# Copyright 2022 Xiaomi Corp. (authors: Daniel Povey) # # See ../../../../LICENSE for clarification regarding multiple 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 typing import Optional, Tuple import torch import torch.nn as nn from torch import Tensor def _activation_balancer_loss( mean_pos: Tensor, mean_neg: Tensor, min_positive: float, # e.g. 0.05 max_positive: float, # e.g. 0.95 max_factor: float, # e.g. 0.01 min_abs: float, # e.g. 0.2 max_abs: float, # e.g. 100.0 eps: float = 1.0e-10, ): """ Returns a loss-function for the ActivationBalancer module. This loss function is not exposed to the user but is used internally, and eventually its derivatives are scaled by some heuristic related to derivative magnitudes, and added to the backpropped deriv. Args: mean_pos: a Tensor of arbitrary dimension, probably something like (1, num_channels, 1, 1), containing the mean of only the positive parts of the input features, i.e. of x.relu(). mean_neg: a Tensor of arbitrary dimension, probably something like (1, num_channels, 1, 1), containing the mean of only the negative parts of the input features, i.e. of (-x).relu(). min_positive: the minimum allowed value of mean_pos / (mean_pos + mean_neg) before we start penalizing. max_positive: the maximum allowed value of mean_pos / (mean_pos + mean_neg) before we start penalizing. """ loss_parts = [] x_mean = mean_pos - mean_neg x_mean_abs = (mean_pos + mean_neg + eps).detach() x_rel_mean = x_mean / x_mean_abs if min_positive != 0.0: # e.g. x_mean_floor = -0.95 + 0.05 = -0.9 x_rel_mean_floor = -(1 - min_positive) + min_positive min_positive_loss = (x_rel_mean_floor - x_rel_mean).relu().sum() * ( 1.0 / (2 * min_positive) ) # this part of the loss would be 1.0 * num_channels if all these constraints were # 100% violated. loss_parts.append(min_positive_loss) if max_positive != 1.0: # e.g. x_mean_floor = -0.05 + 0.95 = 0.8 x_rel_mean_ceil = -(1.0 - max_positive) + max_positive max_positive_loss = (x_rel_mean - x_rel_mean_ceil).relu().sum() * ( 1.0 / (1 - x_rel_mean_ceil) ) # this part of the loss would be 1.0 * num_channels if all these constraints were # 100% violated. loss_parts.append(max_positive_loss) if min_abs != 0.0: min_abs_loss = (min_abs - x_mean_abs).relu().sum() / min_abs # this part of the loss would be 1.0 * num_channels if all these constraints were # 100% violated. loss_parts.append(min_abs_loss) if max_abs != 0.0: max_abs_loss = (x_mean_abs / max_abs).log().relu() # this part of the loss would be [something logarithmic] * num_channels if all these constraints were # 100% violated. loss_parts.append(max_abs_loss) # the min_positive and 1 - max_positive are "ballast" added to the denom = mean_pos + mean_neg + (min_positive + (1 - max_positive)) # num if min_positive != 0.0: pass class ActivationBalancerFunction(torch.autograd.Function): @staticmethod def forward( ctx, x: Tensor, channel_dim: int, min_positive: float, # e.g. 0.05 max_positive: float, # e.g. 0.95 max_factor: float, # e.g. 0.01 min_abs: float, # e.g. 0.2 max_abs: float, # e.g. 100.0 ) -> Tensor: if x.requires_grad: if channel_dim < 0: channel_dim += x.ndim sum_dims = [d for d in range(x.ndim) if d != channel_dim] xgt0 = x > 0 proportion_positive = torch.mean( xgt0.to(x.dtype), dim=sum_dims, keepdim=True ) factor1 = ( (min_positive - proportion_positive).relu() * (max_factor / min_positive) if min_positive != 0.0 else 0.0 ) factor2 = ( (proportion_positive - max_positive).relu() * (max_factor / (max_positive - 1.0)) if max_positive != 1.0 else 0.0 ) factor = factor1 + factor2 if isinstance(factor, float): factor = torch.zeros_like(proportion_positive) mean_abs = torch.mean(x.abs(), dim=sum_dims, keepdim=True) below_threshold = mean_abs < min_abs above_threshold = mean_abs > max_abs ctx.save_for_backward(factor, xgt0, below_threshold, above_threshold) ctx.max_factor = max_factor ctx.sum_dims = sum_dims return x @staticmethod def backward( ctx, x_grad: Tensor ) -> Tuple[Tensor, None, None, None, None, None, None]: factor, xgt0, below_threshold, above_threshold = ctx.saved_tensors dtype = x_grad.dtype scale_factor = ( (below_threshold.to(dtype) - above_threshold.to(dtype)) * (xgt0.to(dtype) - 0.5) * (ctx.max_factor * 2.0) ) neg_delta_grad = x_grad.abs() * (factor + scale_factor) return x_grad - neg_delta_grad, None, None, None, None, None, None class BasicNorm(torch.nn.Module): """ This is intended to be a simpler, and hopefully cheaper, replacement for LayerNorm. The observation this is based on, is that Transformer-type networks, especially with pre-norm, sometimes seem to set one of the feature dimensions to a large constant value (e.g. 50), which "defeats" the LayerNorm because the output magnitude is then not strongly dependent on the other (useful) features. Presumably the weight and bias of the LayerNorm are required to allow it to do this. So the idea is to introduce this large constant value as an explicit parameter, that takes the role of the "eps" in LayerNorm, so the network doesn't have to do this trick. We make the "eps" learnable. Args: num_channels: the number of channels, e.g. 512. channel_dim: the axis/dimension corresponding to the channel, interprted as an offset from the input's ndim if negative. shis is NOT the num_channels; it should typically be one of {-2, -1, 0, 1, 2, 3}. eps: the initial "epsilon" that we add as ballast in: scale = ((input_vec**2).mean() + epsilon)**-0.5 Note: our epsilon is actually large, but we keep the name to indicate the connection with conventional LayerNorm. learn_eps: if true, we learn epsilon; if false, we keep it at the initial value. """ def __init__( self, num_channels: int, channel_dim: int = -1, # CAUTION: see documentation. eps: float = 0.25, learn_eps: bool = True, ) -> None: super(BasicNorm, self).__init__() self.num_channels = num_channels self.channel_dim = channel_dim if learn_eps: self.eps = nn.Parameter(torch.tensor(eps).log().detach()) else: self.register_buffer("eps", torch.tensor(eps).log().detach()) def forward(self, x: Tensor) -> Tensor: assert x.shape[self.channel_dim] == self.num_channels scales = ( torch.mean(x**2, dim=self.channel_dim, keepdim=True) + self.eps.exp() ) ** -0.5 return x * scales class ScaledLinear(nn.Linear): """ A modified version of nn.Linear where the parameters are scaled before use, via: weight = self.weight * self.weight_scale.exp() bias = self.bias * self.bias_scale.exp() Args: Accepts the standard args and kwargs that nn.Linear accepts e.g. in_features, out_features, bias=False. initial_scale: you can override this if you want to increase or decrease the initial magnitude of the module's output (affects the initialization of weight_scale and bias_scale). Another option, if you want to do something like this, is to re-initialize the parameters. Note: it uses the default initialization for the weight and bias, inherited from nn.Linear. For modules with small fan-in, this may be larger than optimal. """ def __init__(self, *args, initial_scale: float = 1.0, **kwargs): super(ScaledLinear, self).__init__(*args, **kwargs) initial_scale = torch.tensor(initial_scale).log() self.weight_scale = nn.Parameter(initial_scale.clone().detach()) if self.bias is not None: self.bias_scale = nn.Parameter(initial_scale.clone().detach()) else: self.register_parameter("bias_scale", None) self._reset_parameters() # Overrides the reset_parameters in nn.Linear def _reset_parameters(self): std = 0.01 a = (3**0.5) * std nn.init.uniform_(self.weight, -a, a) if self.bias is not None: nn.init.constant_(self.bias, 0.0) fan_in = self.weight.shape[1] * self.weight[0][0].numel() scale = fan_in**-0.5 # 1/sqrt(fan_in) with torch.no_grad(): self.weight_scale += torch.tensor(scale / std).log() if self.bias is not None: self.bias_scale += torch.tensor(scale / std).log() def get_weight(self): return self.weight * self.weight_scale.exp() def get_bias(self): return None if self.bias is None else self.bias * self.bias_scale.exp() def forward(self, input: Tensor) -> Tensor: return torch.nn.functional.linear(input, self.get_weight(), self.get_bias()) class ScaledConv1d(nn.Conv1d): def __init__(self, *args, initial_scale=1.0, **kwargs): super(ScaledConv1d, self).__init__(*args, **kwargs) initial_scale = torch.tensor(initial_scale).log() self.weight_scale = nn.Parameter(initial_scale.clone().detach()) if self.bias is not None: self.bias_scale = nn.Parameter(initial_scale.clone().detach()) else: self.register_parameter("bias_scale", None) self._reset_parameters() # Overrides the reset_parameters in base class def _reset_parameters(self): std = 0.01 a = (3**0.5) * std nn.init.uniform_(self.weight, -a, a) if self.bias is not None: nn.init.constant_(self.bias, 0.0) fan_in = self.weight.shape[1] * self.weight[0][0].numel() scale = fan_in**-0.5 # 1/sqrt(fan_in) with torch.no_grad(): self.weight_scale += torch.tensor(scale / std).log() if self.bias is not None: self.bias_scale += torch.tensor(scale / std).log() def get_weight(self): return self.weight * self.weight_scale.exp() def get_bias(self): return None if self.bias is None else self.bias * self.bias_scale.exp() def forward(self, input: Tensor) -> Tensor: F = torch.nn.functional if self.padding_mode != "zeros": return F.conv1d( F.pad( input, self._reversed_padding_repeated_twice, mode=self.padding_mode, ), self.get_weight(), self.get_bias(), self.stride, _single(0), # noqa: F821 self.dilation, self.groups, ) return F.conv1d( input, self.get_weight(), self.get_bias(), self.stride, self.padding, self.dilation, self.groups, ) class ScaledConv2d(nn.Conv2d): def __init__(self, *args, initial_scale=1.0, **kwargs): super(ScaledConv2d, self).__init__(*args, **kwargs) initial_scale = torch.tensor(initial_scale).log() self.weight_scale = nn.Parameter(initial_scale.clone().detach()) if self.bias is not None: self.bias_scale = nn.Parameter(initial_scale.clone().detach()) else: self.register_parameter("bias_scale", None) self._reset_parameters() # Overrides the reset_parameters in base class def _reset_parameters(self): std = 0.01 a = (3**0.5) * std nn.init.uniform_(self.weight, -a, a) if self.bias is not None: nn.init.constant_(self.bias, 0.0) fan_in = self.weight.shape[1] * self.weight[0][0].numel() scale = fan_in**-0.5 # 1/sqrt(fan_in) with torch.no_grad(): self.weight_scale += torch.tensor(scale / std).log() if self.bias is not None: self.bias_scale += torch.tensor(scale / std).log() def get_weight(self): return self.weight * self.weight_scale.exp() def get_bias(self): return None if self.bias is None else self.bias * self.bias_scale.exp() def _conv_forward(self, input, weight): F = torch.nn.functional if self.padding_mode != "zeros": return F.conv2d( F.pad( input, self._reversed_padding_repeated_twice, mode=self.padding_mode, ), weight, self.get_bias(), self.stride, _pair(0), # noqa: F821 self.dilation, self.groups, ) return F.conv2d( input, weight, self.get_bias(), self.stride, self.padding, self.dilation, self.groups, ) def forward(self, input: Tensor) -> Tensor: return self._conv_forward(input, self.get_weight()) class ActivationBalancer(torch.nn.Module): """ Modifies the backpropped derivatives of a function to try to encourage, for each channel, that it is positive at least a proportion `threshold` of the time. It does this by multiplying negative derivative values by up to (1+max_factor), and positive derivative values by up to (1-max_factor), interpolated from 1 at the threshold to those extremal values when none of the inputs are positive. Args: channel_dim: the dimension/axis corresponding to the channel, e.g. -1, 0, 1, 2; will be interpreted as an offset from x.ndim if negative. min_positive: the minimum, per channel, of the proportion of the time that (x > 0), below which we start to modify the derivatives. max_positive: the maximum, per channel, of the proportion of the time that (x > 0), below which we start to modify the derivatives. max_factor: the maximum factor by which we modify the derivatives for either the sign constraint or the magnitude constraint; e.g. with max_factor=0.02, the the derivatives would be multiplied by values in the range [0.98..1.02]. min_abs: the minimum average-absolute-value per channel, which we allow, before we start to modify the derivatives to prevent this. max_abs: the maximum average-absolute-value per channel, which we allow, before we start to modify the derivatives to prevent this. """ def __init__( self, channel_dim: int, min_positive: float = 0.05, max_positive: float = 0.95, max_factor: float = 0.01, min_abs: float = 0.2, max_abs: float = 100.0, ): super(ActivationBalancer, self).__init__() self.channel_dim = channel_dim self.min_positive = min_positive self.max_positive = max_positive self.max_factor = max_factor self.min_abs = min_abs self.max_abs = max_abs def forward(self, x: Tensor) -> Tensor: return ActivationBalancerFunction.apply( x, self.channel_dim, self.min_positive, self.max_positive, self.max_factor, self.min_abs, self.max_abs, ) class DoubleSwishFunction(torch.autograd.Function): """ double_swish(x) = x * torch.sigmoid(x-1) This is a definition, originally motivated by its close numerical similarity to swish(swish(x), where swish(x) = x * sigmoid(x). Memory-efficient derivative computation: double_swish(x) = x * s, where s(x) = torch.sigmoid(x-1) double_swish'(x) = d/dx double_swish(x) = x * s'(x) + x' * s(x) = x * s'(x) + s(x). Now, s'(x) = s(x) * (1-s(x)). double_swish'(x) = x * s'(x) + s(x). = x * s(x) * (1-s(x)) + s(x). = double_swish(x) * (1-s(x)) + s(x) ... so we just need to remember s(x) but not x itself. """ @staticmethod def forward(ctx, x: Tensor) -> Tensor: x = x.detach() s = torch.sigmoid(x - 1.0) y = x * s ctx.save_for_backward(s, y) return y @staticmethod def backward(ctx, y_grad: Tensor) -> Tensor: s, y = ctx.saved_tensors return (y * (1 - s) + s) * y_grad class DoubleSwish(torch.nn.Module): def forward(self, x: Tensor) -> Tensor: """Return double-swish activation function which is an approximation to Swish(Swish(x)), that we approximate closely with x * sigmoid(x-1). """ return DoubleSwishFunction.apply(x) class ScaledEmbedding(nn.Module): r"""A simple lookup table that stores embeddings of a fixed dictionary and size. This module is often used to store word embeddings and retrieve them using indices. The input to the module is a list of indices, and the output is the corresponding word embeddings. Args: num_embeddings (int): size of the dictionary of embeddings embedding_dim (int): the size of each embedding vector padding_idx (int, optional): If given, pads the output with the embedding vector at :attr:`padding_idx` (initialized to zeros) whenever it encounters the index. max_norm (float, optional): If given, each embedding vector with norm larger than :attr:`max_norm` is renormalized to have norm :attr:`max_norm`. norm_type (float, optional): The p of the p-norm to compute for the :attr:`max_norm` option. Default ``2``. scale_grad_by_freq (boolean, optional): If given, this will scale gradients by the inverse of frequency of the words in the mini-batch. Default ``False``. sparse (bool, optional): If ``True``, gradient w.r.t. :attr:`weight` matrix will be a sparse tensor. See Notes for more details regarding sparse gradients. Attributes: weight (Tensor): the learnable weights of the module of shape (num_embeddings, embedding_dim) initialized from :math:`\mathcal{N}(0, 1)` Shape: - Input: :math:`(*)`, LongTensor of arbitrary shape containing the indices to extract - Output: :math:`(*, H)`, where `*` is the input shape and :math:`H=\text{embedding\_dim}` .. note:: Keep in mind that only a limited number of optimizers support sparse gradients: currently it's :class:`optim.SGD` (`CUDA` and `CPU`), :class:`optim.SparseAdam` (`CUDA` and `CPU`) and :class:`optim.Adagrad` (`CPU`) .. note:: With :attr:`padding_idx` set, the embedding vector at :attr:`padding_idx` is initialized to all zeros. However, note that this vector can be modified afterwards, e.g., using a customized initialization method, and thus changing the vector used to pad the output. The gradient for this vector from :class:`~torch.nn.Embedding` is always zero. Examples:: >>> # an Embedding module containing 10 tensors of size 3 >>> embedding = nn.Embedding(10, 3) >>> # a batch of 2 samples of 4 indices each >>> input = torch.LongTensor([[1,2,4,5],[4,3,2,9]]) >>> embedding(input) tensor([[[-0.0251, -1.6902, 0.7172], [-0.6431, 0.0748, 0.6969], [ 1.4970, 1.3448, -0.9685], [-0.3677, -2.7265, -0.1685]], [[ 1.4970, 1.3448, -0.9685], [ 0.4362, -0.4004, 0.9400], [-0.6431, 0.0748, 0.6969], [ 0.9124, -2.3616, 1.1151]]]) >>> # example with padding_idx >>> embedding = nn.Embedding(10, 3, padding_idx=0) >>> input = torch.LongTensor([[0,2,0,5]]) >>> embedding(input) tensor([[[ 0.0000, 0.0000, 0.0000], [ 0.1535, -2.0309, 0.9315], [ 0.0000, 0.0000, 0.0000], [-0.1655, 0.9897, 0.0635]]]) """ __constants__ = [ "num_embeddings", "embedding_dim", "padding_idx", "scale_grad_by_freq", "sparse", ] num_embeddings: int embedding_dim: int padding_idx: int scale_grad_by_freq: bool weight: Tensor sparse: bool def __init__( self, num_embeddings: int, embedding_dim: int, padding_idx: Optional[int] = None, scale_grad_by_freq: bool = False, sparse: bool = False, ) -> None: super(ScaledEmbedding, self).__init__() self.num_embeddings = num_embeddings self.embedding_dim = embedding_dim if padding_idx is not None: if padding_idx > 0: assert ( padding_idx < self.num_embeddings ), "Padding_idx must be within num_embeddings" elif padding_idx < 0: assert ( padding_idx >= -self.num_embeddings ), "Padding_idx must be within num_embeddings" padding_idx = self.num_embeddings + padding_idx self.padding_idx = padding_idx self.scale_grad_by_freq = scale_grad_by_freq self.scale = nn.Parameter(torch.zeros(())) # see reset_parameters() self.sparse = sparse self.weight = nn.Parameter(torch.Tensor(num_embeddings, embedding_dim)) self.reset_parameters() def reset_parameters(self) -> None: std = 0.01 nn.init.normal_(self.weight, std=std) nn.init.constant_(self.scale, torch.tensor(1.0 / std).log()) if self.padding_idx is not None: with torch.no_grad(): self.weight[self.padding_idx].fill_(0) def forward(self, input: Tensor) -> Tensor: F = torch.nn.functional scale = self.scale.exp() if input.numel() < self.num_embeddings: return ( F.embedding( input, self.weight, self.padding_idx, None, 2.0, # None, 2.0 relate to normalization self.scale_grad_by_freq, self.sparse, ) * scale ) else: return F.embedding( input, self.weight * scale, self.padding_idx, None, 2.0, # None, 2.0 relates to normalization self.scale_grad_by_freq, self.sparse, ) def extra_repr(self) -> str: s = "{num_embeddings}, {embedding_dim}, scale={scale}" if self.padding_idx is not None: s += ", padding_idx={padding_idx}" if self.scale_grad_by_freq is not False: s += ", scale_grad_by_freq={scale_grad_by_freq}" if self.sparse is not False: s += ", sparse=True" return s.format(**self.__dict__) def _test_activation_balancer_sign(): channel_dim = 0 probs = torch.arange(0, 1, 0.01) N = 1000 x = 1.0 * (torch.rand(probs.numel(), N) < probs.unsqueeze(-1)) x = x.detach() x.requires_grad = True m = ActivationBalancer( channel_dim=0, min_positive=0.05, max_positive=0.95, max_factor=0.2, min_abs=0.0, ) y_grad = torch.sign(torch.randn(probs.numel(), N)) y = m(x) y.backward(gradient=y_grad) print("_test_activation_balancer_sign: x = ", x) print("_test_activation_balancer_sign: y grad = ", y_grad) print("_test_activation_balancer_sign: x grad = ", x.grad) def _test_activation_balancer_magnitude(): channel_dim = 0 magnitudes = torch.arange(0, 1, 0.01) N = 1000 x = torch.sign(torch.randn(magnitudes.numel(), N)) * magnitudes.unsqueeze(-1) x = x.detach() x.requires_grad = True m = ActivationBalancer( channel_dim=0, min_positive=0.0, max_positive=1.0, max_factor=0.2, min_abs=0.2, max_abs=0.8, ) y_grad = torch.sign(torch.randn(magnitudes.numel(), N)) y = m(x) y.backward(gradient=y_grad) print("_test_activation_balancer_magnitude: x = ", x) print("_test_activation_balancer_magnitude: y grad = ", y_grad) print("_test_activation_balancer_magnitude: x grad = ", x.grad) def _test_basic_norm(): num_channels = 128 m = BasicNorm(num_channels=num_channels, channel_dim=1) x = torch.randn(500, num_channels) y = m(x) assert y.shape == x.shape x_rms = (x**2).mean().sqrt() y_rms = (y**2).mean().sqrt() print("x rms = ", x_rms) print("y rms = ", y_rms) assert y_rms < x_rms assert y_rms > 0.5 * x_rms def _test_double_swish_deriv(): x = torch.randn(10, 12, dtype=torch.double) * 0.5 x.requires_grad = True m = DoubleSwish() torch.autograd.gradcheck(m, x) if __name__ == "__main__": _test_activation_balancer_sign() _test_activation_balancer_magnitude() _test_basic_norm() _test_double_swish_deriv()