7))
expected_shape = ((3, 2), (4, 7))
output_shape = base.check_shape_agreement(layer, input_shape)
self.assertEqual(output_shape, expected_shape)
def test_parallel_div_div(self):
layer = cb.Parallel(core.Div(divisor=0.5), core.Div(divisor=3.0))
input_shape = ((3, 2), (4, 7))
expected_shape = ((3, 2), (4, 7))
output_shape = base.check_shape_agreement(layer, input_shape)
self.assertEqual(output_shape, expected_shape)
def test_parallel_dup_dup(self):
layer = cb.Parallel(cb.Dup(), cb.Dup())
input_shape = ((3, 2), (4, 7))
expected_shape = ((3, 2), (3, 2), (4, 7), (4, 7))
output_shape = base.check_shape_agreement(layer, input_shape)
self.assertEqual(output_shape, expected_shape)
def test_parallel(self):
input_shape = ((2, 3), (2, 3))
expected_shape = ((2, 3), (2, 3))
output_shape = base.check_shape_agreement(cb.Parallel([], []),
input_shape)
self.assertEqual(output_shape, expected_shape)
