repo
stringlengths
1
99
file
stringlengths
13
215
code
stringlengths
12
59.2M
file_length
int64
12
59.2M
avg_line_length
float64
3.82
1.48M
max_line_length
int64
12
2.51M
extension_type
stringclasses
1 value
rankpredictor
rankpredictor-master/sub/gluonts/distribution/box_cox_tranform.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import Dict, List, Tuple # Third-party imports import mxnet as mx # First-party imports from gluonts.core.component import validated from gluonts.distribution.bijection import Bijection, InverseBijection from gluonts.distribution.bijection_output import BijectionOutput from gluonts.model.common import Tensor # Relative imports from .distribution import getF, softplus class BoxCoxTranform(Bijection): r""" Implements Box-Cox transformation of a uni-variate random variable. The Box-Cox transformation of an observation :math:`z` is given by .. math:: BoxCox(z; \lambda_1, \lambda_2) = \begin{cases} ((z + \lambda_2)^{\lambda_1} - 1) / \lambda_1, \quad & \text{if } \lambda_1 \neq 0, \\ \log (z + \lambda_2), \quad & \text{otherwise.} \end{cases} Here, :math:`\lambda_1` and :math:`\lambda_2` are learnable parameters. Note that the domain of the transformation is not restricted. For numerical stability, instead of checking :math:`\lambda_1` is exactly zero, we use the condition .. math:: |\lambda_1| < tol\_lambda\_1 for a pre-specified tolerance `tol_lambda_1`. Inverse of the Box-Cox Transform is given by .. math:: BoxCox^{-1}(y; \lambda_1, \lambda_2) = \begin{cases} (y \lambda_1 + 1)^{(1/\lambda_1)} - \lambda_2, \quad & \text{if } \lambda_1 \neq 0, \\ \exp (y) - \lambda_2, \quad & \text{otherwise.} \end{cases} **Notes on numerical stability:** 1. For the forward transformation, :math:`\lambda_2` must always be chosen such that .. math:: z + \lambda_2 > 0. To achieve this one needs to know a priori the lower bound on the observations. This is set in `BoxCoxTransformOutput`, since :math:`\lambda_2` is learnable. 2. Similarly for the inverse transformation to work reliably, a sufficient condition is .. math:: y \lambda_1 + 1 \geq 0, where :math:`y` is the input to the inverse transformation. This cannot always be guaranteed especially when :math:`y` is a sample from a transformed distribution. Hence we always truncate :math:`y \lambda_1 + 1` at zero. An example showing why this could happen in our case: consider transforming observations from the unit interval (0, 1) with parameters .. math:: \begin{align} \lambda_1 = &\ 1.1, \\ \lambda_2 = &\ 0. \end{align} Then the range of the transformation is (-0.9090, 0.0). If Gaussian is fit to the transformed observations and a sample is drawn from it, then it is likely that the sample is outside this range, e.g., when the mean is close to -0.9. The subsequent inverse transformation of the sample is not a real number anymore. >>> y = -0.91 >>> lambda_1 = 1.1 >>> lambda_2 = 0.0 >>> (y * lambda_1 + 1) ** (1 / lambda_1) + lambda_2 (-0.0017979146510711471+0.0005279153735965289j) Parameters ---------- lambda_1 lambda_2 tol_lambda_1 For numerical stability, treat `lambda_1` as zero if it is less than `tol_lambda_1` F """ arg_names = ["box_cox.lambda_1", "box_cox.lambda_2"] @validated() def __init__( self, lambda_1: Tensor, lambda_2: Tensor, tol_lambda_1: float = 1e-2, F=None, ) -> None: self.lambda_1 = lambda_1 self.lambda_2 = lambda_2 self.tol_lambda_1 = tol_lambda_1 self.F = F if F else getF(lambda_1) # Addressing mxnet madness self._power = self.F.power if self.F == mx.nd else self.F.pow @property def args(self) -> List: r""" List: current values of the parameters """ return [self.lambda_1, self.lambda_2] @property def event_dim(self) -> int: return 0 @property def sign(self) -> Tensor: return 1.0 def f(self, z: Tensor) -> Tensor: r""" Forward transformation of observations `z` Parameters ---------- z observations Returns ------- Tensor Transformed observations """ F = self.F lambda_1 = self.lambda_1 lambda_2 = self.lambda_2 tol_lambda_1 = self.tol_lambda_1 _power = self._power return F.where( condition=(F.abs(lambda_1).__ge__(tol_lambda_1).broadcast_like(z)), x=(_power(z + lambda_2, lambda_1) - 1.0) / lambda_1, y=F.log(z + lambda_2), name="Box_Cox_trans", ) def f_inv(self, y: Tensor) -> Tensor: r"""Inverse of the Box-Cox Transform Parameters ---------- y Transformed observations Returns ------- Tensor Observations """ F = self.F lambda_1 = self.lambda_1 lambda_2 = self.lambda_2 tol_lambda_1 = self.tol_lambda_1 _power = self._power # For numerical stability we truncate :math:`y * \lambda_1 + 1.0` at zero. base = F.relu(y * lambda_1 + 1.0) return F.where( condition=(F.abs(lambda_1).__ge__(tol_lambda_1)).broadcast_like(y), x=_power(base, 1.0 / lambda_1) - lambda_2, y=F.exp(y) - lambda_2, name="Box_Cox_inverse_trans", ) def log_abs_det_jac(self, z: Tensor, y: Tensor = None) -> Tensor: r""" Logarithm of the absolute value of the Jacobian determinant corresponding to the Box-Cox Transform is given by .. math:: \log \frac{d}{dz} BoxCox(z; \lambda_1, \lambda_2) = \begin{cases} \log (z + \lambda_2) (\lambda_1 - 1), \quad & \text{if } \lambda_1 \neq 0, \\ -\log (z + \lambda_2), \quad & \text{otherwise.} \end{cases} Note that the derivative of the transformation is always non-negative. Parameters ---------- z observations y not used Returns ------- Tensor """ F = self.F lambda_1 = self.lambda_1 lambda_2 = self.lambda_2 tol_lambda_1 = self.tol_lambda_1 return F.where( condition=F.abs(lambda_1).__ge__(tol_lambda_1), x=F.log(z + lambda_2) * (lambda_1 - 1.0), y=-F.log(z + lambda_2), name="Box_Cox_trans_log_det_jac", ) class BoxCoxTransformOutput(BijectionOutput): bij_cls: type = BoxCoxTranform args_dim: Dict[str, int] = dict(zip(BoxCoxTranform.arg_names, [1, 1])) @validated() def __init__(self, lb_obs: float = 0.0, fix_lambda_2: bool = True) -> None: super().__init__() self.lb_obs = lb_obs self.fix_lambda_2 = fix_lambda_2 def domain_map(self, F, *args: Tensor) -> Tuple[Tensor, ...]: lambda_1, lambda_2 = args if self.fix_lambda_2: lambda_2 = -self.lb_obs * F.ones_like(lambda_2) else: # This makes sure that :math:`z + \lambda_2 > 0`, where :math:`z > lb_obs` lambda_2 = softplus(F, lambda_2) - self.lb_obs * F.ones_like( lambda_2 ) # we squeeze the output since event_shape is () return lambda_1.squeeze(axis=-1), lambda_2.squeeze(axis=-1) @property def event_shape(self) -> Tuple: return () class InverseBoxCoxTransform(InverseBijection): """ Implements the inverse of Box-Cox transformation as a bijection. """ arg_names = ["box_cox.lambda_1", "box_cox.lambda_2"] @validated() def __init__( self, lambda_1: Tensor, lambda_2: Tensor, tol_lambda_1: float = 1e-2, F=None, ) -> None: super().__init__(BoxCoxTranform(lambda_1, lambda_2, tol_lambda_1, F)) @property def event_dim(self) -> int: return 0 class InverseBoxCoxTransformOutput(BoxCoxTransformOutput): bij_cls: type = InverseBoxCoxTransform args_dim: Dict[str, int] = dict( zip(InverseBoxCoxTransform.arg_names, [1, 1]) ) @property def event_shape(self) -> Tuple: return ()
9,282
29.536184
113
py
rankpredictor
rankpredictor-master/sub/gluonts/distribution/mixture.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import List, Optional, Tuple # Third-party imports from mxnet import gluon import numpy as np # First-party imports from gluonts.core.component import validated from gluonts.model.common import Tensor # Relative imports from .distribution import Distribution, _expand_param, getF from .distribution_output import DistributionOutput class MixtureDistribution(Distribution): r""" A mixture distribution where each component is a Distribution. Parameters ---------- mixture_probs A tensor of mixing probabilities. The entries should all be positive and sum to 1 across the last dimension. Shape: (..., k), where k is the number of distributions to be mixed. All axis except the last one should either coincide with the ones from the component distributions, or be 1 (in which case, the mixing coefficient is shared across the axis). components A list of k Distribution objects representing the mixture components. Distributions can be of different types. Each component's support should be made of tensors of shape (..., d). F A module that can either refer to the Symbol API or the NDArray API in MXNet """ is_reparameterizable = False @validated() def __init__( self, mixture_probs: Tensor, components: List[Distribution], F=None ) -> None: self.F = F if F else getF(mixture_probs) # TODO: handle case with all components of the same type more efficiently when sampling # self.all_same = len(set(c.__class__.__name__ for c in components)) == 1 self.mixture_probs = mixture_probs self.components = components @property def batch_shape(self) -> Tuple: return self.components[0].batch_shape @property def event_shape(self) -> Tuple: return self.components[0].event_shape @property def event_dim(self) -> int: return self.components[0].event_dim def log_prob(self, x: Tensor) -> Tensor: F = self.F log_mix_weights = F.log(self.mixture_probs) # compute log probabilities of components component_log_likelihood = F.stack( *[c.log_prob(x) for c in self.components], axis=-1 ) # compute mixture log probability by log-sum-exp summands = log_mix_weights + component_log_likelihood max_val = F.max_axis(summands, axis=-1, keepdims=True) sum_exp = F.sum( F.exp(F.broadcast_minus(summands, max_val)), axis=-1, keepdims=True ) log_sum_exp = F.log(sum_exp) + max_val return log_sum_exp.squeeze(axis=-1) @property def mean(self) -> Tensor: F = self.F mean_values = F.stack(*[c.mean for c in self.components], axis=-1) return F.sum( F.broadcast_mul(mean_values, self.mixture_probs, axis=-1), axis=-1 ) def cdf(self, x: Tensor) -> Tensor: F = self.F cdf_values = F.stack(*[c.cdf(x) for c in self.components], axis=-1) erg = F.sum( F.broadcast_mul(cdf_values, self.mixture_probs, axis=-1), axis=-1 ) return erg @property def stddev(self) -> Tensor: F = self.F stddev_values = F.stack(*[c.stddev for c in self.components], axis=-1) return F.sum( F.broadcast_mul(stddev_values, self.mixture_probs, axis=-1), axis=-1, ) def sample( self, num_samples: Optional[int] = None, dtype=np.float32 ) -> Tensor: samples_list = [c.sample(num_samples, dtype) for c in self.components] samples = self.F.stack(*samples_list, axis=-1) mixture_probs = _expand_param(self.mixture_probs, num_samples) idx = self.F.random.multinomial(mixture_probs) for _ in range(self.event_dim): idx = idx.expand_dims(axis=-1) idx = idx.broadcast_like(samples_list[0]) selected_samples = self.F.pick(data=samples, index=idx, axis=-1) return selected_samples class MixtureArgs(gluon.HybridBlock): def __init__( self, distr_outputs: List[DistributionOutput], prefix: Optional[str] = None, ) -> None: super().__init__() self.num_components = len(distr_outputs) self.component_projections: List[gluon.HybridBlock] = [] with self.name_scope(): self.proj_mixture_probs = gluon.nn.HybridSequential() self.proj_mixture_probs.add( gluon.nn.Dense( self.num_components, prefix=f"{prefix}_pi_", flatten=False ) ) self.proj_mixture_probs.add(gluon.nn.HybridLambda("softmax")) for k, do in enumerate(distr_outputs): self.component_projections.append( do.get_args_proj(prefix=str(k)) ) self.register_child(self.component_projections[-1]) def hybrid_forward(self, F, x: Tensor) -> Tuple[Tensor, ...]: mixture_probs = self.proj_mixture_probs(x) component_args = [c_proj(x) for c_proj in self.component_projections] return tuple([mixture_probs] + component_args) class MixtureDistributionOutput(DistributionOutput): @validated() def __init__(self, distr_outputs: List[DistributionOutput]) -> None: self.num_components = len(distr_outputs) self.distr_outputs = distr_outputs def get_args_proj(self, prefix: Optional[str] = None) -> MixtureArgs: return MixtureArgs(self.distr_outputs, prefix=prefix) # Overwrites the parent class method. def distribution( self, distr_args, scale: Optional[Tensor] = None, **kwargs ) -> MixtureDistribution: mixture_probs = distr_args[0] component_args = distr_args[1:] return MixtureDistribution( mixture_probs=mixture_probs, components=[ do.distribution(args, scale=scale) for do, args in zip(self.distr_outputs, component_args) ], ) @property def event_shape(self) -> Tuple: return self.distr_outputs[0].event_shape
6,791
33.477157
95
py
rankpredictor
rankpredictor-master/sub/gluonts/distribution/lds.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import Tuple, Optional, NamedTuple # Third-party imports import mxnet as mx import numpy as np # First-party imports from gluonts.distribution import Distribution, Gaussian, MultivariateGaussian from gluonts.distribution.distribution import getF from gluonts.model.common import Tensor from gluonts.core.component import validated from gluonts.support.util import make_nd_diag, _broadcast_param from gluonts.support.linalg_util import jitter_cholesky class ParameterBounds: @validated() def __init__(self, lower, upper) -> None: assert ( lower <= upper ), "lower bound should be smaller or equal to upper bound" self.lower = lower self.upper = upper class LDS(Distribution): r""" Implements Linear Dynamical System (LDS) as a distribution. The LDS is given by .. math:: z_t = A_t l_{t-1} + b_t + \epsilon_t \\ l_t = C_t l_{t-1} + g_t \nu where .. math:: \epsilon_t = N(0, S_v) \\ \nu = N(0, 1) :math:`A_t`, :math:`C_t` and :math:`g_t` are the emission, transition and innovation coefficients respectively. The residual terms are denoted by :math:`b_t`. The target :math:`z_t` can be :math:`d`-dimensional in which case .. math:: A_t \in R^{d \times h}, b_t \in R^{d}, C_t \in R^{h \times h}, g_t \in R^{h} where :math:`h` is dimension of the latent state. Parameters ---------- emission_coeff Tensor of shape (batch_size, seq_length, obs_dim, latent_dim) transition_coeff Tensor of shape (batch_size, seq_length, latent_dim, latent_dim) innovation_coeff Tensor of shape (batch_size, seq_length, latent_dim) noise_std Tensor of shape (batch_size, seq_length, obs_dim) residuals Tensor of shape (batch_size, seq_length, obs_dim) prior_mean Tensor of shape (batch_size, latent_dim) prior_cov Tensor of shape (batch_size, latent_dim, latent_dim) latent_dim Dimension of the latent state output_dim Dimension of the output seq_length Sequence length F """ @validated() def __init__( self, emission_coeff: Tensor, transition_coeff: Tensor, innovation_coeff: Tensor, noise_std: Tensor, residuals: Tensor, prior_mean: Tensor, prior_cov: Tensor, latent_dim: int, output_dim: int, seq_length: int, F=None, ) -> None: self.latent_dim = latent_dim self.output_dim = output_dim self.seq_length = seq_length # Split coefficients along time axis for easy access # emission_coef[t]: (batch_size, obs_dim, latent_dim) self.emission_coeff = emission_coeff.split( axis=1, num_outputs=self.seq_length, squeeze_axis=True ) # innovation_coef[t]: (batch_size, latent_dim) self.innovation_coeff = innovation_coeff.split( axis=1, num_outputs=self.seq_length, squeeze_axis=False ) # transition_coeff: (batch_size, latent_dim, latent_dim) self.transition_coeff = transition_coeff.split( axis=1, num_outputs=self.seq_length, squeeze_axis=True ) # noise_std[t]: (batch_size, obs_dim) self.noise_std = noise_std.split( axis=1, num_outputs=self.seq_length, squeeze_axis=True ) # residuals[t]: (batch_size, obs_dim) self.residuals = residuals.split( axis=1, num_outputs=self.seq_length, squeeze_axis=True ) self.prior_mean = prior_mean self.prior_cov = prior_cov self.F = F if F else getF(noise_std) @property def batch_shape(self) -> Tuple: return self.emission_coeff[0].shape[:1] + (self.seq_length,) @property def event_shape(self) -> Tuple: return (self.output_dim,) @property def event_dim(self) -> int: return 2 def log_prob( self, x: Tensor, scale: Optional[Tensor] = None, observed: Optional[Tensor] = None, ): """ Compute the log probability of observations. This method also returns the final state of the system. Parameters ---------- x Observations, shape (batch_size, seq_length, output_dim) scale Scale of each sequence in x, shape (batch_size, output_dim) observed Flag tensor indicating which observations are genuine (1.0) and which are missing (0.0) Returns ------- Tensor Log probabilities, shape (batch_size, seq_length) Tensor Final mean, shape (batch_size, latent_dim) Tensor Final covariance, shape (batch_size, latent_dim, latent_dim) """ if scale is not None: x = self.F.broadcast_div(x, scale.expand_dims(axis=1)) # TODO: Based on form of the prior decide to do either filtering # or residual-sum-of-squares log_p, final_mean, final_cov = self.kalman_filter(x, observed) return log_p, final_mean, final_cov def kalman_filter( self, targets: Tensor, observed: Tensor ) -> Tuple[Tensor, ...]: """ Performs Kalman filtering given observations. Parameters ---------- targets Observations, shape (batch_size, seq_length, output_dim) observed Flag tensor indicating which observations are genuine (1.0) and which are missing (0.0) Returns ------- Tensor Log probabilities, shape (batch_size, seq_length) Tensor Mean of p(l_T | l_{T-1}), where T is seq_length, with shape (batch_size, latent_dim) Tensor Covariance of p(l_T | l_{T-1}), where T is seq_length, with shape (batch_size, latent_dim, latent_dim) """ F = self.F # targets[t]: (batch_size, obs_dim) targets = targets.split( axis=1, num_outputs=self.seq_length, squeeze_axis=True ) log_p_seq = [] mean = self.prior_mean cov = self.prior_cov observed = ( observed.split( axis=1, num_outputs=self.seq_length, squeeze_axis=True ) if observed is not None else None ) for t in range(self.seq_length): # Compute the filtered distribution # p(l_t | z_1, ..., z_{t + 1}) # and log - probability # log p(z_t | z_0, z_{t - 1}) filtered_mean, filtered_cov, log_p = kalman_filter_step( F, target=targets[t], prior_mean=mean, prior_cov=cov, emission_coeff=self.emission_coeff[t], residual=self.residuals[t], noise_std=self.noise_std[t], latent_dim=self.latent_dim, output_dim=self.output_dim, ) log_p_seq.append(log_p.expand_dims(axis=1)) # Mean of p(l_{t+1} | l_t) mean = F.linalg_gemm2( self.transition_coeff[t], ( filtered_mean.expand_dims(axis=-1) if observed is None else F.where( observed[t], x=filtered_mean, y=mean ).expand_dims(axis=-1) ), ).squeeze(axis=-1) # Covariance of p(l_{t+1} | l_t) cov = F.linalg_gemm2( self.transition_coeff[t], F.linalg_gemm2( ( filtered_cov if observed is None else F.where(observed[t], x=filtered_cov, y=cov) ), self.transition_coeff[t], transpose_b=True, ), ) + F.linalg_gemm2( self.innovation_coeff[t], self.innovation_coeff[t], transpose_a=True, ) # Return sequence of log likelihoods, as well as # final mean and covariance of p(l_T | l_{T-1} where T is seq_length return F.concat(*log_p_seq, dim=1), mean, cov def sample( self, num_samples: Optional[int] = None, scale: Optional[Tensor] = None ) -> Tensor: r""" Generates samples from the LDS: p(z_1, z_2, \ldots, z_{`seq_length`}). Parameters ---------- num_samples Number of samples to generate scale Scale of each sequence in x, shape (batch_size, output_dim) Returns ------- Tensor Samples, shape (num_samples, batch_size, seq_length, output_dim) """ F = self.F # Note on shapes: here we work with tensors of the following shape # in each time step t: (num_samples, batch_size, dim, dim), # where dim can be obs_dim or latent_dim or a constant 1 to facilitate # generalized matrix multiplication (gemm2) # Sample observation noise for all time steps # noise_std: (batch_size, seq_length, obs_dim, 1) noise_std = F.stack(*self.noise_std, axis=1).expand_dims(axis=-1) # samples_eps_obs[t]: (num_samples, batch_size, obs_dim, 1) samples_eps_obs = ( Gaussian(noise_std.zeros_like(), noise_std) .sample(num_samples) .split(axis=-3, num_outputs=self.seq_length, squeeze_axis=True) ) # Sample standard normal for all time steps # samples_eps_std_normal[t]: (num_samples, batch_size, obs_dim, 1) samples_std_normal = ( Gaussian(noise_std.zeros_like(), noise_std.ones_like()) .sample(num_samples) .split(axis=-3, num_outputs=self.seq_length, squeeze_axis=True) ) # Sample the prior state. # samples_lat_state: (num_samples, batch_size, latent_dim, 1) # The prior covariance is observed to be slightly negative definite whenever there is # excessive zero padding at the beginning of the time series. # We add positive tolerance to the diagonal to avoid numerical issues. # Note that `jitter_cholesky` adds positive tolerance only if the decomposition without jitter fails. state = MultivariateGaussian( self.prior_mean, jitter_cholesky( F, self.prior_cov, self.latent_dim, float_type=np.float32 ), ) samples_lat_state = state.sample(num_samples).expand_dims(axis=-1) samples_seq = [] for t in range(self.seq_length): # Expand all coefficients to include samples in axis 0 # emission_coeff_t: (num_samples, batch_size, obs_dim, latent_dim) # transition_coeff_t: # (num_samples, batch_size, latent_dim, latent_dim) # innovation_coeff_t: (num_samples, batch_size, 1, latent_dim) emission_coeff_t, transition_coeff_t, innovation_coeff_t = [ _broadcast_param(coeff, axes=[0], sizes=[num_samples]) if num_samples is not None else coeff for coeff in [ self.emission_coeff[t], self.transition_coeff[t], self.innovation_coeff[t], ] ] # Expand residuals as well # residual_t: (num_samples, batch_size, obs_dim, 1) residual_t = ( _broadcast_param( self.residuals[t].expand_dims(axis=-1), axes=[0], sizes=[num_samples], ) if num_samples is not None else self.residuals[t].expand_dims(axis=-1) ) # (num_samples, batch_size, 1, obs_dim) samples_t = ( F.linalg_gemm2(emission_coeff_t, samples_lat_state) + residual_t + samples_eps_obs[t] ) samples_t = ( samples_t.swapaxes(dim1=2, dim2=3) if num_samples is not None else samples_t.swapaxes(dim1=1, dim2=2) ) samples_seq.append(samples_t) # sample next state: (num_samples, batch_size, latent_dim, 1) samples_lat_state = F.linalg_gemm2( transition_coeff_t, samples_lat_state ) + F.linalg_gemm2( innovation_coeff_t, samples_std_normal[t], transpose_a=True ) # (num_samples, batch_size, seq_length, obs_dim) samples = F.concat(*samples_seq, dim=-2) return ( samples if scale is None else F.broadcast_mul( samples, scale.expand_dims(axis=1).expand_dims(axis=0) if num_samples is not None else scale.expand_dims(axis=1), ) ) def sample_marginals( self, num_samples: Optional[int] = None, scale: Optional[Tensor] = None ) -> Tensor: r""" Generates samples from the marginals p(z_t), t = 1, \ldots, `seq_length`. Parameters ---------- num_samples Number of samples to generate scale Scale of each sequence in x, shape (batch_size, output_dim) Returns ------- Tensor Samples, shape (num_samples, batch_size, seq_length, output_dim) """ F = self.F state_mean = self.prior_mean.expand_dims(axis=-1) state_cov = self.prior_cov output_mean_seq = [] output_cov_seq = [] for t in range(self.seq_length): # compute and store observation mean at time t output_mean = F.linalg_gemm2( self.emission_coeff[t], state_mean ) + self.residuals[t].expand_dims(axis=-1) output_mean_seq.append(output_mean) # compute and store observation cov at time t output_cov = F.linalg_gemm2( self.emission_coeff[t], F.linalg_gemm2( state_cov, self.emission_coeff[t], transpose_b=True ), ) + make_nd_diag( F=F, x=self.noise_std[t] * self.noise_std[t], d=self.output_dim ) output_cov_seq.append(output_cov.expand_dims(axis=1)) state_mean = F.linalg_gemm2(self.transition_coeff[t], state_mean) state_cov = F.linalg_gemm2( self.transition_coeff[t], F.linalg_gemm2( state_cov, self.transition_coeff[t], transpose_b=True ), ) + F.linalg_gemm2( self.innovation_coeff[t], self.innovation_coeff[t], transpose_a=True, ) output_mean = F.concat(*output_mean_seq, dim=1) output_cov = F.concat(*output_cov_seq, dim=1) L = F.linalg_potrf(output_cov) output_distribution = MultivariateGaussian(output_mean, L) samples = output_distribution.sample(num_samples=num_samples) return ( samples if scale is None else F.broadcast_mul(samples, scale.expand_dims(axis=1)) ) class LDSArgsProj(mx.gluon.HybridBlock): def __init__( self, output_dim: int, noise_std_bounds: ParameterBounds, innovation_bounds: ParameterBounds, ) -> None: super().__init__() self.output_dim = output_dim self.dense_noise_std = mx.gluon.nn.Dense( units=1, flatten=False, activation="sigmoid" ) self.dense_innovation = mx.gluon.nn.Dense( units=1, flatten=False, activation="sigmoid" ) self.dense_residual = mx.gluon.nn.Dense( units=output_dim, flatten=False ) self.innovation_bounds = innovation_bounds self.noise_std_bounds = noise_std_bounds # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward(self, F, x: Tensor) -> Tuple[Tensor, Tensor, Tensor]: noise_std = ( self.dense_noise_std(x) * (self.noise_std_bounds.upper - self.noise_std_bounds.lower) + self.noise_std_bounds.lower ) innovation = ( self.dense_innovation(x) * (self.innovation_bounds.upper - self.innovation_bounds.lower) + self.innovation_bounds.lower ) residual = self.dense_residual(x) return noise_std, innovation, residual def kalman_filter_step( F, target: Tensor, prior_mean: Tensor, prior_cov: Tensor, emission_coeff: Tensor, residual: Tensor, noise_std: Tensor, latent_dim: int, output_dim: int, ): """ One step of the Kalman filter. This function computes the filtered state (mean and covariance) given the linear system coefficients the prior state (mean and variance), as well as observations. Parameters ---------- F target Observations of the system output, shape (batch_size, output_dim) prior_mean Prior mean of the latent state, shape (batch_size, latent_dim) prior_cov Prior covariance of the latent state, shape (batch_size, latent_dim, latent_dim) emission_coeff Emission coefficient, shape (batch_size, output_dim, latent_dim) residual Residual component, shape (batch_size, output_dim) noise_std Standard deviation of the output noise, shape (batch_size, output_dim) latent_dim Dimension of the latent state vector Returns ------- Tensor Filtered_mean, shape (batch_size, latent_dim) Tensor Filtered_covariance, shape (batch_size, latent_dim, latent_dim) Tensor Log probability, shape (batch_size, ) """ # output_mean: mean of the target (batch_size, obs_dim) output_mean = F.linalg_gemm2( emission_coeff, prior_mean.expand_dims(axis=-1) ).squeeze(axis=-1) # noise covariance noise_cov = make_nd_diag(F=F, x=noise_std * noise_std, d=output_dim) S_hh_x_A_tr = F.linalg_gemm2(prior_cov, emission_coeff, transpose_b=True) # covariance of the target output_cov = F.linalg_gemm2(emission_coeff, S_hh_x_A_tr) + noise_cov # compute the Cholesky decomposition output_cov = LL^T L_output_cov = F.linalg_potrf(output_cov) # Compute Kalman gain matrix K: # K = S_hh X with X = A^T output_cov^{-1} # We have X = A^T output_cov^{-1} => X output_cov = A^T => X LL^T = A^T # We can thus obtain X by solving two linear systems involving L kalman_gain = F.linalg_trsm( L_output_cov, F.linalg_trsm( L_output_cov, S_hh_x_A_tr, rightside=True, transpose=True ), rightside=True, ) # compute the error target_minus_residual = target - residual delta = target_minus_residual - output_mean # filtered estimates filtered_mean = prior_mean.expand_dims(axis=-1) + F.linalg_gemm2( kalman_gain, delta.expand_dims(axis=-1) ) filtered_mean = filtered_mean.squeeze(axis=-1) # Joseph's symmetrized update for covariance: ImKA = F.broadcast_sub( F.eye(latent_dim), F.linalg_gemm2(kalman_gain, emission_coeff) ) filtered_cov = F.linalg_gemm2( ImKA, F.linalg_gemm2(prior_cov, ImKA, transpose_b=True) ) + F.linalg_gemm2( kalman_gain, F.linalg_gemm2(noise_cov, kalman_gain, transpose_b=True) ) # likelihood term: (batch_size,) log_p = MultivariateGaussian(output_mean, L_output_cov).log_prob( target_minus_residual ) return filtered_mean, filtered_cov, log_p
20,583
31.881789
109
py
rankpredictor
rankpredictor-master/sub/gluonts/distribution/distribution_output.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import Callable, Dict, Optional, Tuple # Third-party imports import numpy as np from mxnet import gluon # First-party imports from gluonts.core.component import DType, validated from gluonts.distribution.bijection import AffineTransformation from gluonts.model.common import Tensor # Relative imports from .distribution import Distribution from .transformed_distribution import TransformedDistribution class ArgProj(gluon.HybridBlock): r""" A block that can be used to project from a dense layer to distribution arguments. Parameters ---------- dim_args Dictionary with string key and int value dimension of each arguments that will be passed to the domain map, the names are used as parameters prefix. domain_map Function returning a tuple containing one tensor a function or a HybridBlock. This will be called with num_args arguments and should return a tuple of outputs that will be used when calling the distribution constructor. """ def __init__( self, args_dim: Dict[str, int], domain_map: Callable[..., Tuple[Tensor]], dtype: DType = np.float32, prefix: Optional[str] = None, **kwargs, ) -> None: super().__init__(**kwargs) self.args_dim = args_dim self.dtype = dtype self.proj = [ gluon.nn.Dense( dim, flatten=False, dtype=self.dtype, prefix=f"{prefix}_distr_{name}_", ) for name, dim in args_dim.items() ] for dense in self.proj: self.register_child(dense) self.domain_map = domain_map # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward(self, F, x: Tensor) -> Tuple[Tensor]: params_unbounded = [proj(x) for proj in self.proj] return self.domain_map(*params_unbounded) class Output: r""" Class to connect a network to some output """ args_dim: Dict[str, int] _dtype: DType = np.float32 @property def dtype(self): return self._dtype @dtype.setter def dtype(self, dtype: DType): self._dtype = dtype def get_args_proj(self, prefix: Optional[str] = None) -> ArgProj: return ArgProj( args_dim=self.args_dim, domain_map=gluon.nn.HybridLambda(self.domain_map), prefix=prefix, dtype=self.dtype, ) def domain_map(self, F, *args: Tensor): raise NotImplementedError() class DistributionOutput(Output): r""" Class to construct a distribution given the output of a network. """ distr_cls: type @validated() def __init__(self) -> None: pass def distribution( self, distr_args, scale: Optional[Tensor] = None ) -> Distribution: r""" Construct the associated distribution, given the collection of constructor arguments and, optionally, a scale tensor. Parameters ---------- distr_args Constructor arguments for the underlying Distribution type. scale Optional tensor, of the same shape as the batch_shape+event_shape of the resulting distribution. """ if scale is None: return self.distr_cls(*distr_args) else: distr = self.distr_cls(*distr_args) return TransformedDistribution( distr, [AffineTransformation(scale=scale)] ) @property def event_shape(self) -> Tuple: r""" Shape of each individual event contemplated by the distributions that this object constructs. """ raise NotImplementedError() @property def event_dim(self) -> int: r""" Number of event dimensions, i.e., length of the `event_shape` tuple, of the distributions that this object constructs. """ return len(self.event_shape) def domain_map(self, F, *args: Tensor): r""" Converts arguments to the right shape and domain. The domain depends on the type of distribution, while the correct shape is obtained by reshaping the trailing axis in such a way that the returned tensors define a distribution of the right event_shape. """ raise NotImplementedError()
5,021
29.253012
76
py
rankpredictor
rankpredictor-master/sub/gluonts/distribution/transformed_distribution_output.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from collections import ChainMap from typing import Optional, Tuple, List # Third-party imports import numpy as np from mxnet import gluon # First-party imports from gluonts.distribution import Distribution from gluonts.distribution.bijection import AffineTransformation from gluonts.distribution.bijection_output import BijectionOutput from gluonts.distribution.distribution_output import ( ArgProj, DistributionOutput, ) from gluonts.distribution.transformed_distribution import ( TransformedDistribution, ) from gluonts.model.common import Tensor from gluonts.core.component import validated class TransformedDistributionOutput(DistributionOutput): r""" Class to connect a network to a distribution that is transformed by a sequence of learnable bijections. """ @validated() def __init__( self, base_distr_output: DistributionOutput, transforms_output: List[BijectionOutput], ) -> None: super().__init__() self.base_distr_output = base_distr_output self.transforms_output = transforms_output self.base_distr_args_dim = base_distr_output.args_dim self.transforms_args_dim = [ transform.args_dim for transform in transforms_output ] def _fuse(t1: Tuple, t2: Tuple) -> Tuple: if len(t1) > len(t2): t1, t2 = t2, t1 # from here on len(t2) >= len(t1) assert t2[-len(t1) :] == t1 return t2 self._event_shape: Tuple[int, ...] = () for to in self.transforms_output: self._event_shape = _fuse(self._event_shape, to.event_shape) def get_args_proj(self, prefix: Optional[str] = None) -> ArgProj: return ArgProj( args_dim=dict( self.base_distr_args_dim, **dict(ChainMap(*self.transforms_args_dim)), ), domain_map=gluon.nn.HybridLambda(self.domain_map), prefix=prefix, ) def _split_args(self, args): # Since hybrid_forward does not support dictionary, # we have to separate the raw outputs of the network based on the indices # and map them to the learnable parameters num_distr_args = len(self.base_distr_args_dim) distr_args = args[0:num_distr_args] num_transforms_args = [ len(transform_dim_args) for transform_dim_args in self.transforms_args_dim ] # starting indices of arguments for each transformation num_args_cumsum = np.cumsum([num_distr_args] + num_transforms_args) # get the arguments for each of the transformations transforms_args = list( map( lambda ixs: args[ixs[0] : ixs[1]], zip(num_args_cumsum, num_args_cumsum[1:]), ) ) return distr_args, transforms_args def domain_map(self, F, *args: Tensor): distr_args, transforms_args = self._split_args(args) distr_params = self.base_distr_output.domain_map(F, *distr_args) transforms_params = [ transform_output.domain_map(F, *transform_args) for transform_output, transform_args in zip( self.transforms_output, transforms_args ) ] # flatten the nested tuple return sum(tuple([distr_params] + transforms_params), ()) def distribution( self, distr_args, scale: Optional[Tensor] = None, **kwargs ) -> Distribution: distr_args, transforms_args = self._split_args(distr_args) distr = self.base_distr_output.distr_cls(*distr_args) transforms = [ transform_output.bij_cls(*bij_args) for transform_output, bij_args in zip( self.transforms_output, transforms_args ) ] trans_distr = TransformedDistribution(distr, transforms) # Apply scaling as well at the end if scale is not None! if scale is None: return trans_distr else: return TransformedDistribution( trans_distr, [AffineTransformation(scale=scale)] ) @property def event_shape(self) -> Tuple: return self._event_shape
4,855
33.197183
81
py
rankpredictor
rankpredictor-master/sub/gluonts/distribution/piecewise_linear.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import Dict, Optional, Tuple, cast, List # Third-party imports import numpy as np # First-party imports from gluonts.core.component import validated from gluonts.distribution.bijection import AffineTransformation, Bijection from gluonts.distribution.distribution import Distribution, getF from gluonts.distribution.distribution_output import DistributionOutput from gluonts.distribution.transformed_distribution import ( TransformedDistribution, ) from gluonts.model.common import Tensor from gluonts.support import util class PiecewiseLinear(Distribution): r""" Piecewise linear distribution. This class represents the *quantile function* (i.e., the inverse CDF) associated with the a distribution, as a continuous, non-decreasing, piecewise linear function defined in the [0, 1] interval: .. math:: q(x; \gamma, b, d) = \gamma + \sum_{l=0}^L b_l (x_l - d_l)_+ where the input :math:`x \in [0,1]` and the parameters are - :math:`\gamma`: intercept at 0 - :math:`b`: differences of the slopes in consecutive pieces - :math:`d`: knot positions Parameters ---------- gamma Tensor containing the intercepts at zero slopes Tensor containing the slopes of each linear piece. All coefficients must be positive. Shape: ``(*gamma.shape, num_pieces)`` knot_spacings Tensor containing the spacings between knots in the splines. All coefficients must be positive and sum to one on the last axis. Shape: ``(*gamma.shape, num_pieces)`` F """ is_reparameterizable = False @validated() def __init__( self, gamma: Tensor, slopes: Tensor, knot_spacings: Tensor, F=None ) -> None: self.F = F if F else getF(gamma) self.gamma = gamma # Since most of the calculations are easily expressed in the original parameters, we transform the # learned parameters back self.b, self.knot_positions = PiecewiseLinear._to_orig_params( self.F, slopes, knot_spacings ) @staticmethod def _to_orig_params( F, slopes: Tensor, knot_spacings: Tensor ) -> Tuple[Tensor, Tensor]: r""" Convert the trainable parameters to the original parameters of the splines, i.e., convert the slopes of each piece to the difference between slopes of consecutive pieces and knot spacings to knot positions. Parameters ---------- F slopes Tensor of shape (*gamma.shape, num_pieces) knot_spacings Tensor of shape (*gamma.shape, num_pieces) Returns ------- Tensor Tensor of shape (*gamma.shape, num_pieces) Tensor Tensor of shape (*gamma.shape, num_pieces) """ # b: the difference between slopes of consecutive pieces # shape (..., num_pieces - 1) b = F.slice_axis(slopes, axis=-1, begin=1, end=None) - F.slice_axis( slopes, axis=-1, begin=0, end=-1 ) # Add slope of first piece to b: b_0 = m_0 m_0 = F.slice_axis(slopes, axis=-1, begin=0, end=1) b = F.concat(m_0, b, dim=-1) # The actual position of the knots is obtained by cumulative sum of # the knot spacings. The first knot position is always 0 for quantile # functions; cumsum will take care of that. knot_positions = util.cumsum(F, knot_spacings, exclusive=True) return b, knot_positions def sample( self, num_samples: Optional[int] = None, dtype=np.float32 ) -> Tensor: F = self.F # if num_samples=None then u should have the same shape as gamma, i.e., (dim,) # else u should be (num_samples, dim) # Note: there is no need to extend the parameters to (num_samples, dim, ...) # Thankfully samples returned by `uniform_like` have the expected datatype. u = F.random.uniform_like( data=( self.gamma if num_samples is None else self.gamma.expand_dims(axis=0).repeat( axis=0, repeats=num_samples ) ) ) sample = self.quantile(u) if num_samples is None: sample = F.squeeze(sample, axis=0) return sample # overwrites the loss method of the Distribution class def loss(self, x: Tensor) -> Tensor: return self.crps(x) def crps(self, x: Tensor) -> Tensor: r""" Compute CRPS in analytical form. Parameters ---------- x Observation to evaluate. Shape equals to gamma.shape. Returns ------- Tensor Tensor containing the CRPS. """ F = self.F gamma, b, knot_positions = self.gamma, self.b, self.knot_positions a_tilde = self.cdf(x) max_a_tilde_knots = F.broadcast_maximum( a_tilde.expand_dims(axis=-1), knot_positions ) knots_cubed = F.broadcast_power(self.knot_positions, F.ones(1) * 3.0) coeff = ( (1.0 - knots_cubed) / 3.0 - knot_positions - F.square(max_a_tilde_knots) + 2 * max_a_tilde_knots * knot_positions ) crps = ( (2 * a_tilde - 1) * x + (1 - 2 * a_tilde) * gamma + F.sum(b * coeff, axis=-1, keepdims=False) ) return crps def cdf(self, x: Tensor) -> Tensor: r""" Computes the quantile level :math:`\alpha` such that :math:`q(\alpha) = x`. Parameters ---------- x Tensor of shape gamma.shape Returns ------- Tensor Tensor of shape gamma.shape """ F = self.F gamma, b, knot_positions = self.gamma, self.b, self.knot_positions quantiles_at_knots = self.quantile_internal(knot_positions, axis=-2) # Mask to nullify the terms corresponding to knots larger than l_0, which is the largest knot # (quantile level) such that the quantile at l_0, s(l_0) < x. # (..., num_pieces) mask = F.broadcast_lesser(quantiles_at_knots, x.expand_dims(axis=-1)) slope_l0 = F.sum(b * mask, axis=-1, keepdims=False) # slope_l0 can be zero in which case a_tilde = 0. # The following is to circumvent mxnet issue with "where" operator which returns nan even if the statement # you are interested in does not result in nan (but the "else" statement evaluates to nan). slope_l0_nz = F.where( slope_l0 == F.zeros_like(slope_l0), F.ones_like(x), slope_l0 ) a_tilde = F.where( slope_l0 == F.zeros_like(slope_l0), F.zeros_like(x), ( x - gamma + F.sum(b * knot_positions * mask, axis=-1, keepdims=False) ) / slope_l0_nz, ) return a_tilde def quantile(self, level: Tensor) -> Tensor: return self.quantile_internal(level, axis=0) def quantile_internal( self, x: Tensor, axis: Optional[int] = None ) -> Tensor: r""" Evaluates the quantile function at the quantile levels contained in `x`. Parameters ---------- x Tensor of shape ``*gamma.shape`` if axis=None, or containing an additional axis on the specified position, otherwise. axis Index of the axis containing the different quantile levels which are to be computed. Returns ------- Tensor Quantiles tensor, of the same shape as x. """ F = self.F # shapes of self # self.gamma: (*batch_shape) # self.knot_positions, self.b: (*batch_shape, num_pieces) # axis=None - passed at inference when num_samples is None # The shape of x is (*batch_shape). # The shapes of the parameters should be: # gamma: (*batch_shape), knot_positions, b: (*batch_shape, num_pieces) # They match the self. counterparts so no reshaping is needed # axis=0 - passed at inference when num_samples is not None # The shape of x is (num_samples, *batch_shape). # The shapes of the parameters should be: # gamma: (num_samples, *batch_shape), knot_positions, b: (num_samples, *batch_shape, num_pieces), # They do not match the self. counterparts and we need to expand the axis=0 to all of them. # axis=-2 - passed at training when we evaluate quantiles at knot_positions in order to compute a_tilde # The shape of x is shape(x) = shape(knot_positions) = (*batch_shape, num_pieces). # The shape of the parameters shopuld be: # gamma: (*batch_shape, 1), knot_positions: (*batch_shape, 1, num_pieces), b: (*batch_shape, 1, num_pieces) # They do not match the self. counterparts and we need to expand axis=-1 for gamma and axis=-2 for the rest. if axis is not None: gamma = self.gamma.expand_dims(axis=axis if axis == 0 else -1) knot_positions = self.knot_positions.expand_dims(axis=axis) b = self.b.expand_dims(axis=axis) else: gamma, knot_positions, b = self.gamma, self.knot_positions, self.b x_minus_knots = F.broadcast_minus( x.expand_dims(axis=-1), knot_positions ) quantile = F.broadcast_add( gamma, F.sum(F.broadcast_mul(b, F.relu(x_minus_knots)), axis=-1) ) return quantile @property def batch_shape(self) -> Tuple: return self.gamma.shape @property def event_shape(self) -> Tuple: return () @property def event_dim(self) -> int: return 0 class PiecewiseLinearOutput(DistributionOutput): distr_cls: type = PiecewiseLinear @validated() def __init__(self, num_pieces: int) -> None: assert ( isinstance(num_pieces, int) and num_pieces > 1 ), "num_pieces should be an integer larger than 1" self.num_pieces = num_pieces self.args_dim = cast( Dict[str, int], {"gamma": 1, "slopes": num_pieces, "knot_spacings": num_pieces}, ) @classmethod def domain_map(cls, F, gamma, slopes, knot_spacings): # slopes of the pieces are non-negative thresh = F.zeros_like(slopes) + 20.0 I_grt_thresh = F.broadcast_greater_equal(slopes, thresh) I_sml_thresh = F.broadcast_lesser(slopes, thresh) slopes_proj = ( I_sml_thresh * F.Activation(data=(I_sml_thresh * slopes), act_type="softrelu") + I_grt_thresh * slopes ) # slopes = F.Activation(slopes, 'softrelu') # the spacing between the knots should be in [0, 1] and sum to 1 knot_spacings_proj = F.softmax(knot_spacings) return gamma.squeeze(axis=-1), slopes_proj, knot_spacings_proj def distribution( self, distr_args, scale: Optional[Tensor] = None, **kwargs ) -> PiecewiseLinear: if scale is None: return self.distr_cls(*distr_args) else: distr = self.distr_cls(*distr_args) return TransformedPiecewiseLinear( distr, [AffineTransformation(scale=scale)] ) @property def event_shape(self) -> Tuple: return () # Need to inherit from PiecewiseLinear to get the overwritten loss method. class TransformedPiecewiseLinear(TransformedDistribution, PiecewiseLinear): @validated() def __init__( self, base_distribution: PiecewiseLinear, transforms: List[Bijection] ) -> None: super().__init__(base_distribution, transforms) def crps(self, y: Tensor) -> Tensor: # TODO: use event_shape F = getF(y) for t in self.transforms[::-1]: assert isinstance( t, AffineTransformation ), "Not an AffineTransformation" assert ( t.scale is not None and t.loc is None ), "Not a scaling transformation" scale = t.scale x = t.f_inv(y) # (..., 1) p = self.base_distribution.crps(x) return F.broadcast_mul(p, scale)
12,984
31.873418
116
py
rankpredictor
rankpredictor-master/sub/gluonts/support/linalg_util.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import Optional # Third-party imports import mxnet as mx import numpy as np # First-party imports from gluonts.core.component import DType from gluonts.model.common import Tensor def batch_diagonal( F, matrix: Tensor, num_data_points: Optional[int] = None, ctx=mx.cpu(), float_type=np.float32, ) -> Tensor: """ This function extracts the diagonal of a batch matrix. Parameters ---------- F A module that can either refer to the Symbol API or the NDArray API in MXNet. matrix matrix of shape (batch_size, num_data_points, num_data_points). num_data_points Number of rows in the kernel_matrix. Returns ------- Tensor Diagonals of kernel_matrix of shape (batch_size, num_data_points, 1). """ return F.linalg.gemm2( F.broadcast_mul( F.eye(num_data_points, ctx=ctx, dtype=float_type), matrix ), F.ones_like(F.slice_axis(matrix, axis=2, begin=0, end=1)), ) def lower_triangular_ones(F, d: int, offset: int = 0) -> Tensor: """ Constructs a lower triangular matrix consisting of ones. Parameters ---------- F d Dimension of the output tensor, whose shape will be (d, d). offset Indicates how many diagonals to set to zero in the lower triangular part. By default, offset = 0, so the output matrix contains also the main diagonal. For example, if offset = 1 then the output will be a strictly lower triangular matrix (i.e. the main diagonal will be zero). Returns ------- Tensor Tensor of shape (d, d) consisting of ones in the strictly lower triangular part, and zeros elsewhere. """ mask = F.zeros_like(F.eye(d)) for k in range(offset, d): mask = mask + F.eye(d, d, -k) return mask # noinspection PyMethodOverriding,PyPep8Naming def jitter_cholesky_eig( F, matrix: Tensor, num_data_points: Optional[int] = None, ctx: mx.Context = mx.Context("cpu"), float_type: DType = np.float64, diag_weight: float = 1e-6, ) -> Tensor: """ This function applies the jitter method using the eigenvalue decomposition. The eigenvalues are bound below by the jitter, which is proportional to the mean of the diagonal elements Parameters ---------- F A module that can either refer to the Symbol API or the NDArray API in MXNet. matrix Matrix of shape (batch_size, num_data_points, num_data_points). num_data_points Number of rows in the kernel_matrix. ctx Determines whether to compute on the cpu or gpu. float_type Determines whether to use single or double precision. Returns ------- Tensor Returns the approximate lower triangular Cholesky factor `L` of shape (batch_size, num_data_points, num_data_points) """ diag = batch_diagonal( F, matrix, num_data_points, ctx, float_type ) # shape (batch_size, num_data_points, 1) diag_mean = diag.mean(axis=1).expand_dims( axis=2 ) # shape (batch_size, 1, 1) U, Lambda = F.linalg.syevd(matrix) jitter = F.broadcast_mul(diag_mean, F.ones_like(diag)) * diag_weight # `K = U^TLambdaU`, where the rows of `U` are the eigenvectors of `K`. # The eigendecomposition :math:`U^TLambdaU` is used instead of :math: ULambdaU^T, sine # to utilize row-based computation (see Section 4, Seeger et al., 2018) return F.linalg.potrf( F.linalg.gemm2( U, F.linalg.gemm2( F.broadcast_mul( F.eye(num_data_points, ctx=ctx, dtype=float_type), F.maximum(jitter, Lambda.expand_dims(axis=2)), ), U, ), transpose_a=True, ) ) # noinspection PyMethodOverriding,PyPep8Naming def jitter_cholesky( F, matrix: Tensor, num_data_points: Optional[int] = None, ctx: mx.Context = mx.Context("cpu"), float_type: DType = np.float64, max_iter_jitter: int = 10, neg_tol: float = -1e-8, diag_weight: float = 1e-6, increase_jitter: int = 10, ) -> Optional[Tensor]: """ This function applies the jitter method. It iteratively tries to compute the Cholesky decomposition and adds a positive tolerance to the diagonal that increases at each iteration until the matrix is positive definite or the maximum number of iterations has been reached. Parameters ---------- matrix Kernel matrix of shape (batch_size, num_data_points, num_data_points). num_data_points Number of rows in the kernel_matrix. ctx Determines whether to compute on the cpu or gpu. float_type Determines whether to use single or double precision. max_iter_jitter Maximum number of iterations for jitter to iteratively make the matrix positive definite. neg_tol Parameter in the jitter methods to eliminate eliminate matrices with diagonal elements smaller than this when checking if a matrix is positive definite. diag_weight Multiple of mean of diagonal entries to initialize the jitter. increase_jitter Each iteration multiply by jitter by this amount Returns ------- Optional[Tensor] The method either fails to make the matrix positive definite within the maximum number of iterations and outputs an error or succeeds and returns the lower triangular Cholesky factor `L` of shape (batch_size, num_data_points, num_data_points) """ num_iter = 0 diag = batch_diagonal( F, matrix, num_data_points, ctx, float_type ) # shape (batch_size, num_data_points, 1) diag_mean = diag.mean(axis=1).expand_dims( axis=2 ) # shape (batch_size, 1, 1) jitter = F.zeros_like(diag) # shape (batch_size, num_data_points, 1) # Ensure that diagonal entries are numerically non-negative, as defined by neg_tol # TODO: Add support for symbolic case: Cannot use < operator with symbolic variables if F.sum(diag <= neg_tol) > 0: raise mx.base.MXNetError( " Matrix is not positive definite: negative diagonal elements" ) while num_iter <= max_iter_jitter: try: L = F.linalg.potrf( F.broadcast_add( matrix, F.broadcast_mul( F.eye(num_data_points, ctx=ctx, dtype=float_type), jitter, ), ) ) # gpu will not throw error but will store nans. If nan, L.sum() = nan and # L.nansum() computes the sum treating nans as zeros so the error tolerance can be large. # for axis = Null, nansum() and sum() will sum over all elements and return scalar array with shape (1,) # TODO: Add support for symbolic case: Cannot use <= operator with symbolic variables assert F.abs(L.nansum() - L.sum()) <= 1e-1 return L except: if num_iter == 0: # Initialize the jitter: constant jitter per each batch jitter = ( F.broadcast_mul(diag_mean, F.ones_like(jitter)) * diag_weight ) else: jitter = jitter * increase_jitter finally: num_iter += 1 raise mx.base.MXNetError( f" Matrix is not positive definite after the maximum number of iterations = {max_iter_jitter} " f"with a maximum jitter = {F.max(jitter)}" )
8,278
33.932489
116
py
rankpredictor
rankpredictor-master/sub/gluonts/support/util.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports import inspect import os import signal import tempfile import time from pathlib import Path from typing import Any, Callable, Dict, List, Optional, cast # Third-party imports import mxnet as mx # First-party imports from gluonts.core.serde import dump_json, load_json from gluonts.model.common import Tensor MXNET_HAS_ERF = hasattr(mx.nd, "erf") MXNET_HAS_ERFINV = hasattr(mx.nd, "erfinv") class Timer: """Context manager for measuring the time of enclosed code fragments.""" def __enter__(self): self.start = time.clock() self.interval = None return self def __exit__(self, *args): self.end = time.clock() self.interval = self.end - self.start class SignalHandler: """ A context manager that attaches a set of signal handlers within its scope. Parameters ---------- handlers_map A dictionary mapping signal numbers to associated signal handlers to be attached within the scope of the enclosing `SignalHandler` instance. """ Callback = Optional[Callable[[int, Any], None]] def __init__(self, handlers_map: Dict[int, Callback]) -> None: self.handlers_map = handlers_map def __enter__(self): self.default_handlers = { s: signal.signal(s, h) for s, h in self.handlers_map.items() } return self def __exit__(self, *args): for s, h in self.default_handlers.items(): signal.signal(s, h) class HybridContext: """ A context manager that ensures that an MXNet network is operating in a hybridized / not hybridized mode. Parameters ---------- net The network whose hybrid mode has to be modified within the enclosing context. hybridize A boolean flag inidicating whether the hybrid mode should be set or not. kwargs A dictionary of optional arguments to pass to the `hybridize()` call of the enclosed `HybridBlock` network. """ def __init__( self, net: mx.gluon.HybridBlock, hybridize: bool, data_batch: Optional[List[mx.nd.NDArray]] = None, **kwargs, ) -> None: self.net = net self.required_mode = hybridize self.original_mode = getattr(net, "_active", False) self.data_batch = data_batch self.kwargs = kwargs def __enter__(self): self.net.hybridize(active=self.required_mode, **self.kwargs) if self.data_batch is not None: self.net(*self.data_batch) def __exit__(self, *args): self.net.hybridize(active=self.original_mode, **self.kwargs) def copy_parameters( net_source: mx.gluon.Block, net_dest: mx.gluon.Block, ignore_extra: bool = False, allow_missing: bool = False, ) -> None: """ Copies parameters from one network to another. Parameters ---------- net_source Input network. net_dest Output network. ignore_extra Whether to ignore parameters from the source that are not present in the target. allow_missing Whether to allow additional parameters in the target not present in the source. """ with tempfile.TemporaryDirectory( prefix="gluonts-estimator-temp-" ) as model_dir: model_dir_path = str(Path(model_dir) / "tmp_model") net_source.save_parameters(model_dir_path) net_dest.load_parameters( model_dir_path, ctx=mx.current_context(), allow_missing=allow_missing, ignore_extra=ignore_extra, ) def get_hybrid_forward_input_names(hb: mx.gluon.HybridBlock): params = inspect.signature(hb.hybrid_forward).parameters param_names = list(params) assert param_names[0] == "F", ( f"Expected first argument of HybridBlock to be `F`, " f"but found `{param_names[0]}`" ) return param_names[1:] # skip: F def hybrid_block_to_symbol_block( hb: mx.gluon.HybridBlock, data_batch: List[mx.nd.NDArray] ) -> mx.gluon.SymbolBlock: """ Converts a Gluon `HybridBlock` to a `SymbolBlock`. Following the Gluon API, this is achieved by a `hybridize()` call on the passed `HybridBlock`, a single forward pass (using the provided data batch), and a combination of an `export()` and an `import()` calls of the input block. Note that MXNet has `problems with this method <https://github.com/apache/incubator-mxnet/issues/12783>`_. Parameters ---------- hb The Gluon `HybridBlock` to convert. data_batch Data to use for the forward pass after the `hybridize()` call. Returns ------- mx.gluon.SymbolBlock The resulting Gluon block backed by an MXNet symbol graph. """ with tempfile.TemporaryDirectory( prefix="gluonts-estimator-temp-" ) as model_dir: num_inputs = len(data_batch) model_dir_path = Path(model_dir) model_name = "gluonts-model" with HybridContext( net=hb, hybridize=True, data_batch=data_batch, static_alloc=True, static_shape=True, ): export_symb_block(hb, model_dir_path, model_name) sb = import_symb_block(num_inputs, model_dir_path, model_name) return sb def export_symb_block( hb: mx.gluon.HybridBlock, model_dir: Path, model_name: str, epoch: int = 0 ) -> None: """ Serializes a hybridized Gluon `HybridBlock`. Parameters ---------- hb The block to export. model_dir The path where the model will be saved. model_name The name identifying the model. epoch The epoch number, which together with the `model_name` identifies the model parameters. """ hb.export(path=str(model_dir / model_name), epoch=epoch) def import_symb_block( num_inputs: int, model_dir: Path, model_name: str, epoch: int = 0 ) -> mx.gluon.SymbolBlock: """ Deserializes a hybridized Gluon `HybridBlock` as a `SymbolBlock`. Parameters ---------- num_inputs The number of inputs of the serialized block. model_dir The path where the model is saved. model_name The name identifying the model. epoch The epoch number, which together with the `model_name` identifies the model parameters. Returns ------- mx.gluon.SymbolBlock The deserialized block. """ if num_inputs == 1: input_names = ["data"] else: input_names = [f"data{i}" for i in range(num_inputs)] # FIXME: mx.gluon.SymbolBlock cannot infer float_type and uses default np.float32 # FIXME: https://github.com/apache/incubator-mxnet/issues/11849 return mx.gluon.SymbolBlock.imports( symbol_file=str(model_dir / f"{model_name}-symbol.json"), input_names=input_names, param_file=str(model_dir / f"{model_name}-{epoch:04}.params"), ctx=mx.current_context(), ) def export_repr_block( rb: mx.gluon.HybridBlock, model_dir: Path, model_name: str, epoch: int = 0 ) -> None: """ Serializes a representable Gluon block. Parameters ---------- rb The block to export. model_dir The path where the model will be saved. model_name The name identifying the model. epoch The epoch number, which together with the `model_name` identifies the model parameters. """ with (model_dir / f"{model_name}-network.json").open("w") as fp: print(dump_json(rb), file=fp) rb.save_parameters(str(model_dir / f"{model_name}-{epoch:04}.params")) def import_repr_block( model_dir: Path, model_name: str, epoch: int = 0 ) -> mx.gluon.HybridBlock: """ Deserializes a representable Gluon block. Parameters ---------- model_dir The path where the model is saved. model_name The name identifying the model. epoch The epoch number, which together with the `model_name` identifies the model parameters. Returns ------- mx.gluon.HybridBlock: The deserialized block. """ with (model_dir / f"{model_name}-network.json").open("r") as fp: rb = cast(mx.gluon.HybridBlock, load_json(fp.read())) rb.load_parameters( str(model_dir / f"{model_name}-{epoch:04}.params"), ctx=mx.current_context(), allow_missing=False, ignore_extra=False, ) return rb def cumsum( F, x: Tensor, exclusive: bool = False, reverse: bool = False ) -> Tensor: r""" Find cumulative sum on the last axis by multiplying with lower triangular ones-matrix: .. math:: \operatorname{cumsum}(x) = \begin{cases} \operatorname{ltr\_ones} \times x & \text{for cumulative sum}\\ x \times \operatorname{ltr\_ones} & \text{for cumulative sum in the reverse order} \end{cases} Also supports `exclusive` flag to start the cumsum with zero. For example, if :math:`x = [a, b, c]`, we have .. math:: \operatorname{cumsum}(x) = \begin{cases} [a, a + b, a + b + c] & \text{if }\mathit{reverse = False, exclusive = False}\\ [0, a, a + b] & \text{if }\mathit{reverse = False, exclusive = True}\\ [a + b + c, b + c, c] & \text{if }\mathit{reverse = True, exclusive = False}\\ [b + c, c, 0] & \text{if }\mathit{reverse = True, exclusive = True}\\ \end{cases} Parameters ---------- F The function space to use. x A tensor with shape :math:`(..., n)`. exclusive If `True`, the cumulative sum starts with zero. reverse If `True`, the cumulative sum is performed in the opposite direction. Returns ------- Tensor: A modified tensor with identical shape and cumulative sums in the last axis. """ # Create a new axis (for matrix multiplication) either at last location or # last-but-one location (for reverse mode) exp_dim = -2 if reverse else -1 # (..., 1, n) if reverse is True and (..., n, 1) otherwise x = x.expand_dims(axis=exp_dim) # Ones_matrix (..., n, n) ones_matrix = F.linalg_gemm2( F.ones_like(x), F.ones_like(x), transpose_a=reverse, transpose_b=not reverse, ) cumulative_sum = F.linalg_trmm(ones_matrix, x, rightside=reverse) if exclusive: cumulative_sum = cumulative_sum - x return cumulative_sum.squeeze(axis=exp_dim) def weighted_average( F, x: Tensor, weights: Optional[Tensor] = None, axis=None ) -> Tensor: """ Computes the weighted average of a given tensor across a given axis. Parameters ---------- F The function space to use. x Input tensor, of which the average must be computed. weights Weights tensor, of the same shape as `x`. axis The axis along which to average `x` Returns ------- Tensor: The tensor with values averaged along the specified `axis`. """ if weights is not None: weighted_tensor = x * weights sum_weights = F.maximum(1.0, weights.sum(axis=axis)) return weighted_tensor.sum(axis=axis) / sum_weights else: return x.mean(axis=axis) def make_nd_diag(F, x: Tensor, d: int) -> Tensor: """ Make a diagonal tensor, given the diagonal Parameters ---------- F The function space to use. x Diagonal to use, shape :math:`(..., d)`. d Last dimension of `x`. Returns ------- Tensor A tensor y of shape :math:`(..., d, d)` such that :math:`y[..., i, i] = x[..., i]`. """ return F.broadcast_mul(F.eye(d), x.expand_dims(axis=-1)) def _broadcast_param(param, axes, sizes): for axis, size in zip(axes, sizes): param = param.expand_dims(axis=axis).broadcast_axes( axis=axis, size=size ) return param def erf(F, x: Tensor): if MXNET_HAS_ERF: return F.erf(x) # Using numerical recipes approximation for erf function # accurate to 1E-7 ones = x.ones_like() zeros = x.zeros_like() t = ones / (ones + 0.5 * x.abs()) coefficients = [ 1.00002368, 0.37409196, 0.09678418, -0.18628806, 0.27886807, -1.13520398, 1.48851587, -0.82215223, 0.17087277, ] inner = zeros for c in coefficients[::-1]: inner = t * (c + inner) res = ones - t * (inner - 1.26551223 - x.square()).exp() return F.where(F.broadcast_greater_equal(x, zeros), res, -1.0 * res) def erfinv(F, x: Tensor) -> Tensor: if MXNET_HAS_ERFINV: return F.erfinv(x) zeros = x.zeros_like() w = -F.log(F.broadcast_mul((1.0 - x), (1.0 + x))) mask_lesser = F.broadcast_lesser(w, zeros + 5.0) w = F.where(mask_lesser, w - 2.5, F.sqrt(w) - 3.0) coefficients_lesser = [ 2.81022636e-08, 3.43273939e-07, -3.5233877e-06, -4.39150654e-06, 0.00021858087, -0.00125372503, -0.00417768164, 0.246640727, 1.50140941, ] coefficients_greater_equal = [ -0.000200214257, 0.000100950558, 0.00134934322, -0.00367342844, 0.00573950773, -0.0076224613, 0.00943887047, 1.00167406, 2.83297682, ] p = F.where( mask_lesser, coefficients_lesser[0] + zeros, coefficients_greater_equal[0] + zeros, ) for c_l, c_ge in zip( coefficients_lesser[1:], coefficients_greater_equal[1:] ): c = F.where(mask_lesser, c_l + zeros, c_ge + zeros) p = c + F.broadcast_mul(p, w) return F.broadcast_mul(p, x) def get_download_path() -> Path: """ Returns ------- Path default path to download datasets or models of gluon-ts. The path is either $MXNET_HOME if the environment variable is defined or /home/username/.mxnet/gluon-ts/ """ return Path( os.environ.get("MXNET_HOME", str(Path.home() / ".mxnet" / "gluon-ts")) ) def map_dct_values(fn: Callable, dct: dict) -> dict: """Maps `fn` over a dicts values.""" return {key: fn(value) for key, value in dct.items()}
15,017
26.255898
85
py
rankpredictor
rankpredictor-master/sub/gluonts/gp/gaussian_process.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import List, Optional, Tuple # Third-party imports import mxnet as mx import numpy as np # First-party imports from gluonts.core.component import DType from gluonts.distribution import MultivariateGaussian from gluonts.distribution.distribution import getF from gluonts.kernels import Kernel from gluonts.model.common import Tensor from gluonts.support.linalg_util import ( batch_diagonal, jitter_cholesky, jitter_cholesky_eig, ) class GaussianProcess: # noinspection PyMethodOverriding, PyPep8Naming def __init__( self, sigma: Tensor, kernel: Kernel, prediction_length: Optional[int] = None, context_length: Optional[int] = None, num_samples: Optional[int] = None, ctx: mx.Context = mx.Context("cpu"), float_type: DType = np.float64, jitter_method: str = "iter", max_iter_jitter: int = 10, neg_tol: float = -1e-8, diag_weight: float = 1e-6, increase_jitter: int = 10, sample_noise: bool = True, F=None, ) -> None: r""" Parameters ---------- sigma Noise parameter of shape (batch_size, num_data_points, 1), where num_data_points is the number of rows in the Cholesky matrix. kernel Kernel object. prediction_length Prediction length. context_length Training length. num_samples The number of samples to be drawn. ctx Determines whether to compute on the cpu or gpu. float_type Determines whether to use single or double precision. jitter_method Iteratively jitter method or use eigenvalue decomposition depending on problem size. max_iter_jitter Maximum number of iterations for jitter to iteratively make the matrix positive definite. neg_tol Parameter in the jitter methods to eliminate eliminate matrices with diagonal elements smaller than this when checking if a matrix is positive definite. diag_weight Multiple of mean of diagonal entries to initialize the jitter. increase_jitter Each iteration multiply by jitter by this amount sample_noise Boolean to determine whether to add :math:`\sigma^2I` to the predictive covariance matrix. F A module that can either refer to the Symbol API or the NDArray API in MXNet. """ assert ( prediction_length is None or prediction_length > 0 ), "The value of `prediction_length` should be > 0" assert ( context_length is None or context_length > 0 ), "The value of `context_length` should be > 0" assert ( num_samples is None or num_samples > 0 ), "The value of `num_samples` should be > 0" self.sigma = sigma self.kernel = kernel self.prediction_length = prediction_length self.context_length = ( context_length if context_length is not None else prediction_length ) self.num_samples = num_samples self.F = F if F else getF(sigma) self.ctx = ctx self.float_type = float_type self.jitter_method = jitter_method self.max_iter_jitter = max_iter_jitter self.neg_tol = neg_tol self.diag_weight = diag_weight self.increase_jitter = increase_jitter self.sample_noise = sample_noise # noinspection PyMethodOverriding,PyPep8Naming def _compute_cholesky_gp( self, kernel_matrix: Tensor, num_data_points: Optional[int] = None, noise: bool = True, ) -> Tensor: r""" Parameters -------------------- kernel_matrix Kernel matrix of shape (batch_size, num_data_points, num_data_points). num_data_points Number of rows in the kernel_matrix. noise Boolean to determine whether to add :math:`\sigma^2I` to the kernel matrix. This is used in the predictive step if you would like to sample the predictive covariance matrix without noise. It is set to True in every other case. Returns -------------------- Tensor Cholesky factor :math:`L` of the kernel matrix with added noise :math:`LL^T = K + \sigma^2 I` of shape (batch_size, num_data_points, num_data_points). """ if noise: # Add sigma kernel_matrix = self.F.broadcast_plus( kernel_matrix, self.F.broadcast_mul( self.sigma ** 2, self.F.eye( num_data_points, ctx=self.ctx, dtype=self.float_type ), ), ) # Warning: This method is more expensive than the iterative jitter # but it works for mx.sym if self.jitter_method == "eig": return jitter_cholesky_eig( self.F, kernel_matrix, num_data_points, self.ctx, self.float_type, self.diag_weight, ) elif self.jitter_method == "iter" and self.F is mx.nd: return jitter_cholesky( self.F, kernel_matrix, num_data_points, self.ctx, self.float_type, self.max_iter_jitter, self.neg_tol, self.diag_weight, self.increase_jitter, ) else: return self.F.linalg.potrf(kernel_matrix) def log_prob(self, x_train: Tensor, y_train: Tensor) -> Tensor: r""" This method computes the negative marginal log likelihood .. math:: :nowrap: \begin{aligned} \frac{1}{2} [d \log(2\pi) + \log(|K|) + y^TK^{-1}y], \end{aligned} where :math:`d` is the number of data points. This can be written in terms of the Cholesky factor :math:`L` as .. math:: :nowrap: \begin{aligned} \log(|K|) = \log(|LL^T|) &= \log(|L||L|^T) = \log(|L|^2) = 2\log(|L|) \\ &= 2\log\big(\prod_i^n L_{ii}\big) = 2 \sum_i^N \log(L_{ii}) \end{aligned} and .. math:: :nowrap: \begin{aligned} y^TK^{-1}y = (y^TL^{-T})(L^{-1}y) = (L^{-1}y)^T(L^{-1}y) = ||L^{-1}y||_2^2. \end{aligned} Parameters -------------------- x_train Training set of features of shape (batch_size, context_length, num_features). y_train Training labels of shape (batch_size, context_length). Returns -------------------- Tensor The negative log marginal likelihood of shape (batch_size,) """ assert ( self.context_length is not None ), "The value of `context_length` must be set." return -MultivariateGaussian( self.F.zeros_like(y_train), # 0 mean gaussian process prior self._compute_cholesky_gp( self.kernel.kernel_matrix(x_train, x_train), self.context_length, ), ).log_prob(y_train) def sample(self, mean: Tensor, covariance: Tensor) -> Tensor: r""" Parameters ---------- covariance The covariance matrix of the GP of shape (batch_size, prediction_length, prediction_length). mean The mean vector of the GP of shape (batch_size, prediction_length). Returns ------- Tensor Samples from a Gaussian Process of shape (batch_size, prediction_length, num_samples), where :math:`L` is the matrix square root, Cholesky Factor of the covariance matrix with the added noise tolerance on the diagonal, :math:`Lz`, where :math:`z \sim N(0,I)` and assumes the mean is zero. """ assert ( self.num_samples is not None ), "The value of `num_samples` must be set." assert ( self.prediction_length is not None ), "The value of `prediction_length` must be set." samples = MultivariateGaussian( mean, self._compute_cholesky_gp( covariance, self.prediction_length, self.sample_noise ), ).sample_rep( self.num_samples, dtype=self.float_type ) # Shape (num_samples, batch_size, prediction_length) return self.F.transpose(samples, axes=(1, 2, 0)) # noinspection PyMethodOverriding,PyPep8Naming def exact_inference( self, x_train: Tensor, y_train: Tensor, x_test: Tensor ) -> Tuple[Tensor, Tensor, Tensor]: """ Parameters ---------- x_train Training set of features of shape (batch_size, context_length, num_features). y_train Training labels of shape (batch_size, context_length). x_test Test set of features of shape (batch_size, prediction_length, num_features). Returns ------- Tuple Tensor Predictive GP samples of shape (batch_size, prediction_length, num_samples). Tensor Predictive mean of the GP of shape (batch_size, prediction_length). Tensor Predictive standard deviation of the GP of shape (batch_size, prediction_length). """ assert ( self.context_length is not None ), "The value of `context_length` must be set." assert ( self.prediction_length is not None ), "The value of `prediction_length` must be set." # Compute Cholesky factorization of training kernel matrix l_train = self._compute_cholesky_gp( self.kernel.kernel_matrix(x_train, x_train), self.context_length ) lower_tri_solve = self.F.linalg.trsm( l_train, self.kernel.kernel_matrix(x_train, x_test) ) predictive_mean = self.F.linalg.gemm2( lower_tri_solve, self.F.linalg.trsm(l_train, y_train.expand_dims(axis=-1)), transpose_a=True, ).squeeze(axis=-1) # Can rewrite second term as # :math:`||L^-1 * K(x_train,x_test||_2^2` # and only solve 1 equation predictive_covariance = self.kernel.kernel_matrix( x_test, x_test ) - self.F.linalg.gemm2( lower_tri_solve, lower_tri_solve, transpose_a=True ) # Extract diagonal entries of covariance matrix predictive_std = batch_diagonal( self.F, predictive_covariance, self.prediction_length, self.ctx, self.float_type, ) # If self.sample_noise = True, predictive covariance has sigma^2 on the diagonal if self.sample_noise: predictive_std = self.F.broadcast_add( predictive_std, self.sigma ** 2 ) predictive_std = self.F.sqrt(predictive_std).squeeze(axis=-1) # Compute sample from GP predictive distribution return ( self.sample(predictive_mean, predictive_covariance), predictive_mean, predictive_std, ) @staticmethod def plot( ts_idx: int, x_train: Optional[Tensor] = None, y_train: Optional[Tensor] = None, x_test: Optional[Tensor] = None, mean: Optional[Tensor] = None, std: Optional[Tensor] = None, samples: Optional[Tensor] = None, axis: Optional[List] = None, ) -> None: """ This method plots the sampled GP distribution at the test points in solid colors, as well as the predictive mean as the dashed red line. Plus and minus 2 predictive standard deviations are shown in the grey region. The training points are shown as the blue dots. Parameters ---------- ts_idx Time series index to plot x_train Training set of features of shape (batch_size, context_length, num_features). y_train Training labels of shape (batch_size, context_length). x_test Test set of features of shape (batch_size, prediction_length, num_features). mean Mean of the GP of shape (batch_size, prediction_length). std Standard deviation of the GP of shape (batch_size, prediction_length, 1). samples GP samples of shape (batch_size, prediction_length, num_samples). axis Plot axes limits """ # matplotlib==2.0.* gives errors in Brazil builds and has to be # imported locally import matplotlib.pyplot as plt if x_train is not None: x_train = x_train[ts_idx, :, :].asnumpy() if y_train is not None: y_train = y_train[ts_idx, :].asnumpy() plt.plot(x_train, y_train, "bs", ms=8) if x_test is not None: x_test = x_test[ts_idx, :, :].asnumpy() if samples is not None: samples = samples[ts_idx, :, :].asnumpy() plt.plot(x_test, samples) if mean is not None: mean = mean[ts_idx, :].asnumpy() plt.plot(x_test, mean, "r--", lw=2) if std is not None: std = std[ts_idx, :].asnumpy() plt.gca().fill_between( x_test.flat, mean - 2 * std, mean + 2 * std, color="#dddddd", ) if axis is not None: plt.axis(axis) plt.title(f"Samples from GP for time series {ts_idx}") plt.show()
14,687
36.279188
117
py
rankpredictor
rankpredictor-master/sub/gluonts/model/forecast-raw.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports import re from enum import Enum from typing import Dict, List, NamedTuple, Optional, Set, Union, Callable # Third-party imports import mxnet as mx import numpy as np import pandas as pd import pydantic # First-party imports from gluonts.core.exception import GluonTSUserError from gluonts.distribution import Distribution from gluonts.core.component import validated class Quantile(NamedTuple): value: float name: str @property def loss_name(self): return f"QuantileLoss[{self.name}]" @property def weighted_loss_name(self): return f"wQuantileLoss[{self.name}]" @property def coverage_name(self): return f"Coverage[{self.name}]" @classmethod def checked(cls, value: float, name: str) -> "Quantile": if not 0 <= value <= 1: raise GluonTSUserError( f"quantile value should be in [0, 1] but found {value}" ) return Quantile(value, name) @classmethod def from_float(cls, quantile: float) -> "Quantile": assert isinstance(quantile, float) return cls.checked(value=quantile, name=str(quantile)) @classmethod def from_str(cls, quantile: str) -> "Quantile": assert isinstance(quantile, str) try: return cls.checked(value=float(quantile), name=quantile) except ValueError: m = re.match(r"^p(\d{2})$", quantile) if m is None: raise GluonTSUserError( "Quantile string should be of the form " f'"p10", "p50", ... or "0.1", "0.5", ... but found {quantile}' ) else: quantile: float = int(m.group(1)) / 100 return cls(value=quantile, name=str(quantile)) @classmethod def parse(cls, quantile: Union["Quantile", float, str]) -> "Quantile": """Produces equivalent float and string representation of a given quantile level. >>> Quantile.parse(0.1) Quantile(value=0.1, name='0.1') >>> Quantile.parse('0.2') Quantile(value=0.2, name='0.2') >>> Quantile.parse('0.20') Quantile(value=0.2, name='0.20') >>> Quantile.parse('p99') Quantile(value=0.99, name='0.99') Parameters ---------- quantile Quantile, can be a float a str representing a float e.g. '0.1' or a quantile string of the form 'p0.1'. Returns ------- Quantile A tuple containing both a float and a string representation of the input quantile level. """ if isinstance(quantile, Quantile): return quantile elif isinstance(quantile, float): return cls.from_float(quantile) else: return cls.from_str(quantile) class Forecast: """ A abstract class representing predictions. """ start_date: pd.Timestamp freq: str item_id: Optional[str] info: Optional[Dict] prediction_length: int mean: np.ndarray _index = None def quantile(self, q: Union[float, str]) -> np.ndarray: """ Computes a quantile from the predicted distribution. Parameters ---------- q Quantile to compute. Returns ------- numpy.ndarray Value of the quantile across the prediction range. """ raise NotImplementedError() @property def median(self) -> np.ndarray: return self.quantile(0.5) def plot( self, prediction_intervals=(50.0, 90.0), show_mean=False, color="b", label=None, output_file=None, *args, **kwargs, ): """ Plots the median of the forecast as well as confidence bounds. (requires matplotlib and pandas). Parameters ---------- prediction_intervals : float or list of floats in [0, 100] Confidence interval size(s). If a list, it will stack the error plots for each confidence interval. Only relevant for error styles with "ci" in the name. show_mean : boolean Whether to also show the mean of the forecast. color : matplotlib color name or dictionary The color used for plotting the forecast. label : string A label (prefix) that is used for the forecast output_file : str or None, default None Output path for the plot file. If None, plot is not saved to file. args : Other arguments are passed to main plot() call kwargs : Other keyword arguments are passed to main plot() call """ # matplotlib==2.0.* gives errors in Brazil builds and has to be # imported locally import matplotlib.pyplot as plt label_prefix = "" if label is None else label + "-" for c in prediction_intervals: assert 0.0 <= c <= 100.0 ps = [50.0] + [ 50.0 + f * c / 2.0 for c in prediction_intervals for f in [-1.0, +1.0] ] percentiles_sorted = sorted(set(ps)) def alpha_for_percentile(p): return (p / 100.0) ** 0.3 ps_data = [self.quantile(p / 100.0) for p in percentiles_sorted] i_p50 = len(percentiles_sorted) // 2 p50_data = ps_data[i_p50] p50_series = pd.Series(data=p50_data, index=self.index) p50_series.plot(color=color, ls="-", label=f"{label_prefix}median") if show_mean: mean_data = np.mean(self._sorted_samples, axis=0) pd.Series(data=mean_data, index=self.index).plot( color=color, ls=":", label=f"{label_prefix}mean", *args, **kwargs, ) for i in range(len(percentiles_sorted) // 2): ptile = percentiles_sorted[i] alpha = alpha_for_percentile(ptile) plt.fill_between( self.index, ps_data[i], ps_data[-i - 1], facecolor=color, alpha=alpha, interpolate=True, *args, **kwargs, ) # Hack to create labels for the error intervals. # Doesn't actually plot anything, because we only pass a single data point pd.Series(data=p50_data[:1], index=self.index[:1]).plot( color=color, alpha=alpha, linewidth=10, label=f"{label_prefix}{100 - ptile * 2}%", *args, **kwargs, ) if output_file: plt.savefig(output_file) @property def index(self) -> pd.DatetimeIndex: if self._index is None: self._index = pd.date_range( self.start_date, periods=self.prediction_length, freq=self.freq ) return self._index def dim(self) -> int: """ Returns the dimensionality of the forecast object. """ raise NotImplementedError() def copy_dim(self, dim: int): """ Returns a new Forecast object with only the selected sub-dimension. Parameters ---------- dim The returned forecast object will only represent this dimension. """ raise NotImplementedError() def copy_aggregate(self, agg_fun: Callable): """ Returns a new Forecast object with a time series aggregated over the dimension axis. Parameters ---------- agg_fun Aggregation function that defines the aggregation operation (typically mean or sum). """ raise NotImplementedError() def as_json_dict(self, config: "Config") -> dict: result = {} if OutputType.mean in config.output_types: result["mean"] = self.mean.tolist() if OutputType.quantiles in config.output_types: quantiles = map(Quantile.parse, config.quantiles) result["quantiles"] = { quantile.name: self.quantile(quantile.value).tolist() for quantile in quantiles } if OutputType.samples in config.output_types: result["samples"] = [] return result class SampleForecast(Forecast): """ A `Forecast` object, where the predicted distribution is represented internally as samples. Parameters ---------- samples Array of size (num_samples, prediction_length) (1D case) or (num_samples, prediction_length, target_dim) (multivariate case) start_date start of the forecast freq forecast frequency info additional information that the forecaster may provide e.g. estimated parameters, number of iterations ran etc. """ @validated() def __init__( self, samples: Union[mx.nd.NDArray, np.ndarray], start_date, freq, item_id: Optional[str] = None, info: Optional[Dict] = None, ): assert isinstance( samples, (np.ndarray, mx.ndarray.ndarray.NDArray) ), "samples should be either a numpy or an mxnet array" assert ( len(np.shape(samples)) == 2 or len(np.shape(samples)) == 3 ), "samples should be a 2-dimensional or 3-dimensional array. Dimensions found: {}".format( len(np.shape(samples)) ) self.samples = ( samples if (isinstance(samples, np.ndarray)) else samples.asnumpy() ) self._sorted_samples_value = None self._mean = None self._dim = None self.item_id = item_id self.info = info assert isinstance( start_date, pd.Timestamp ), "start_date should be a pandas Timestamp object" self.start_date = start_date assert isinstance(freq, str), "freq should be a string" self.freq = freq @property def _sorted_samples(self): if self._sorted_samples_value is None: self._sorted_samples_value = np.sort(self.samples, axis=0) return self._sorted_samples_value @property def num_samples(self): """ The number of samples representing the forecast. """ return self.samples.shape[0] @property def prediction_length(self): """ Time length of the forecast. """ return self.samples.shape[-1] @property def mean(self): """ Forecast mean. """ if self._mean is not None: return self._mean else: return np.mean(self.samples, axis=0) @property def mean_ts(self): """ Forecast mean, as a pandas.Series object. """ return pd.Series(self.index, self.mean) def quantile(self, q): q = Quantile.parse(q).value sample_idx = int(np.round((self.num_samples - 1) * q)) return self._sorted_samples[sample_idx, :] def copy_dim(self, dim: int): if len(self.samples.shape) == 2: samples = self.samples else: target_dim = self.samples.shape[2] assert dim < target_dim, ( f"must set 0 <= dim < target_dim, but got dim={dim}," f" target_dim={target_dim}" ) samples = self.samples[:, :, dim] return SampleForecast( samples=samples, start_date=self.start_date, freq=self.freq, item_id=self.item_id, info=self.info, ) def copy_aggregate(self, agg_fun: Callable): if len(self.samples.shape) == 2: samples = self.samples else: # Aggregate over target dimension axis samples = agg_fun(self.samples, axis=2) return SampleForecast( samples=samples, start_date=self.start_date, freq=self.freq, item_id=self.item_id, info=self.info, ) def dim(self) -> int: if self._dim is not None: return self._dim else: if len(self.samples.shape) == 2: # univariate target # shape: (num_samples, prediction_length) return 1 else: # multivariate target # shape: (num_samples, prediction_length, target_dim) return self.samples.shape[2] def as_json_dict(self, config: "Config") -> dict: result = super().as_json_dict(config) if OutputType.samples in config.output_types: result["samples"] = self.samples.tolist() return result def __repr__(self): return ", ".join( [ f"SampleForecast({self.samples!r})", f"{self.start_date!r}", f"{self.freq!r}", f"item_id={self.item_id!r}", f"info={self.info!r})", ] ) class QuantileForecast(Forecast): """ A Forecast that contains arrays (i.e. time series) for quantiles and mean Parameters ---------- forecast_arrays An array of forecasts start_date start of the forecast freq forecast frequency forecast_keys A list of quantiles of the form '0.1', '0.9', etc., and potentially 'mean'. Each entry corresponds to one array in forecast_arrays. info additional information that the forecaster may provide e.g. estimated parameters, number of iterations ran etc. """ def __init__( self, forecast_arrays: np.ndarray, start_date: pd.Timestamp, freq: str, forecast_keys: List[str], item_id: Optional[str] = None, info: Optional[Dict] = None, ): self.forecast_array = forecast_arrays self.start_date = pd.Timestamp(start_date, freq=freq) self.freq = freq # normalize keys self.forecast_keys = [ Quantile.from_str(key).name if key != "mean" else key for key in forecast_keys ] self.item_id = item_id self.info = info self._dim = None shape = self.forecast_array.shape assert shape[0] == len(self.forecast_keys), ( f"The forecast_array (shape={shape} should have the same " f"length as the forecast_keys (len={len(self.forecast_keys)})." ) self.prediction_length = shape[-1] self._forecast_dict = { k: self.forecast_array[i] for i, k in enumerate(self.forecast_keys) } self._nan_out = np.array([np.nan] * self.prediction_length) def quantile(self, q: Union[float, str]) -> np.ndarray: q_str = Quantile.parse(q).name # We return nan here such that evaluation runs through return self._forecast_dict.get(q_str, self._nan_out) @property def mean(self): """ Forecast mean. """ return self._forecast_dict.get("mean", self._nan_out) def dim(self) -> int: if self._dim is not None: return self._dim else: if ( len(self.forecast_array.shape) == 2 ): # 1D target. shape: (num_samples, prediction_length) return 1 else: return self.forecast_array.shape[ 1 ] # 2D target. shape: (num_samples, target_dim, prediction_length) def __repr__(self): return ", ".join( [ f"QuantileForecast({self.forecast_array!r})", f"start_date={self.start_date!r}", f"freq={self.freq!r}", f"forecast_keys={self.forecast_keys!r}", f"item_id={self.item_id!r}", f"info={self.info!r})", ] ) class DistributionForecast(Forecast): """ A `Forecast` object that uses a GluonTS distribution directly. This can for instance be used to represent marginal probability distributions for each time point -- although joint distributions are also possible, e.g. when using MultiVariateGaussian). Parameters ---------- distribution Distribution object. This should represent the entire prediction length, i.e., if we draw `num_samples` samples from the distribution, the sample shape should be samples = trans_dist.sample(num_samples) samples.shape -> (num_samples, prediction_length) start_date start of the forecast freq forecast frequency info additional information that the forecaster may provide e.g. estimated parameters, number of iterations ran etc. """ @validated() def __init__( self, distribution: Distribution, start_date, freq, item_id: Optional[str] = None, info: Optional[Dict] = None, ): self.distribution = distribution self.shape = ( self.distribution.batch_shape + self.distribution.event_shape ) self.prediction_length = self.shape[0] self.item_id = item_id self.info = info assert isinstance( start_date, pd.Timestamp ), "start_date should be a pandas Timestamp object" self.start_date = start_date assert isinstance(freq, str), "freq should be a string" self.freq = freq self._mean = None @property def mean(self): """ Forecast mean. """ if self._mean is not None: return self._mean else: self._mean = self.distribution.mean.asnumpy() return self._mean @property def mean_ts(self): """ Forecast mean, as a pandas.Series object. """ return pd.Series(self.index, self.mean) def quantile(self, level): level = Quantile.parse(level).value q = self.distribution.quantile(mx.nd.array([level])).asnumpy()[0] return q def to_sample_forecast(self, num_samples: int = 200) -> SampleForecast: return SampleForecast( samples=self.distribution.sample(num_samples), start_date=self.start_date, freq=self.freq, item_id=self.item_id, info=self.info, ) class OutputType(str, Enum): mean = "mean" samples = "samples" quantiles = "quantiles" class Config(pydantic.BaseModel): num_samples: int = pydantic.Field(100, alias="num_eval_samples") output_types: Set[OutputType] = {"quantiles", "mean"} # FIXME: validate list elements quantiles: List[str] = ["0.1", "0.5", "0.9"] class Config: allow_population_by_field_name = True # store additional fields extra = "allow"
19,665
28.933029
99
py
rankpredictor
rankpredictor-master/sub/gluonts/model/predictor.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports import functools import itertools import logging import multiprocessing as mp import sys import traceback from pathlib import Path from pydoc import locate from tempfile import TemporaryDirectory import json from typing import ( TYPE_CHECKING, Tuple, Union, Any, Callable, Dict, Iterator, List, Optional, Type, ) # Third-party imports import mxnet as mx import numpy as np # First-party imports import gluonts from gluonts.distribution import Distribution, DistributionOutput from gluonts.core.component import ( DType, equals, from_hyperparameters, get_mxnet_context, validated, ) from gluonts.core.exception import GluonTSException from gluonts.core.serde import dump_json, fqname_for, load_json from gluonts.dataset.common import DataEntry, Dataset, ListDataset from .forecast_generator import ForecastGenerator, SampleForecastGenerator from gluonts.dataset.loader import DataBatch, InferenceDataLoader from gluonts.model.forecast import Forecast from gluonts.support.util import ( export_repr_block, export_symb_block, get_hybrid_forward_input_names, hybrid_block_to_symbol_block, import_repr_block, import_symb_block, ) from gluonts.transform import Transformation if TYPE_CHECKING: # avoid circular import from gluonts.model.estimator import Estimator # noqa OutputTransform = Callable[[DataEntry, np.ndarray], np.ndarray] class Predictor: """ Abstract class representing predictor objects. Parameters ---------- prediction_length Prediction horizon. freq Frequency of the predicted data. """ __version__: str = gluonts.__version__ def __init__(self, prediction_length: int, freq: str) -> None: assert ( prediction_length > 0 ), "The value of `prediction_length` should be > 0" self.prediction_length = prediction_length self.freq = freq def predict(self, dataset: Dataset, **kwargs) -> Iterator[Forecast]: """ Compute forecasts for the time series in the provided dataset. This method is not implemented in this abstract class; please use one of the subclasses. Parameters ---------- dataset The dataset containing the time series to predict. Returns ------- Iterator[Forecast] Iterator over the forecasts, in the same order as the dataset iterable was provided. """ raise NotImplementedError def serialize(self, path: Path) -> None: # serialize Predictor type with (path / "type.txt").open("w") as fp: fp.write(fqname_for(self.__class__)) with (path / "version.json").open("w") as fp: json.dump( {"model": self.__version__, "gluonts": gluonts.__version__}, fp ) @classmethod def deserialize( cls, path: Path, ctx: Optional[mx.Context] = None ) -> "Predictor": """ Load a serialized predictor from the given path Parameters ---------- path Path to the serialized files predictor. ctx Optional mxnet context to be used with the predictor. If nothing is passed will use the GPU if available and CPU otherwise. """ # deserialize Predictor type with (path / "type.txt").open("r") as fp: tpe = locate(fp.readline()) # ensure that predictor_cls is a subtype of Predictor if not issubclass(tpe, Predictor): raise IOError( f"Class {fqname_for(tpe)} is not " f"a subclass of {fqname_for(Predictor)}" ) # call deserialize() for the concrete Predictor type return tpe.deserialize(path, ctx) @classmethod def from_hyperparameters(cls, **hyperparameters): return from_hyperparameters(cls, **hyperparameters) class RepresentablePredictor(Predictor): """ An abstract predictor that can be subclassed by models that are not based on Gluon. Subclasses should have @validated() constructors. (De)serialization and value equality are all implemented on top of the @validated() logic. Parameters ---------- prediction_length Prediction horizon. freq Frequency of the predicted data. """ @validated() def __init__(self, prediction_length: int, freq: str) -> None: super().__init__(prediction_length, freq) def predict(self, dataset: Dataset, **kwargs) -> Iterator[Forecast]: for item in dataset: yield self.predict_item(item) def predict_item(self, item: DataEntry) -> Forecast: raise NotImplementedError def __eq__(self, that): """ Two RepresentablePredictor instances are considered equal if they have the same constructor arguments. """ return equals(self, that) def serialize(self, path: Path) -> None: # call Predictor.serialize() in order to serialize the class name super().serialize(path) with (path / "predictor.json").open("w") as fp: print(dump_json(self), file=fp) @classmethod def deserialize( cls, path: Path, ctx: Optional[mx.Context] = None ) -> "RepresentablePredictor": with (path / "predictor.json").open("r") as fp: return load_json(fp.read()) class GluonPredictor(Predictor): """ Base predictor type for Gluon-based models. Parameters ---------- input_names Input tensor names for the graph prediction_net Network that will be called for prediction batch_size Number of time series to predict in a single batch prediction_length Number of time steps to predict freq Frequency of the input data input_transform Input transformation pipeline output_transform Output transformation ctx MXNet context to use for computation forecast_generator Class to generate forecasts from network ouputs """ BlockType = mx.gluon.Block def __init__( self, input_names: List[str], prediction_net: BlockType, batch_size: int, prediction_length: int, freq: str, ctx: mx.Context, input_transform: Transformation, forecast_generator: ForecastGenerator = SampleForecastGenerator(), output_transform: Optional[OutputTransform] = None, dtype: DType = np.float32, ) -> None: super().__init__(prediction_length, freq) self.input_names = input_names self.prediction_net = prediction_net self.batch_size = batch_size self.input_transform = input_transform self.forecast_generator = forecast_generator self.output_transform = output_transform self.ctx = ctx self.dtype = dtype def hybridize(self, batch: DataBatch) -> None: """ Hybridizes the underlying prediction network. Parameters ---------- batch A batch of data to use for the required forward pass after the `hybridize()` call. """ self.prediction_net.hybridize(active=True) self.prediction_net(*[batch[k] for k in self.input_names]) def as_symbol_block_predictor( self, batch: DataBatch ) -> "SymbolBlockPredictor": """ Returns a variant of the current :class:`GluonPredictor` backed by a Gluon `SymbolBlock`. If the current predictor is already a :class:`SymbolBlockPredictor`, it just returns itself. Parameters ---------- batch A batch of data to use for the required forward pass after the `hybridize()` call of the underlying network. Returns ------- SymbolBlockPredictor A predictor derived from the current one backed by a `SymbolBlock`. """ raise NotImplementedError def predict( self, dataset: Dataset, num_samples: Optional[int] = None ) -> Iterator[Forecast]: #print('predict') inference_data_loader = InferenceDataLoader( dataset, self.input_transform, self.batch_size, ctx=self.ctx, dtype=self.dtype, ) yield from self.forecast_generator( inference_data_loader=inference_data_loader, prediction_net=self.prediction_net, input_names=self.input_names, freq=self.freq, output_transform=self.output_transform, num_samples=num_samples, ) def __eq__(self, that): if type(self) != type(that): return False # TODO: also consider equality of the pipelines # if not equals(self.input_transform, that.input_transform): # return False return equals( self.prediction_net.collect_params(), that.prediction_net.collect_params(), ) def serialize(self, path: Path) -> None: # call Predictor.serialize() in order to serialize the class name super().serialize(path) # serialize every GluonPredictor-specific parameters # serialize the prediction network self.serialize_prediction_net(path) # serialize transformation chain with (path / "input_transform.json").open("w") as fp: print(dump_json(self.input_transform), file=fp) # FIXME: also needs to serialize the output_transform # serialize all remaining constructor parameters with (path / "parameters.json").open("w") as fp: parameters = dict( batch_size=self.batch_size, prediction_length=self.prediction_length, freq=self.freq, ctx=self.ctx, dtype=self.dtype, forecast_generator=self.forecast_generator, input_names=self.input_names, ) print(dump_json(parameters), file=fp) def serialize_prediction_net(self, path: Path) -> None: raise NotImplementedError() class SymbolBlockPredictor(GluonPredictor): """ A predictor which serializes the network structure as an MXNet symbolic graph. Should be used for models deployed in production in order to ensure forward-compatibility as GluonTS models evolve. Used by the training shell if training is invoked with a hyperparameter `use_symbol_block_predictor = True`. """ BlockType = mx.gluon.SymbolBlock def as_symbol_block_predictor( self, batch: DataBatch ) -> "SymbolBlockPredictor": return self def serialize_prediction_net(self, path: Path) -> None: export_symb_block(self.prediction_net, path, "prediction_net") @classmethod def deserialize( cls, path: Path, ctx: Optional[mx.Context] = None ) -> "SymbolBlockPredictor": ctx = ctx if ctx is not None else get_mxnet_context() with mx.Context(ctx): # deserialize constructor parameters with (path / "parameters.json").open("r") as fp: parameters = load_json(fp.read()) parameters["ctx"] = ctx # deserialize transformation chain with (path / "input_transform.json").open("r") as fp: transform = load_json(fp.read()) # deserialize prediction network num_inputs = len(parameters["input_names"]) prediction_net = import_symb_block( num_inputs, path, "prediction_net" ) return SymbolBlockPredictor( input_transform=transform, prediction_net=prediction_net, **parameters, ) class RepresentableBlockPredictor(GluonPredictor): """ A predictor which serializes the network structure using the JSON-serialization methods located in `gluonts.core.serde`. Use the following logic to create a `RepresentableBlockPredictor` from a trained prediction network. >>> def create_representable_block_predictor( ... prediction_network: mx.gluon.HybridBlock, ... **kwargs ... ) -> RepresentableBlockPredictor: ... return RepresentableBlockPredictor( ... prediction_net=prediction_network, ... **kwargs ... ) """ BlockType = mx.gluon.HybridBlock def __init__( self, prediction_net: BlockType, batch_size: int, prediction_length: int, freq: str, ctx: mx.Context, input_transform: Transformation, forecast_generator: ForecastGenerator = SampleForecastGenerator(), output_transform: Optional[ Callable[[DataEntry, np.ndarray], np.ndarray] ] = None, dtype: DType = np.float32, ) -> None: super().__init__( input_names=get_hybrid_forward_input_names(prediction_net), prediction_net=prediction_net, batch_size=batch_size, prediction_length=prediction_length, freq=freq, ctx=ctx, input_transform=input_transform, forecast_generator=forecast_generator, output_transform=output_transform, dtype=dtype, ) def as_symbol_block_predictor( self, batch: DataBatch ) -> SymbolBlockPredictor: symbol_block_net = hybrid_block_to_symbol_block( hb=self.prediction_net, data_batch=[batch[k] for k in self.input_names], ) return SymbolBlockPredictor( input_names=self.input_names, prediction_net=symbol_block_net, batch_size=self.batch_size, prediction_length=self.prediction_length, freq=self.freq, ctx=self.ctx, input_transform=self.input_transform, forecast_generator=self.forecast_generator, output_transform=self.output_transform, dtype=self.dtype, ) def serialize(self, path: Path) -> None: logging.warning( "Serializing RepresentableBlockPredictor instances does not save " "the prediction network structure in a backwards-compatible " "manner. Be careful not to use this method in production." ) super().serialize(path) def serialize_prediction_net(self, path: Path) -> None: export_repr_block(self.prediction_net, path, "prediction_net") @classmethod def deserialize( cls, path: Path, ctx: Optional[mx.Context] = None ) -> "RepresentableBlockPredictor": ctx = ctx if ctx is not None else get_mxnet_context() with mx.Context(ctx): # deserialize constructor parameters with (path / "parameters.json").open("r") as fp: parameters = load_json(fp.read()) # deserialize transformation chain with (path / "input_transform.json").open("r") as fp: transform = load_json(fp.read()) # deserialize prediction network prediction_net = import_repr_block(path, "prediction_net") # input_names is derived from the prediction_net if "input_names" in parameters: del parameters["input_names"] parameters["ctx"] = ctx return RepresentableBlockPredictor( input_transform=transform, prediction_net=prediction_net, **parameters, ) class WorkerError: def __init__(self, msg): self.msg = msg def _worker_loop( predictor_path: Path, input_queue: mp.Queue, output_queue: mp.Queue, worker_id, **kwargs, ): """ Worker loop for multiprocessing Predictor. Loads the predictor serialized in predictor_path reads inputs from input_queue and writes forecasts to output_queue """ predictor = Predictor.deserialize(predictor_path) while True: idx, data_chunk = input_queue.get() if idx is None: output_queue.put((None, None, None)) break try: result = list(predictor.predict(data_chunk, **kwargs)) except Exception: we = WorkerError( "".join(traceback.format_exception(*sys.exc_info())) ) output_queue.put((we, None, None)) break output_queue.put((idx, worker_id, result)) class ParallelizedPredictor(Predictor): """ Runs multiple instances (workers) of a predictor in parallel. Exceptions are propagated from the workers. Note: That there is currently an issue with tqdm that will cause things to hang if the ParallelizedPredictor is used with tqdm and an exception occurs during prediction. https://github.com/tqdm/tqdm/issues/548 Parameters ---------- base_predictor A representable predictor that will be used num_workers Number of workers (processes) to use. If set to None, one worker per CPU will be used. chunk_size Number of items to pass per call """ def __init__( self, base_predictor: Predictor, num_workers: Optional[int] = None, chunk_size=1, ) -> None: super().__init__(base_predictor.prediction_length, base_predictor.freq) self._base_predictor = base_predictor self._num_workers = ( num_workers if num_workers is not None else mp.cpu_count() ) self._chunk_size = chunk_size self._num_running_workers = 0 self._input_queues = [] self._output_queue = None def _grouper(self, iterable, n): iterator = iter(iterable) group = tuple(itertools.islice(iterator, n)) while group: yield group group = tuple(itertools.islice(iterator, n)) def terminate(self): for q in self._input_queues: q.put((None, None)) for w in self._workers: w.terminate() for i, w in enumerate(self._workers): w.join() def predict(self, dataset: Dataset, **kwargs) -> Iterator[Forecast]: with TemporaryDirectory() as tempdir: predictor_path = Path(tempdir) self._base_predictor.serialize(predictor_path) # TODO: Consider using shared memory for the data transfer. self._input_queues = [mp.Queue() for _ in range(self._num_workers)] self._output_queue = mp.Queue() workers = [] for worker_id, in_q in enumerate(self._input_queues): worker = mp.Process( target=_worker_loop, args=(predictor_path, in_q, self._output_queue, worker_id), kwargs=kwargs, ) worker.daemon = True worker.start() workers.append(worker) self._num_running_workers += 1 self._workers = workers chunked_data = self._grouper(dataset, self._chunk_size) self._send_idx = 0 self._next_idx = 0 self._data_buffer = {} worker_ids = list(range(self._num_workers)) def receive(): idx, worker_id, result = self._output_queue.get() if isinstance(idx, WorkerError): self._num_running_workers -= 1 self.terminate() raise Exception(idx.msg) if idx is not None: self._data_buffer[idx] = result return idx, worker_id, result def get_next_from_buffer(): while self._next_idx in self._data_buffer: result_batch = self._data_buffer.pop(self._next_idx) self._next_idx += 1 for result in result_batch: yield result def send(worker_id, chunk): q = self._input_queues[worker_id] q.put((self._send_idx, chunk)) self._send_idx += 1 try: # prime the queues for wid in worker_ids: chunk = next(chunked_data) send(wid, chunk) while True: idx, wid, result = receive() for res in get_next_from_buffer(): yield res chunk = next(chunked_data) send(wid, chunk) except StopIteration: # signal workers end of data for q in self._input_queues: q.put((None, None)) # collect any outstanding results while self._num_running_workers > 0: idx, worker_id, result = receive() if idx is None: self._num_running_workers -= 1 continue for res in get_next_from_buffer(): yield res assert len(self._data_buffer) == 0 assert self._send_idx == self._next_idx class Localizer(Predictor): """ A Predictor that uses an estimator to train a local model per time series and immediatly calls this to predict. Parameters ---------- estimator The estimator object to train on each dataset entry at prediction time. """ def __init__(self, estimator: "Estimator"): super().__init__(estimator.prediction_length, estimator.freq) self.estimator = estimator def predict(self, dataset: Dataset, **kwargs) -> Iterator[Forecast]: logger = logging.getLogger(__name__) for i, ts in enumerate(dataset, start=1): logger.info(f"training for time series {i} / {len(dataset)}") local_ds = ListDataset([ts], freq=self.freq) trained_pred = self.estimator.train(local_ds) logger.info(f"predicting for time series {i} / {len(dataset)}") predictions = trained_pred.predict(local_ds, **kwargs) for pred in predictions: yield pred class FallbackPredictor(Predictor): @classmethod def from_predictor( cls, base: RepresentablePredictor, **overrides ) -> Predictor: # Create predictor based on an existing predictor. # This let's us create a MeanPredictor as a fallback on the fly. return cls.from_hyperparameters( **getattr(base, "__init_args__"), **overrides ) def fallback(fallback_cls: Type[FallbackPredictor]): def decorator(predict_item): @functools.wraps(predict_item) def fallback_predict(self, item: DataEntry) -> Forecast: try: return predict_item(self, item) except GluonTSException: raise except Exception: logging.warning( f"Base predictor failed with: {traceback.format_exc()}" ) fallback_predictor = fallback_cls.from_predictor(self) return fallback_predictor.predict_item(item) return fallback_predict return decorator
23,995
30.994667
81
py
rankpredictor
rankpredictor-master/sub/gluonts/model/forecast_generator.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. import logging from typing import Any, Callable, Iterator, List, Optional # Third-party imports import mxnet as mx import numpy as np # First-party imports from gluonts.distribution import Distribution, DistributionOutput from gluonts.core.component import validated from gluonts.dataset.common import DataEntry from gluonts.dataset.field_names import FieldName from gluonts.dataset.loader import InferenceDataLoader from gluonts.model.forecast import ( Forecast, SampleForecast, QuantileForecast, DistributionForecast, ) OutputTransform = Callable[[DataEntry, np.ndarray], np.ndarray] BlockType = mx.gluon.Block LOG_CACHE = set([]) def log_once(msg): global LOG_CACHE if msg not in LOG_CACHE: logging.info(msg) LOG_CACHE.add(msg) def _extract_instances(x: Any) -> Any: """ Helper function to extract individual instances from batched mxnet results. For a tensor `a` _extract_instances(a) -> [a[0], a[1], ...] For (nested) tuples of tensors `(a, (b, c))` _extract_instances((a, (b, c)) -> [(a[0], (b[0], c[0])), (a[1], (b[1], c[1])), ...] """ if isinstance(x, (np.ndarray, mx.nd.NDArray)): for i in range(x.shape[0]): # yield x[i: i + 1] yield x[i] elif isinstance(x, tuple): for m in zip(*[_extract_instances(y) for y in x]): yield tuple([r for r in m]) elif isinstance(x, list): for m in zip(*[_extract_instances(y) for y in x]): yield [r for r in m] elif x is None: while True: yield None else: assert False class ForecastGenerator: """ Classes used to bring the output of a network into a class. """ def __call__( self, inference_data_loader: InferenceDataLoader, prediction_net: BlockType, input_names: List[str], freq: str, output_transform: Optional[OutputTransform], num_samples: Optional[int], **kwargs ) -> Iterator[Forecast]: raise NotImplementedError() class DistributionForecastGenerator(ForecastGenerator): @validated() def __init__(self, distr_output: DistributionOutput) -> None: self.distr_output = distr_output def __call__( self, inference_data_loader: InferenceDataLoader, prediction_net: BlockType, input_names: List[str], freq: str, output_transform: Optional[OutputTransform], num_samples: Optional[int], **kwargs ) -> Iterator[DistributionForecast]: for batch in inference_data_loader: inputs = [batch[k] for k in input_names] outputs = prediction_net(*inputs) if output_transform is not None: outputs = output_transform(batch, outputs) if num_samples: log_once( "Forecast is not sample based. Ignoring parameter `num_samples` from predict method." ) distributions = [ self.distr_output.distribution(*u) for u in _extract_instances(outputs) ] i = -1 for i, distr in enumerate(distributions): yield DistributionForecast( distr, start_date=batch["forecast_start"][i], freq=freq, item_id=batch[FieldName.ITEM_ID][i] if FieldName.ITEM_ID in batch else None, info=batch["info"][i] if "info" in batch else None, ) assert i + 1 == len(batch["forecast_start"]) class QuantileForecastGenerator(ForecastGenerator): @validated() def __init__(self, quantiles: List[str]) -> None: self.quantiles = quantiles def __call__( self, inference_data_loader: InferenceDataLoader, prediction_net: BlockType, input_names: List[str], freq: str, output_transform: Optional[OutputTransform], num_samples: Optional[int], **kwargs ) -> Iterator[Forecast]: for batch in inference_data_loader: inputs = [batch[k] for k in input_names] outputs = prediction_net(*inputs).asnumpy() if output_transform is not None: outputs = output_transform(batch, outputs) if num_samples: log_once( "Forecast is not sample based. Ignoring parameter `num_samples` from predict method." ) i = -1 for i, output in enumerate(outputs): yield QuantileForecast( output, start_date=batch["forecast_start"][i], freq=freq, item_id=batch[FieldName.ITEM_ID][i] if FieldName.ITEM_ID in batch else None, info=batch["info"][i] if "info" in batch else None, forecast_keys=self.quantiles, ) assert i + 1 == len(batch["forecast_start"]) class SampleForecastGenerator(ForecastGenerator): @validated() def __init__(self): pass def __call__( self, inference_data_loader: InferenceDataLoader, prediction_net: BlockType, input_names: List[str], freq: str, output_transform: Optional[OutputTransform], num_samples: Optional[int], **kwargs ) -> Iterator[Forecast]: for batch in inference_data_loader: inputs = [batch[k] for k in input_names] outputs = prediction_net(*inputs).asnumpy() if output_transform is not None: outputs = output_transform(batch, outputs) if num_samples: num_collected_samples = outputs[0].shape[0] collected_samples = [outputs] while num_collected_samples < num_samples: outputs = prediction_net(*inputs).asnumpy() if output_transform is not None: outputs = output_transform(batch, outputs) collected_samples.append(outputs) num_collected_samples += outputs[0].shape[0] outputs = [ np.concatenate(s)[:num_samples] for s in zip(*collected_samples) ] assert len(outputs[0]) == num_samples i = -1 for i, output in enumerate(outputs): yield SampleForecast( output, start_date=batch["forecast_start"][i], freq=freq, item_id=batch[FieldName.ITEM_ID][i] if FieldName.ITEM_ID in batch else None, info=batch["info"][i] if "info" in batch else None, ) assert i + 1 == len(batch["forecast_start"])
7,578
32.535398
105
py
rankpredictor
rankpredictor-master/sub/gluonts/model/forecast.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports import re from enum import Enum from typing import Dict, List, NamedTuple, Optional, Set, Union, Callable # Third-party imports import mxnet as mx import numpy as np import pandas as pd import pydantic # First-party imports from gluonts.core.exception import GluonTSUserError from gluonts.distribution import Distribution from gluonts.core.component import validated class Quantile(NamedTuple): value: float name: str @property def loss_name(self): return f"QuantileLoss[{self.name}]" @property def weighted_loss_name(self): return f"wQuantileLoss[{self.name}]" @property def coverage_name(self): return f"Coverage[{self.name}]" @classmethod def checked(cls, value: float, name: str) -> "Quantile": if not 0 <= value <= 1: raise GluonTSUserError( f"quantile value should be in [0, 1] but found {value}" ) return Quantile(value, name) @classmethod def from_float(cls, quantile: float) -> "Quantile": assert isinstance(quantile, float) return cls.checked(value=quantile, name=str(quantile)) @classmethod def from_str(cls, quantile: str) -> "Quantile": assert isinstance(quantile, str) try: return cls.checked(value=float(quantile), name=quantile) except ValueError: m = re.match(r"^p(\d{2})$", quantile) if m is None: raise GluonTSUserError( "Quantile string should be of the form " f'"p10", "p50", ... or "0.1", "0.5", ... but found {quantile}' ) else: quantile: float = int(m.group(1)) / 100 return cls(value=quantile, name=str(quantile)) @classmethod def parse(cls, quantile: Union["Quantile", float, str]) -> "Quantile": """Produces equivalent float and string representation of a given quantile level. >>> Quantile.parse(0.1) Quantile(value=0.1, name='0.1') >>> Quantile.parse('0.2') Quantile(value=0.2, name='0.2') >>> Quantile.parse('0.20') Quantile(value=0.2, name='0.20') >>> Quantile.parse('p99') Quantile(value=0.99, name='0.99') Parameters ---------- quantile Quantile, can be a float a str representing a float e.g. '0.1' or a quantile string of the form 'p0.1'. Returns ------- Quantile A tuple containing both a float and a string representation of the input quantile level. """ if isinstance(quantile, Quantile): return quantile elif isinstance(quantile, float): return cls.from_float(quantile) else: return cls.from_str(quantile) class Forecast: """ A abstract class representing predictions. """ start_date: pd.Timestamp freq: str item_id: Optional[str] info: Optional[Dict] prediction_length: int mean: np.ndarray _index = None def quantile(self, q: Union[float, str]) -> np.ndarray: """ Computes a quantile from the predicted distribution. Parameters ---------- q Quantile to compute. Returns ------- numpy.ndarray Value of the quantile across the prediction range. """ raise NotImplementedError() @property def median(self) -> np.ndarray: return self.quantile(0.5) def plot( self, prediction_intervals=(50.0, 90.0), show_mean=False, color="b", label=None, output_file=None, *args, **kwargs, ): """ Plots the median of the forecast as well as confidence bounds. (requires matplotlib and pandas). Parameters ---------- prediction_intervals : float or list of floats in [0, 100] Confidence interval size(s). If a list, it will stack the error plots for each confidence interval. Only relevant for error styles with "ci" in the name. show_mean : boolean Whether to also show the mean of the forecast. color : matplotlib color name or dictionary The color used for plotting the forecast. label : string A label (prefix) that is used for the forecast output_file : str or None, default None Output path for the plot file. If None, plot is not saved to file. args : Other arguments are passed to main plot() call kwargs : Other keyword arguments are passed to main plot() call """ # matplotlib==2.0.* gives errors in Brazil builds and has to be # imported locally import matplotlib.pyplot as plt label_prefix = "" if label is None else label + "-" for c in prediction_intervals: assert 0.0 <= c <= 100.0 ps = [50.0] + [ 50.0 + f * c / 2.0 for c in prediction_intervals for f in [-1.0, +1.0] ] percentiles_sorted = sorted(set(ps)) def alpha_for_percentile(p): return (p / 100.0) ** 0.3 ps_data = [self.quantile(p / 100.0) for p in percentiles_sorted] i_p50 = len(percentiles_sorted) // 2 p50_data = ps_data[i_p50] p50_series = pd.Series(data=p50_data, index=self.index) p50_series.plot(color=color, ls="-", label=f"{label_prefix}median", **kwargs) if show_mean: mean_data = np.mean(self._sorted_samples, axis=0) pd.Series(data=mean_data, index=self.index).plot( color=color, ls=":", label=f"{label_prefix}mean", *args, **kwargs, ) for i in range(len(percentiles_sorted) // 2): ptile = percentiles_sorted[i] alpha = alpha_for_percentile(ptile) plt.fill_between( self.index, ps_data[i], ps_data[-i - 1], facecolor=color, alpha=alpha, interpolate=True, *args, **kwargs, ) # Hack to create labels for the error intervals. # Doesn't actually plot anything, because we only pass a single data point pd.Series(data=p50_data[:1], index=self.index[:1]).plot( color=color, alpha=alpha, #linewidth=10, label=f"{label_prefix}{100 - ptile * 2}%", *args, **kwargs, ) if output_file: plt.savefig(output_file) @property def index(self) -> pd.DatetimeIndex: if self._index is None: self._index = pd.date_range( self.start_date, periods=self.prediction_length, freq=self.freq ) return self._index def dim(self) -> int: """ Returns the dimensionality of the forecast object. """ raise NotImplementedError() def copy_dim(self, dim: int): """ Returns a new Forecast object with only the selected sub-dimension. Parameters ---------- dim The returned forecast object will only represent this dimension. """ raise NotImplementedError() def copy_aggregate(self, agg_fun: Callable): """ Returns a new Forecast object with a time series aggregated over the dimension axis. Parameters ---------- agg_fun Aggregation function that defines the aggregation operation (typically mean or sum). """ raise NotImplementedError() def as_json_dict(self, config: "Config") -> dict: result = {} if OutputType.mean in config.output_types: result["mean"] = self.mean.tolist() if OutputType.quantiles in config.output_types: quantiles = map(Quantile.parse, config.quantiles) result["quantiles"] = { quantile.name: self.quantile(quantile.value).tolist() for quantile in quantiles } if OutputType.samples in config.output_types: result["samples"] = [] return result class SampleForecast(Forecast): """ A `Forecast` object, where the predicted distribution is represented internally as samples. Parameters ---------- samples Array of size (num_samples, prediction_length) (1D case) or (num_samples, prediction_length, target_dim) (multivariate case) start_date start of the forecast freq forecast frequency info additional information that the forecaster may provide e.g. estimated parameters, number of iterations ran etc. """ @validated() def __init__( self, samples: Union[mx.nd.NDArray, np.ndarray], start_date, freq, item_id: Optional[str] = None, info: Optional[Dict] = None, ): assert isinstance( samples, (np.ndarray, mx.ndarray.ndarray.NDArray) ), "samples should be either a numpy or an mxnet array" assert ( len(np.shape(samples)) == 2 or len(np.shape(samples)) == 3 ), "samples should be a 2-dimensional or 3-dimensional array. Dimensions found: {}".format( len(np.shape(samples)) ) self.samples = ( samples if (isinstance(samples, np.ndarray)) else samples.asnumpy() ) self._sorted_samples_value = None self._mean = None self._dim = None self.item_id = item_id self.info = info assert isinstance( start_date, pd.Timestamp ), "start_date should be a pandas Timestamp object" self.start_date = start_date assert isinstance(freq, str), "freq should be a string" self.freq = freq @property def _sorted_samples(self): if self._sorted_samples_value is None: self._sorted_samples_value = np.sort(self.samples, axis=0) return self._sorted_samples_value @property def num_samples(self): """ The number of samples representing the forecast. """ return self.samples.shape[0] @property def prediction_length(self): """ Time length of the forecast. """ return self.samples.shape[-1] @property def mean(self): """ Forecast mean. """ if self._mean is not None: return self._mean else: return np.mean(self.samples, axis=0) @property def mean_ts(self): """ Forecast mean, as a pandas.Series object. """ return pd.Series(self.index, self.mean) def quantile(self, q): q = Quantile.parse(q).value sample_idx = int(np.round((self.num_samples - 1) * q)) return self._sorted_samples[sample_idx, :] def copy_dim(self, dim: int): if len(self.samples.shape) == 2: samples = self.samples else: target_dim = self.samples.shape[2] assert dim < target_dim, ( f"must set 0 <= dim < target_dim, but got dim={dim}," f" target_dim={target_dim}" ) samples = self.samples[:, :, dim] return SampleForecast( samples=samples, start_date=self.start_date, freq=self.freq, item_id=self.item_id, info=self.info, ) def copy_aggregate(self, agg_fun: Callable): if len(self.samples.shape) == 2: samples = self.samples else: # Aggregate over target dimension axis samples = agg_fun(self.samples, axis=2) return SampleForecast( samples=samples, start_date=self.start_date, freq=self.freq, item_id=self.item_id, info=self.info, ) def dim(self) -> int: if self._dim is not None: return self._dim else: if len(self.samples.shape) == 2: # univariate target # shape: (num_samples, prediction_length) return 1 else: # multivariate target # shape: (num_samples, prediction_length, target_dim) return self.samples.shape[2] def as_json_dict(self, config: "Config") -> dict: result = super().as_json_dict(config) if OutputType.samples in config.output_types: result["samples"] = self.samples.tolist() return result def __repr__(self): return ", ".join( [ f"SampleForecast({self.samples!r})", f"{self.start_date!r}", f"{self.freq!r}", f"item_id={self.item_id!r}", f"info={self.info!r})", ] ) class QuantileForecast(Forecast): """ A Forecast that contains arrays (i.e. time series) for quantiles and mean Parameters ---------- forecast_arrays An array of forecasts start_date start of the forecast freq forecast frequency forecast_keys A list of quantiles of the form '0.1', '0.9', etc., and potentially 'mean'. Each entry corresponds to one array in forecast_arrays. info additional information that the forecaster may provide e.g. estimated parameters, number of iterations ran etc. """ def __init__( self, forecast_arrays: np.ndarray, start_date: pd.Timestamp, freq: str, forecast_keys: List[str], item_id: Optional[str] = None, info: Optional[Dict] = None, ): self.forecast_array = forecast_arrays self.start_date = pd.Timestamp(start_date, freq=freq) self.freq = freq # normalize keys self.forecast_keys = [ Quantile.from_str(key).name if key != "mean" else key for key in forecast_keys ] self.item_id = item_id self.info = info self._dim = None shape = self.forecast_array.shape assert shape[0] == len(self.forecast_keys), ( f"The forecast_array (shape={shape} should have the same " f"length as the forecast_keys (len={len(self.forecast_keys)})." ) self.prediction_length = shape[-1] self._forecast_dict = { k: self.forecast_array[i] for i, k in enumerate(self.forecast_keys) } self._nan_out = np.array([np.nan] * self.prediction_length) def quantile(self, q: Union[float, str]) -> np.ndarray: q_str = Quantile.parse(q).name # We return nan here such that evaluation runs through return self._forecast_dict.get(q_str, self._nan_out) @property def mean(self): """ Forecast mean. """ return self._forecast_dict.get("mean", self._nan_out) def dim(self) -> int: if self._dim is not None: return self._dim else: if ( len(self.forecast_array.shape) == 2 ): # 1D target. shape: (num_samples, prediction_length) return 1 else: return self.forecast_array.shape[ 1 ] # 2D target. shape: (num_samples, target_dim, prediction_length) def __repr__(self): return ", ".join( [ f"QuantileForecast({self.forecast_array!r})", f"start_date={self.start_date!r}", f"freq={self.freq!r}", f"forecast_keys={self.forecast_keys!r}", f"item_id={self.item_id!r}", f"info={self.info!r})", ] ) class DistributionForecast(Forecast): """ A `Forecast` object that uses a GluonTS distribution directly. This can for instance be used to represent marginal probability distributions for each time point -- although joint distributions are also possible, e.g. when using MultiVariateGaussian). Parameters ---------- distribution Distribution object. This should represent the entire prediction length, i.e., if we draw `num_samples` samples from the distribution, the sample shape should be samples = trans_dist.sample(num_samples) samples.shape -> (num_samples, prediction_length) start_date start of the forecast freq forecast frequency info additional information that the forecaster may provide e.g. estimated parameters, number of iterations ran etc. """ @validated() def __init__( self, distribution: Distribution, start_date, freq, item_id: Optional[str] = None, info: Optional[Dict] = None, ): self.distribution = distribution self.shape = ( self.distribution.batch_shape + self.distribution.event_shape ) self.prediction_length = self.shape[0] self.item_id = item_id self.info = info assert isinstance( start_date, pd.Timestamp ), "start_date should be a pandas Timestamp object" self.start_date = start_date assert isinstance(freq, str), "freq should be a string" self.freq = freq self._mean = None @property def mean(self): """ Forecast mean. """ if self._mean is not None: return self._mean else: self._mean = self.distribution.mean.asnumpy() return self._mean @property def mean_ts(self): """ Forecast mean, as a pandas.Series object. """ return pd.Series(self.index, self.mean) def quantile(self, level): level = Quantile.parse(level).value q = self.distribution.quantile(mx.nd.array([level])).asnumpy()[0] return q def to_sample_forecast(self, num_samples: int = 200) -> SampleForecast: return SampleForecast( samples=self.distribution.sample(num_samples), start_date=self.start_date, freq=self.freq, item_id=self.item_id, info=self.info, ) class OutputType(str, Enum): mean = "mean" samples = "samples" quantiles = "quantiles" class Config(pydantic.BaseModel): num_samples: int = pydantic.Field(100, alias="num_eval_samples") output_types: Set[OutputType] = {"quantiles", "mean"} # FIXME: validate list elements quantiles: List[str] = ["0.1", "0.5", "0.9"] class Config: allow_population_by_field_name = True # store additional fields extra = "allow"
19,676
28.949772
99
py
rankpredictor
rankpredictor-master/sub/gluonts/model/common.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports import types import typing # Third-party imports import mxnet as mx import numpy as np # Tensor type for HybridBlocks in Gluon Tensor = typing.Union[mx.nd.NDArray, mx.sym.Symbol] # Type of tensor-transforming functions in Gluon TensorTransformer = typing.Callable[[types.ModuleType, Tensor], Tensor] # untyped global configuration passed to model components GlobalConfig = typing.Dict[str, typing.Any] # to annotate Numpy parameter NPArrayLike = typing.Union[int, float, np.ndarray]
1,091
32.090909
75
py
rankpredictor
rankpredictor-master/sub/gluonts/model/estimator.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import NamedTuple, Optional # Third-party imports import numpy as np from mxnet.gluon import HybridBlock from pydantic import ValidationError # First-party imports import gluonts from gluonts.core import fqname_for from gluonts.core.component import DType, from_hyperparameters, validated from gluonts.core.exception import GluonTSHyperparametersError from gluonts.dataset.common import Dataset from gluonts.dataset.loader import TrainDataLoader, ValidationDataLoader from gluonts.model.predictor import Predictor from gluonts.support.util import get_hybrid_forward_input_names from gluonts.trainer import Trainer from gluonts.transform import Transformation class Estimator: """ An abstract class representing a trainable model. The underlying model is trained by calling the `train` method with a training `Dataset`, producing a `Predictor` object. """ __version__: str = gluonts.__version__ prediction_length: int freq: str def train( self, training_data: Dataset, validation_data: Optional[Dataset] = None ) -> Predictor: """ Train the estimator on the given data. Parameters ---------- training_data Dataset to train the model on. validation_data Dataset to validate the model on during training. Returns ------- Predictor The predictor containing the trained model. """ raise NotImplementedError @classmethod def from_hyperparameters(cls, **hyperparameters): return from_hyperparameters(cls, **hyperparameters) class DummyEstimator(Estimator): """ An `Estimator` that, upon training, simply returns a pre-constructed `Predictor`. Parameters ---------- predictor_cls `Predictor` class to instantiate. **kwargs Keyword arguments to pass to the predictor constructor. """ @validated() def __init__(self, predictor_cls: type, **kwargs) -> None: self.predictor = predictor_cls(**kwargs) def train( self, training_data: Dataset, validation_dataset: Optional[Dataset] = None, ) -> Predictor: return self.predictor class TrainOutput(NamedTuple): transformation: Transformation trained_net: HybridBlock predictor: Predictor class GluonEstimator(Estimator): """ An `Estimator` type with utilities for creating Gluon-based models. To extend this class, one needs to implement three methods: `create_transformation`, `create_training_network`, `create_predictor`. """ @validated() def __init__(self, trainer: Trainer, dtype: DType = np.float32) -> None: self.trainer = trainer self.dtype = dtype @classmethod def from_hyperparameters(cls, **hyperparameters) -> "GluonEstimator": Model = getattr(cls.__init__, "Model", None) if not Model: raise AttributeError( f"Cannot find attribute Model attached to the " f"{fqname_for(cls)}. Most probably you have forgotten to mark " f"the class constructor as @validated()." ) try: trainer = from_hyperparameters(Trainer, **hyperparameters) return cls( **Model(**{**hyperparameters, "trainer": trainer}).__dict__ ) except ValidationError as e: raise GluonTSHyperparametersError from e def create_transformation(self) -> Transformation: """ Create and return the transformation needed for training and inference. Returns ------- Transformation The transformation that will be applied entry-wise to datasets, at training and inference time. """ raise NotImplementedError def create_training_network(self) -> HybridBlock: """ Create and return the network used for training (i.e., computing the loss). Returns ------- HybridBlock The network that computes the loss given input data. """ raise NotImplementedError def create_predictor( self, transformation: Transformation, trained_network: HybridBlock ) -> Predictor: """ Create and return a predictor object. Returns ------- Predictor A predictor wrapping a `HybridBlock` used for inference. """ raise NotImplementedError def train_model( self, training_data: Dataset, validation_data: Optional[Dataset] = None ) -> TrainOutput: transformation = self.create_transformation() transformation.estimate(iter(training_data)) training_data_loader = TrainDataLoader( dataset=training_data, transform=transformation, batch_size=self.trainer.batch_size, num_batches_per_epoch=self.trainer.num_batches_per_epoch, ctx=self.trainer.ctx, dtype=self.dtype, ) validation_data_loader = None if validation_data is not None: validation_data_loader = ValidationDataLoader( dataset=validation_data, transform=transformation, batch_size=self.trainer.batch_size, ctx=self.trainer.ctx, dtype=self.dtype, ) # ensure that the training network is created within the same MXNet # context as the one that will be used during training with self.trainer.ctx: trained_net = self.create_training_network() self.trainer( net=trained_net, input_names=get_hybrid_forward_input_names(trained_net), train_iter=training_data_loader, validation_iter=validation_data_loader, ) with self.trainer.ctx: # ensure that the prediction network is created within the same MXNet # context as the one that was used during training return TrainOutput( transformation=transformation, trained_net=trained_net, predictor=self.create_predictor(transformation, trained_net), ) def train( self, training_data: Dataset, validation_data: Optional[Dataset] = None ) -> Predictor: return self.train_model(training_data, validation_data).predictor
7,061
30.526786
81
py
rankpredictor
rankpredictor-master/sub/gluonts/model/canonical/_network.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Third-party imports from mxnet.gluon import HybridBlock # First-party imports from gluonts.block.feature import FeatureEmbedder from gluonts.block.scaler import MeanScaler from gluonts.core.component import validated from gluonts.distribution import DistributionOutput from gluonts.model.common import Tensor class CanonicalNetworkBase(HybridBlock): @validated() def __init__( self, model: HybridBlock, embedder: FeatureEmbedder, distr_output: DistributionOutput, is_sequential: bool, **kwargs, ) -> None: super().__init__(**kwargs) self.distr_output = distr_output self.is_sequential = is_sequential self.model = model self.embedder = embedder with self.name_scope(): self.proj_distr_args = self.distr_output.get_args_proj() self.scaler = MeanScaler(keepdims=True) def assemble_features( self, F, feat_static_cat: Tensor, # (batch_size, num_features) time_feat: Tensor, # (batch_size, history_length, num_features) ) -> Tensor: embedded_cat = self.embedder( feat_static_cat ) # (batch_size, num_features * embedding_size) # a workaround when you wish to repeat without knowing the number # of repeats helper_ones = F.ones_like( F.slice_axis(time_feat, axis=2, begin=-1, end=None) ) # (batch_size, history_length, num_features * embedding_size) repeated_cat = F.batch_dot( helper_ones, F.expand_dims(embedded_cat, axis=1) ) # putting together all the features input_feat = F.concat(repeated_cat, time_feat, dim=2) return input_feat def hybrid_forward(self, F, x, *args, **kwargs): raise NotImplementedError class CanonicalTrainingNetwork(CanonicalNetworkBase): # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, feat_static_cat: Tensor, # (batch_size, num_features) past_time_feat: Tensor, # (batch_size, num_features, history_length) past_target: Tensor, # (batch_size, history_length) ) -> Tensor: """ Parameters ---------- F Function space feat_static_cat Shape: (batch_size, num_features) past_time_feat Shape: (batch_size, history_length, num_features) past_target Shape: (batch_size, history_length) Returns ------- Tensor A batch of negative log likelihoods. """ _, target_scale = self.scaler( past_target, F.ones_like(past_target), # TODO: pass the actual observed here ) input_feat = self.assemble_features(F, feat_static_cat, past_time_feat) outputs = self.model(input_feat) distr = self.distr_output.distribution( self.proj_distr_args(outputs), scale=target_scale ) loss = distr.loss(past_target) return loss class CanonicalPredictionNetwork(CanonicalNetworkBase): @validated() def __init__( self, prediction_len: int, num_parallel_samples: int, **kwargs ) -> None: super().__init__(**kwargs) self.prediction_len = prediction_len self.num_parallel_samples = num_parallel_samples # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, feat_static_cat: Tensor, past_time_feat: Tensor, future_time_feat: Tensor, past_target: Tensor, ) -> Tensor: """ Parameters ---------- F Function space module. feat_static_cat Shape: (batch_size, num_features). past_time_feat Shape: (batch_size, history_length, num_features). future_time_feat Shape: (batch_size, history_length, num_features). past_target Shape: (batch_size, history_length). Returns ------- Tensor a batch of prediction samples Shape: (batch_size, prediction_length, num_sample_paths) """ _, target_scale = self.scaler( past_target, F.ones_like(past_target), # TODO: pass the actual observed here ) time_feat = ( F.concat(past_time_feat, future_time_feat, dim=1) if self.is_sequential else future_time_feat ) input_feat = self.assemble_features(F, feat_static_cat, time_feat) outputs = self.model(input_feat) if self.is_sequential: outputs = F.slice_axis( outputs, axis=1, begin=-self.prediction_len, end=None ) distr = self.distr_output.distribution( self.proj_distr_args(outputs), scale=target_scale ) samples = distr.sample( self.num_parallel_samples ) # (num_samples, batch_size, prediction_length, 1) return samples.swapaxes(0, 1)
5,687
29.745946
79
py
rankpredictor
rankpredictor-master/sub/gluonts/model/canonical/_estimator.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import List # Third-party imports from mxnet.gluon import HybridBlock, nn # First-party imports from gluonts.block.feature import FeatureEmbedder from gluonts.block.rnn import RNN from gluonts.core.component import validated from gluonts.dataset.field_names import FieldName from gluonts.distribution import DistributionOutput, StudentTOutput from gluonts.model.estimator import GluonEstimator from gluonts.model.predictor import Predictor, RepresentableBlockPredictor from gluonts.time_feature import time_features_from_frequency_str from gluonts.trainer import Trainer from gluonts.transform import ( AddTimeFeatures, AsNumpyArray, Chain, InstanceSplitter, SetFieldIfNotPresent, TestSplitSampler, Transformation, ) # Relative imports from ._network import CanonicalPredictionNetwork, CanonicalTrainingNetwork class CanonicalEstimator(GluonEstimator): @validated() def __init__( self, model: HybridBlock, is_sequential: bool, freq: str, context_length: int, prediction_length: int, trainer: Trainer = Trainer(), num_parallel_samples: int = 100, cardinality: List[int] = list([1]), embedding_dimension: int = 10, distr_output: DistributionOutput = StudentTOutput(), ) -> None: super().__init__(trainer=trainer) # TODO: error checking self.freq = freq self.context_length = context_length self.prediction_length = prediction_length self.distr_output = distr_output self.num_parallel_samples = num_parallel_samples self.cardinality = cardinality self.embedding_dimensions = [embedding_dimension for _ in cardinality] self.model = model self.is_sequential = is_sequential def create_transformation(self) -> Transformation: return Chain( trans=[ AsNumpyArray(field=FieldName.TARGET, expected_ndim=1), AddTimeFeatures( start_field=FieldName.START, target_field=FieldName.TARGET, output_field=FieldName.FEAT_TIME, time_features=time_features_from_frequency_str(self.freq), pred_length=self.prediction_length, ), SetFieldIfNotPresent( field=FieldName.FEAT_STATIC_CAT, value=[0.0] ), AsNumpyArray(field=FieldName.FEAT_STATIC_CAT, expected_ndim=1), InstanceSplitter( target_field=FieldName.TARGET, is_pad_field=FieldName.IS_PAD, start_field=FieldName.START, forecast_start_field=FieldName.FORECAST_START, train_sampler=TestSplitSampler(), time_series_fields=[FieldName.FEAT_TIME], past_length=self.context_length, future_length=self.prediction_length, ), ] ) def create_training_network(self) -> CanonicalTrainingNetwork: return CanonicalTrainingNetwork( embedder=FeatureEmbedder( cardinalities=self.cardinality, embedding_dims=self.embedding_dimensions, ), model=self.model, distr_output=self.distr_output, is_sequential=self.is_sequential, ) def create_predictor( self, transformation: Transformation, trained_network: CanonicalTrainingNetwork, ) -> Predictor: prediction_net = CanonicalPredictionNetwork( embedder=trained_network.embedder, model=trained_network.model, distr_output=trained_network.distr_output, is_sequential=trained_network.is_sequential, prediction_len=self.prediction_length, num_parallel_samples=self.num_parallel_samples, params=trained_network.collect_params(), ) return RepresentableBlockPredictor( input_transform=transformation, prediction_net=prediction_net, batch_size=self.trainer.batch_size, freq=self.freq, prediction_length=self.prediction_length, ctx=self.trainer.ctx, ) class CanonicalRNNEstimator(CanonicalEstimator): @validated() def __init__( self, freq: str, context_length: int, prediction_length: int, trainer: Trainer = Trainer(), num_layers: int = 1, num_cells: int = 50, cell_type: str = "lstm", num_parallel_samples: int = 100, cardinality: List[int] = list([1]), embedding_dimension: int = 10, distr_output: DistributionOutput = StudentTOutput(), ) -> None: model = RNN( mode=cell_type, num_layers=num_layers, num_hidden=num_cells ) super(CanonicalRNNEstimator, self).__init__( model=model, is_sequential=True, freq=freq, context_length=context_length, prediction_length=prediction_length, trainer=trainer, num_parallel_samples=num_parallel_samples, cardinality=cardinality, embedding_dimension=embedding_dimension, distr_output=distr_output, ) class MLPForecasterEstimator(CanonicalEstimator): @validated() def __init__( self, freq: str, context_length: int, prediction_length: int, trainer: Trainer = Trainer(), hidden_dim_sequence=list([50]), num_parallel_samples: int = 100, cardinality: List[int] = list([1]), embedding_dimension: int = 10, distr_output: DistributionOutput = StudentTOutput(), ) -> None: model = nn.HybridSequential() for layer, layer_dim in enumerate(hidden_dim_sequence): model.add( nn.Dense( layer_dim, flatten=False, activation="relu", prefix="mlp_%d_" % layer, ) ) super(MLPForecasterEstimator, self).__init__( model=model, is_sequential=False, freq=freq, context_length=context_length, prediction_length=prediction_length, trainer=trainer, num_parallel_samples=num_parallel_samples, cardinality=cardinality, embedding_dimension=embedding_dimension, distr_output=distr_output, )
7,229
33.759615
79
py
rankpredictor
rankpredictor-master/sub/gluonts/model/deepar-original/predictor.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports import functools import itertools import logging import multiprocessing as mp import sys import traceback from pathlib import Path from pydoc import locate from tempfile import TemporaryDirectory import json from typing import ( TYPE_CHECKING, Tuple, Union, Any, Callable, Dict, Iterator, List, Optional, Type, ) # Third-party imports import mxnet as mx import numpy as np # First-party imports import gluonts from gluonts.distribution import Distribution, DistributionOutput from gluonts.core.component import ( DType, equals, from_hyperparameters, get_mxnet_context, validated, ) from gluonts.core.exception import GluonTSException from gluonts.core.serde import dump_json, fqname_for, load_json from gluonts.dataset.common import DataEntry, Dataset, ListDataset from .forecast_generator import ForecastGenerator, SampleForecastGenerator from gluonts.dataset.loader import DataBatch, InferenceDataLoader from gluonts.model.forecast import Forecast from gluonts.support.util import ( export_repr_block, export_symb_block, get_hybrid_forward_input_names, hybrid_block_to_symbol_block, import_repr_block, import_symb_block, ) from gluonts.transform import Transformation if TYPE_CHECKING: # avoid circular import from gluonts.model.estimator import Estimator # noqa OutputTransform = Callable[[DataEntry, np.ndarray], np.ndarray] class Predictor: """ Abstract class representing predictor objects. Parameters ---------- prediction_length Prediction horizon. freq Frequency of the predicted data. """ __version__: str = gluonts.__version__ def __init__(self, prediction_length: int, freq: str) -> None: assert ( prediction_length > 0 ), "The value of `prediction_length` should be > 0" self.prediction_length = prediction_length self.freq = freq def predict(self, dataset: Dataset, **kwargs) -> Iterator[Forecast]: """ Compute forecasts for the time series in the provided dataset. This method is not implemented in this abstract class; please use one of the subclasses. Parameters ---------- dataset The dataset containing the time series to predict. Returns ------- Iterator[Forecast] Iterator over the forecasts, in the same order as the dataset iterable was provided. """ raise NotImplementedError def serialize(self, path: Path) -> None: # serialize Predictor type with (path / "type.txt").open("w") as fp: fp.write(fqname_for(self.__class__)) with (path / "version.json").open("w") as fp: json.dump( {"model": self.__version__, "gluonts": gluonts.__version__}, fp ) @classmethod def deserialize( cls, path: Path, ctx: Optional[mx.Context] = None ) -> "Predictor": """ Load a serialized predictor from the given path Parameters ---------- path Path to the serialized files predictor. ctx Optional mxnet context to be used with the predictor. If nothing is passed will use the GPU if available and CPU otherwise. """ # deserialize Predictor type with (path / "type.txt").open("r") as fp: tpe = locate(fp.readline()) # ensure that predictor_cls is a subtype of Predictor if not issubclass(tpe, Predictor): raise IOError( f"Class {fqname_for(tpe)} is not " f"a subclass of {fqname_for(Predictor)}" ) # call deserialize() for the concrete Predictor type return tpe.deserialize(path, ctx) @classmethod def from_hyperparameters(cls, **hyperparameters): return from_hyperparameters(cls, **hyperparameters) class RepresentablePredictor(Predictor): """ An abstract predictor that can be subclassed by models that are not based on Gluon. Subclasses should have @validated() constructors. (De)serialization and value equality are all implemented on top of the @validated() logic. Parameters ---------- prediction_length Prediction horizon. freq Frequency of the predicted data. """ @validated() def __init__(self, prediction_length: int, freq: str) -> None: super().__init__(prediction_length, freq) def predict(self, dataset: Dataset, **kwargs) -> Iterator[Forecast]: for item in dataset: yield self.predict_item(item) def predict_item(self, item: DataEntry) -> Forecast: raise NotImplementedError def __eq__(self, that): """ Two RepresentablePredictor instances are considered equal if they have the same constructor arguments. """ return equals(self, that) def serialize(self, path: Path) -> None: # call Predictor.serialize() in order to serialize the class name super().serialize(path) with (path / "predictor.json").open("w") as fp: print(dump_json(self), file=fp) @classmethod def deserialize( cls, path: Path, ctx: Optional[mx.Context] = None ) -> "RepresentablePredictor": with (path / "predictor.json").open("r") as fp: return load_json(fp.read()) class GluonPredictor(Predictor): """ Base predictor type for Gluon-based models. Parameters ---------- input_names Input tensor names for the graph prediction_net Network that will be called for prediction batch_size Number of time series to predict in a single batch prediction_length Number of time steps to predict freq Frequency of the input data input_transform Input transformation pipeline output_transform Output transformation ctx MXNet context to use for computation forecast_generator Class to generate forecasts from network ouputs """ BlockType = mx.gluon.Block def __init__( self, input_names: List[str], prediction_net: BlockType, batch_size: int, prediction_length: int, freq: str, ctx: mx.Context, input_transform: Transformation, forecast_generator: ForecastGenerator = SampleForecastGenerator(), output_transform: Optional[OutputTransform] = None, dtype: DType = np.float32, ) -> None: super().__init__(prediction_length, freq) self.input_names = input_names self.prediction_net = prediction_net self.batch_size = batch_size self.input_transform = input_transform self.forecast_generator = forecast_generator self.output_transform = output_transform self.ctx = ctx self.dtype = dtype def hybridize(self, batch: DataBatch) -> None: """ Hybridizes the underlying prediction network. Parameters ---------- batch A batch of data to use for the required forward pass after the `hybridize()` call. """ self.prediction_net.hybridize(active=True) self.prediction_net(*[batch[k] for k in self.input_names]) def as_symbol_block_predictor( self, batch: DataBatch ) -> "SymbolBlockPredictor": """ Returns a variant of the current :class:`GluonPredictor` backed by a Gluon `SymbolBlock`. If the current predictor is already a :class:`SymbolBlockPredictor`, it just returns itself. Parameters ---------- batch A batch of data to use for the required forward pass after the `hybridize()` call of the underlying network. Returns ------- SymbolBlockPredictor A predictor derived from the current one backed by a `SymbolBlock`. """ raise NotImplementedError def predict( self, dataset: Dataset, num_samples: Optional[int] = None ) -> Iterator[Forecast]: #print('predict') inference_data_loader = InferenceDataLoader( dataset, self.input_transform, self.batch_size, ctx=self.ctx, dtype=self.dtype, ) yield from self.forecast_generator( inference_data_loader=inference_data_loader, prediction_net=self.prediction_net, input_names=self.input_names, freq=self.freq, output_transform=self.output_transform, num_samples=num_samples, ) def __eq__(self, that): if type(self) != type(that): return False # TODO: also consider equality of the pipelines # if not equals(self.input_transform, that.input_transform): # return False return equals( self.prediction_net.collect_params(), that.prediction_net.collect_params(), ) def serialize(self, path: Path) -> None: # call Predictor.serialize() in order to serialize the class name super().serialize(path) # serialize every GluonPredictor-specific parameters # serialize the prediction network self.serialize_prediction_net(path) # serialize transformation chain with (path / "input_transform.json").open("w") as fp: print(dump_json(self.input_transform), file=fp) # FIXME: also needs to serialize the output_transform # serialize all remaining constructor parameters with (path / "parameters.json").open("w") as fp: parameters = dict( batch_size=self.batch_size, prediction_length=self.prediction_length, freq=self.freq, ctx=self.ctx, dtype=self.dtype, forecast_generator=self.forecast_generator, input_names=self.input_names, ) print(dump_json(parameters), file=fp) def serialize_prediction_net(self, path: Path) -> None: raise NotImplementedError() class SymbolBlockPredictor(GluonPredictor): """ A predictor which serializes the network structure as an MXNet symbolic graph. Should be used for models deployed in production in order to ensure forward-compatibility as GluonTS models evolve. Used by the training shell if training is invoked with a hyperparameter `use_symbol_block_predictor = True`. """ BlockType = mx.gluon.SymbolBlock def as_symbol_block_predictor( self, batch: DataBatch ) -> "SymbolBlockPredictor": return self def serialize_prediction_net(self, path: Path) -> None: export_symb_block(self.prediction_net, path, "prediction_net") @classmethod def deserialize( cls, path: Path, ctx: Optional[mx.Context] = None ) -> "SymbolBlockPredictor": ctx = ctx if ctx is not None else get_mxnet_context() with mx.Context(ctx): # deserialize constructor parameters with (path / "parameters.json").open("r") as fp: parameters = load_json(fp.read()) parameters["ctx"] = ctx # deserialize transformation chain with (path / "input_transform.json").open("r") as fp: transform = load_json(fp.read()) # deserialize prediction network num_inputs = len(parameters["input_names"]) prediction_net = import_symb_block( num_inputs, path, "prediction_net" ) return SymbolBlockPredictor( input_transform=transform, prediction_net=prediction_net, **parameters, ) class RepresentableBlockPredictor(GluonPredictor): """ A predictor which serializes the network structure using the JSON-serialization methods located in `gluonts.core.serde`. Use the following logic to create a `RepresentableBlockPredictor` from a trained prediction network. >>> def create_representable_block_predictor( ... prediction_network: mx.gluon.HybridBlock, ... **kwargs ... ) -> RepresentableBlockPredictor: ... return RepresentableBlockPredictor( ... prediction_net=prediction_network, ... **kwargs ... ) """ BlockType = mx.gluon.HybridBlock def __init__( self, prediction_net: BlockType, batch_size: int, prediction_length: int, freq: str, ctx: mx.Context, input_transform: Transformation, forecast_generator: ForecastGenerator = SampleForecastGenerator(), output_transform: Optional[ Callable[[DataEntry, np.ndarray], np.ndarray] ] = None, dtype: DType = np.float32, ) -> None: super().__init__( input_names=get_hybrid_forward_input_names(prediction_net), prediction_net=prediction_net, batch_size=batch_size, prediction_length=prediction_length, freq=freq, ctx=ctx, input_transform=input_transform, forecast_generator=forecast_generator, output_transform=output_transform, dtype=dtype, ) def as_symbol_block_predictor( self, batch: DataBatch ) -> SymbolBlockPredictor: symbol_block_net = hybrid_block_to_symbol_block( hb=self.prediction_net, data_batch=[batch[k] for k in self.input_names], ) return SymbolBlockPredictor( input_names=self.input_names, prediction_net=symbol_block_net, batch_size=self.batch_size, prediction_length=self.prediction_length, freq=self.freq, ctx=self.ctx, input_transform=self.input_transform, forecast_generator=self.forecast_generator, output_transform=self.output_transform, dtype=self.dtype, ) def serialize(self, path: Path) -> None: logging.warning( "Serializing RepresentableBlockPredictor instances does not save " "the prediction network structure in a backwards-compatible " "manner. Be careful not to use this method in production." ) super().serialize(path) def serialize_prediction_net(self, path: Path) -> None: export_repr_block(self.prediction_net, path, "prediction_net") @classmethod def deserialize( cls, path: Path, ctx: Optional[mx.Context] = None ) -> "RepresentableBlockPredictor": ctx = ctx if ctx is not None else get_mxnet_context() with mx.Context(ctx): # deserialize constructor parameters with (path / "parameters.json").open("r") as fp: parameters = load_json(fp.read()) # deserialize transformation chain with (path / "input_transform.json").open("r") as fp: transform = load_json(fp.read()) # deserialize prediction network prediction_net = import_repr_block(path, "prediction_net") # input_names is derived from the prediction_net if "input_names" in parameters: del parameters["input_names"] parameters["ctx"] = ctx return RepresentableBlockPredictor( input_transform=transform, prediction_net=prediction_net, **parameters, ) class WorkerError: def __init__(self, msg): self.msg = msg def _worker_loop( predictor_path: Path, input_queue: mp.Queue, output_queue: mp.Queue, worker_id, **kwargs, ): """ Worker loop for multiprocessing Predictor. Loads the predictor serialized in predictor_path reads inputs from input_queue and writes forecasts to output_queue """ predictor = Predictor.deserialize(predictor_path) while True: idx, data_chunk = input_queue.get() if idx is None: output_queue.put((None, None, None)) break try: result = list(predictor.predict(data_chunk, **kwargs)) except Exception: we = WorkerError( "".join(traceback.format_exception(*sys.exc_info())) ) output_queue.put((we, None, None)) break output_queue.put((idx, worker_id, result)) class ParallelizedPredictor(Predictor): """ Runs multiple instances (workers) of a predictor in parallel. Exceptions are propagated from the workers. Note: That there is currently an issue with tqdm that will cause things to hang if the ParallelizedPredictor is used with tqdm and an exception occurs during prediction. https://github.com/tqdm/tqdm/issues/548 Parameters ---------- base_predictor A representable predictor that will be used num_workers Number of workers (processes) to use. If set to None, one worker per CPU will be used. chunk_size Number of items to pass per call """ def __init__( self, base_predictor: Predictor, num_workers: Optional[int] = None, chunk_size=1, ) -> None: super().__init__(base_predictor.prediction_length, base_predictor.freq) self._base_predictor = base_predictor self._num_workers = ( num_workers if num_workers is not None else mp.cpu_count() ) self._chunk_size = chunk_size self._num_running_workers = 0 self._input_queues = [] self._output_queue = None def _grouper(self, iterable, n): iterator = iter(iterable) group = tuple(itertools.islice(iterator, n)) while group: yield group group = tuple(itertools.islice(iterator, n)) def terminate(self): for q in self._input_queues: q.put((None, None)) for w in self._workers: w.terminate() for i, w in enumerate(self._workers): w.join() def predict(self, dataset: Dataset, **kwargs) -> Iterator[Forecast]: with TemporaryDirectory() as tempdir: predictor_path = Path(tempdir) self._base_predictor.serialize(predictor_path) # TODO: Consider using shared memory for the data transfer. self._input_queues = [mp.Queue() for _ in range(self._num_workers)] self._output_queue = mp.Queue() workers = [] for worker_id, in_q in enumerate(self._input_queues): worker = mp.Process( target=_worker_loop, args=(predictor_path, in_q, self._output_queue, worker_id), kwargs=kwargs, ) worker.daemon = True worker.start() workers.append(worker) self._num_running_workers += 1 self._workers = workers chunked_data = self._grouper(dataset, self._chunk_size) self._send_idx = 0 self._next_idx = 0 self._data_buffer = {} worker_ids = list(range(self._num_workers)) def receive(): idx, worker_id, result = self._output_queue.get() if isinstance(idx, WorkerError): self._num_running_workers -= 1 self.terminate() raise Exception(idx.msg) if idx is not None: self._data_buffer[idx] = result return idx, worker_id, result def get_next_from_buffer(): while self._next_idx in self._data_buffer: result_batch = self._data_buffer.pop(self._next_idx) self._next_idx += 1 for result in result_batch: yield result def send(worker_id, chunk): q = self._input_queues[worker_id] q.put((self._send_idx, chunk)) self._send_idx += 1 try: # prime the queues for wid in worker_ids: chunk = next(chunked_data) send(wid, chunk) while True: idx, wid, result = receive() for res in get_next_from_buffer(): yield res chunk = next(chunked_data) send(wid, chunk) except StopIteration: # signal workers end of data for q in self._input_queues: q.put((None, None)) # collect any outstanding results while self._num_running_workers > 0: idx, worker_id, result = receive() if idx is None: self._num_running_workers -= 1 continue for res in get_next_from_buffer(): yield res assert len(self._data_buffer) == 0 assert self._send_idx == self._next_idx class Localizer(Predictor): """ A Predictor that uses an estimator to train a local model per time series and immediatly calls this to predict. Parameters ---------- estimator The estimator object to train on each dataset entry at prediction time. """ def __init__(self, estimator: "Estimator"): super().__init__(estimator.prediction_length, estimator.freq) self.estimator = estimator def predict(self, dataset: Dataset, **kwargs) -> Iterator[Forecast]: logger = logging.getLogger(__name__) for i, ts in enumerate(dataset, start=1): logger.info(f"training for time series {i} / {len(dataset)}") local_ds = ListDataset([ts], freq=self.freq) trained_pred = self.estimator.train(local_ds) logger.info(f"predicting for time series {i} / {len(dataset)}") predictions = trained_pred.predict(local_ds, **kwargs) for pred in predictions: yield pred class FallbackPredictor(Predictor): @classmethod def from_predictor( cls, base: RepresentablePredictor, **overrides ) -> Predictor: # Create predictor based on an existing predictor. # This let's us create a MeanPredictor as a fallback on the fly. return cls.from_hyperparameters( **getattr(base, "__init_args__"), **overrides ) def fallback(fallback_cls: Type[FallbackPredictor]): def decorator(predict_item): @functools.wraps(predict_item) def fallback_predict(self, item: DataEntry) -> Forecast: try: return predict_item(self, item) except GluonTSException: raise except Exception: logging.warning( f"Base predictor failed with: {traceback.format_exc()}" ) fallback_predictor = fallback_cls.from_predictor(self) return fallback_predictor.predict_item(item) return fallback_predict return decorator
23,995
30.994667
81
py
rankpredictor
rankpredictor-master/sub/gluonts/model/deepar-original/_network.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports import numpy as np from typing import List, Optional, Tuple # Third-party imports import mxnet as mx # First-party imports from gluonts.block.feature import FeatureEmbedder from gluonts.block.scaler import MeanScaler, NOPScaler from gluonts.core.component import DType, validated from gluonts.distribution import DistributionOutput, Distribution from gluonts.distribution.distribution import getF from gluonts.model.common import Tensor from gluonts.support.util import weighted_average def prod(xs): p = 1 for x in xs: p *= x return p class DeepARNetwork(mx.gluon.HybridBlock): @validated() def __init__( self, num_layers: int, num_cells: int, cell_type: str, history_length: int, context_length: int, prediction_length: int, distr_output: DistributionOutput, dropout_rate: float, cardinality: List[int], embedding_dimension: List[int], lags_seq: List[int], scaling: bool = True, dtype: DType = np.float32, **kwargs, ) -> None: super().__init__(**kwargs) self.num_layers = num_layers self.num_cells = num_cells self.cell_type = cell_type self.history_length = history_length self.context_length = context_length self.prediction_length = prediction_length self.dropout_rate = dropout_rate self.cardinality = cardinality self.embedding_dimension = embedding_dimension self.num_cat = len(cardinality) self.scaling = scaling self.dtype = dtype assert len(cardinality) == len( embedding_dimension ), "embedding_dimension should be a list with the same size as cardinality" assert len(set(lags_seq)) == len( lags_seq ), "no duplicated lags allowed!" lags_seq.sort() self.lags_seq = lags_seq self.distr_output = distr_output RnnCell = {"lstm": mx.gluon.rnn.LSTMCell, "gru": mx.gluon.rnn.GRUCell}[ self.cell_type ] self.target_shape = distr_output.event_shape # TODO: is the following restriction needed? assert ( len(self.target_shape) <= 1 ), "Argument `target_shape` should be a tuple with 1 element at most" with self.name_scope(): self.proj_distr_args = distr_output.get_args_proj() self.rnn = mx.gluon.rnn.HybridSequentialRNNCell() for k in range(num_layers): cell = RnnCell(hidden_size=num_cells) cell = mx.gluon.rnn.ResidualCell(cell) if k > 0 else cell cell = ( mx.gluon.rnn.ZoneoutCell(cell, zoneout_states=dropout_rate) if dropout_rate > 0.0 else cell ) self.rnn.add(cell) self.rnn.cast(dtype=dtype) self.embedder = FeatureEmbedder( cardinalities=cardinality, embedding_dims=embedding_dimension, dtype=self.dtype, ) if scaling: self.scaler = MeanScaler(keepdims=True) else: self.scaler = NOPScaler(keepdims=True) @staticmethod def get_lagged_subsequences( F, sequence: Tensor, sequence_length: int, indices: List[int], subsequences_length: int = 1, ) -> Tensor: """ Returns lagged subsequences of a given sequence. Parameters ---------- sequence : Tensor the sequence from which lagged subsequences should be extracted. Shape: (N, T, C). sequence_length : int length of sequence in the T (time) dimension (axis = 1). indices : List[int] list of lag indices to be used. subsequences_length : int length of the subsequences to be extracted. Returns -------- lagged : Tensor a tensor of shape (N, S, C, I), where S = subsequences_length and I = len(indices), containing lagged subsequences. Specifically, lagged[i, j, :, k] = sequence[i, -indices[k]-S+j, :]. """ # we must have: sequence_length - lag_index - subsequences_length >= 0 # for all lag_index, hence the following assert assert max(indices) + subsequences_length <= sequence_length, ( f"lags cannot go further than history length, " f"found lag {max(indices)} while history length is only " f"{sequence_length}" ) assert all(lag_index >= 0 for lag_index in indices) lagged_values = [] for lag_index in indices: begin_index = -lag_index - subsequences_length end_index = -lag_index if lag_index > 0 else None lagged_values.append( F.slice_axis( sequence, axis=1, begin=begin_index, end=end_index ) ) return F.stack(*lagged_values, axis=-1) def unroll_encoder( self, F, feat_static_cat: Tensor, # (batch_size, num_features) feat_static_real: Tensor, # (batch_size, num_features) past_time_feat: Tensor, # (batch_size, history_length, num_features) past_target: Tensor, # (batch_size, history_length, *target_shape) past_observed_values: Tensor, # (batch_size, history_length, *target_shape) future_time_feat: Optional[ Tensor ], # (batch_size, prediction_length, num_features) future_target: Optional[ Tensor ], # (batch_size, prediction_length, *target_shape) ) -> Tuple[Tensor, List, Tensor, Tensor]: """ Unrolls the LSTM encoder over past and, if present, future data. Returns outputs and state of the encoder, plus the scale of past_target and a vector of static features that was constructed and fed as input to the encoder. All tensor arguments should have NTC layout. """ if future_time_feat is None or future_target is None: time_feat = past_time_feat.slice_axis( axis=1, begin=self.history_length - self.context_length, end=None, ) sequence = past_target sequence_length = self.history_length subsequences_length = self.context_length else: time_feat = F.concat( past_time_feat.slice_axis( axis=1, begin=self.history_length - self.context_length, end=None, ), future_time_feat, dim=1, ) sequence = F.concat(past_target, future_target, dim=1) sequence_length = self.history_length + self.prediction_length subsequences_length = self.context_length + self.prediction_length # (batch_size, sub_seq_len, *target_shape, num_lags) lags = self.get_lagged_subsequences( F=F, sequence=sequence, sequence_length=sequence_length, indices=self.lags_seq, subsequences_length=subsequences_length, ) # scale is computed on the context length last units of the past target # scale shape is (batch_size, 1, *target_shape) _, scale = self.scaler( past_target.slice_axis( axis=1, begin=-self.context_length, end=None ), past_observed_values.slice_axis( axis=1, begin=-self.context_length, end=None ), ) # (batch_size, num_features) embedded_cat = self.embedder(feat_static_cat) # in addition to embedding features, use the log scale as it can help # prediction too # (batch_size, num_features + prod(target_shape)) static_feat = F.concat( embedded_cat, feat_static_real, F.log(scale) if len(self.target_shape) == 0 else F.log(scale.squeeze(axis=1)), dim=1, ) # (batch_size, subsequences_length, num_features + 1) repeated_static_feat = static_feat.expand_dims(axis=1).repeat( axis=1, repeats=subsequences_length ) # (batch_size, sub_seq_len, *target_shape, num_lags) lags_scaled = F.broadcast_div(lags, scale.expand_dims(axis=-1)) # from (batch_size, sub_seq_len, *target_shape, num_lags) # to (batch_size, sub_seq_len, prod(target_shape) * num_lags) input_lags = F.reshape( data=lags_scaled, shape=( -1, subsequences_length, len(self.lags_seq) * prod(self.target_shape), ), ) # (batch_size, sub_seq_len, input_dim) inputs = F.concat(input_lags, time_feat, repeated_static_feat, dim=-1) # unroll encoder outputs, state = self.rnn.unroll( inputs=inputs, length=subsequences_length, layout="NTC", merge_outputs=True, begin_state=self.rnn.begin_state( func=F.zeros, dtype=self.dtype, batch_size=inputs.shape[0] if isinstance(inputs, mx.nd.NDArray) else 0, ), ) # outputs: (batch_size, seq_len, num_cells) # state: list of (batch_size, num_cells) tensors # scale: (batch_size, 1, *target_shape) # static_feat: (batch_size, num_features + prod(target_shape)) return outputs, state, scale, static_feat class DeepARTrainingNetwork(DeepARNetwork): def distribution( self, feat_static_cat: Tensor, feat_static_real: Tensor, past_time_feat: Tensor, past_target: Tensor, past_observed_values: Tensor, future_time_feat: Tensor, future_target: Tensor, future_observed_values: Tensor, ) -> Distribution: """ Returns the distribution predicted by the model on the range of past_target and future_target. The distribution is obtained by unrolling the network with the true target, this is also the distribution that is being minimized during training. This can be used in anomaly detection, see for instance examples/anomaly_detection.py. Input arguments are the same as for the hybrid_forward method. Returns ------- Distribution a distribution object whose mean has shape: (batch_size, context_length + prediction_length). """ # unroll the decoder in "training mode" # i.e. by providing future data as well F = getF(feat_static_cat) rnn_outputs, _, scale, _ = self.unroll_encoder( F=F, feat_static_cat=feat_static_cat, feat_static_real=feat_static_real, past_time_feat=past_time_feat, past_target=past_target, past_observed_values=past_observed_values, future_time_feat=future_time_feat, future_target=future_target, ) distr_args = self.proj_distr_args(rnn_outputs) return self.distr_output.distribution(distr_args, scale=scale) # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, feat_static_cat: Tensor, feat_static_real: Tensor, past_time_feat: Tensor, past_target: Tensor, past_observed_values: Tensor, future_time_feat: Tensor, future_target: Tensor, future_observed_values: Tensor, ) -> Tensor: """ Computes the loss for training DeepAR, all inputs tensors representing time series have NTC layout. Parameters ---------- F feat_static_cat : (batch_size, num_features) feat_static_real : (batch_size, num_features) past_time_feat : (batch_size, history_length, num_features) past_target : (batch_size, history_length, *target_shape) past_observed_values : (batch_size, history_length, *target_shape, seq_len) future_time_feat : (batch_size, prediction_length, num_features) future_target : (batch_size, prediction_length, *target_shape) future_observed_values : (batch_size, prediction_length, *target_shape) Returns loss with shape (batch_size, context + prediction_length, 1) ------- """ distr = self.distribution( feat_static_cat=feat_static_cat, feat_static_real=feat_static_real, past_time_feat=past_time_feat, past_target=past_target, past_observed_values=past_observed_values, future_time_feat=future_time_feat, future_target=future_target, future_observed_values=future_observed_values, ) # put together target sequence # (batch_size, seq_len, *target_shape) target = F.concat( past_target.slice_axis( axis=1, begin=self.history_length - self.context_length, end=None, ), future_target, dim=1, ) # (batch_size, seq_len) loss = distr.loss(target) # (batch_size, seq_len, *target_shape) observed_values = F.concat( past_observed_values.slice_axis( axis=1, begin=self.history_length - self.context_length, end=self.history_length, ), future_observed_values, dim=1, ) # mask the loss at one time step iff one or more observations is missing in the target dimensions # (batch_size, seq_len) loss_weights = ( observed_values if (len(self.target_shape) == 0) else observed_values.min(axis=-1, keepdims=False) ) weighted_loss = weighted_average( F=F, x=loss, weights=loss_weights, axis=1 ) return weighted_loss, loss class DeepARPredictionNetwork(DeepARNetwork): @validated() def __init__(self, num_parallel_samples: int = 100, **kwargs) -> None: super().__init__(**kwargs) self.num_parallel_samples = num_parallel_samples # for decoding the lags are shifted by one, at the first time-step # of the decoder a lag of one corresponds to the last target value self.shifted_lags = [l - 1 for l in self.lags_seq] def sampling_decoder( self, F, static_feat: Tensor, past_target: Tensor, time_feat: Tensor, scale: Tensor, begin_states: List, ) -> Tensor: """ Computes sample paths by unrolling the LSTM starting with a initial input and state. Parameters ---------- static_feat : Tensor static features. Shape: (batch_size, num_static_features). past_target : Tensor target history. Shape: (batch_size, history_length). time_feat : Tensor time features. Shape: (batch_size, prediction_length, num_time_features). scale : Tensor tensor containing the scale of each element in the batch. Shape: (batch_size, 1, 1). begin_states : List list of initial states for the LSTM layers. the shape of each tensor of the list should be (batch_size, num_cells) Returns -------- Tensor A tensor containing sampled paths. Shape: (batch_size, num_sample_paths, prediction_length). """ # blows-up the dimension of each tensor to batch_size * self.num_parallel_samples for increasing parallelism repeated_past_target = past_target.repeat( repeats=self.num_parallel_samples, axis=0 ) repeated_time_feat = time_feat.repeat( repeats=self.num_parallel_samples, axis=0 ) repeated_static_feat = static_feat.repeat( repeats=self.num_parallel_samples, axis=0 ).expand_dims(axis=1) repeated_scale = scale.repeat( repeats=self.num_parallel_samples, axis=0 ) repeated_states = [ s.repeat(repeats=self.num_parallel_samples, axis=0) for s in begin_states ] future_samples = [] # for each future time-units we draw new samples for this time-unit and update the state for k in range(self.prediction_length): # (batch_size * num_samples, 1, *target_shape, num_lags) lags = self.get_lagged_subsequences( F=F, sequence=repeated_past_target, sequence_length=self.history_length + k, indices=self.shifted_lags, subsequences_length=1, ) # (batch_size * num_samples, 1, *target_shape, num_lags) lags_scaled = F.broadcast_div( lags, repeated_scale.expand_dims(axis=-1) ) # from (batch_size * num_samples, 1, *target_shape, num_lags) # to (batch_size * num_samples, 1, prod(target_shape) * num_lags) input_lags = F.reshape( data=lags_scaled, shape=(-1, 1, prod(self.target_shape) * len(self.lags_seq)), ) # (batch_size * num_samples, 1, prod(target_shape) * num_lags + num_time_features + num_static_features) decoder_input = F.concat( input_lags, repeated_time_feat.slice_axis(axis=1, begin=k, end=k + 1), repeated_static_feat, dim=-1, ) # output shape: (batch_size * num_samples, 1, num_cells) # state shape: (batch_size * num_samples, num_cells) rnn_outputs, repeated_states = self.rnn.unroll( inputs=decoder_input, length=1, begin_state=repeated_states, layout="NTC", merge_outputs=True, ) distr_args = self.proj_distr_args(rnn_outputs) # compute likelihood of target given the predicted parameters distr = self.distr_output.distribution( distr_args, scale=repeated_scale ) # (batch_size * num_samples, 1, *target_shape) new_samples = distr.sample(dtype=self.dtype) # (batch_size * num_samples, seq_len, *target_shape) repeated_past_target = F.concat( repeated_past_target, new_samples, dim=1 ) future_samples.append(new_samples) # (batch_size * num_samples, prediction_length, *target_shape) samples = F.concat(*future_samples, dim=1) # (batch_size, num_samples, prediction_length, *target_shape) return samples.reshape( shape=( (-1, self.num_parallel_samples) + (self.prediction_length,) + self.target_shape ) ) # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, feat_static_cat: Tensor, # (batch_size, num_features) feat_static_real: Tensor, # (batch_size, num_features) past_time_feat: Tensor, # (batch_size, history_length, num_features) past_target: Tensor, # (batch_size, history_length, *target_shape) past_observed_values: Tensor, # (batch_size, history_length, *target_shape) future_time_feat: Tensor, # (batch_size, prediction_length, num_features) ) -> Tensor: """ Predicts samples, all tensors should have NTC layout. Parameters ---------- F feat_static_cat : (batch_size, num_features) feat_static_real : (batch_size, num_features) past_time_feat : (batch_size, history_length, num_features) past_target : (batch_size, history_length, *target_shape) past_observed_values : (batch_size, history_length, *target_shape) future_time_feat : (batch_size, prediction_length, num_features) Returns ------- Tensor Predicted samples """ # unroll the decoder in "prediction mode", i.e. with past data only _, state, scale, static_feat = self.unroll_encoder( F=F, feat_static_cat=feat_static_cat, feat_static_real=feat_static_real, past_time_feat=past_time_feat, past_target=past_target, past_observed_values=past_observed_values, future_time_feat=None, future_target=None, ) return self.sampling_decoder( F=F, past_target=past_target, time_feat=future_time_feat, static_feat=static_feat, scale=scale, begin_states=state, )
21,761
34.85173
116
py
rankpredictor
rankpredictor-master/sub/gluonts/model/deepar-original/_estimator.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports import numpy as np from typing import List, Optional # Third-party imports from mxnet.gluon import HybridBlock # First-party imports from gluonts.core.component import DType, validated from gluonts.dataset.field_names import FieldName from gluonts.distribution import DistributionOutput, StudentTOutput from gluonts.model.estimator import GluonEstimator from gluonts.model.predictor import Predictor, RepresentableBlockPredictor from gluonts.support.util import copy_parameters from gluonts.time_feature import ( TimeFeature, time_features_from_frequency_str, get_lags_for_frequency, ) from gluonts.trainer import Trainer from gluonts.transform import ( AddAgeFeature, AddObservedValuesIndicator, AddTimeFeatures, AsNumpyArray, Chain, ExpectedNumInstanceSampler, InstanceSplitter, RemoveFields, SetField, Transformation, VstackFeatures, ) # Relative imports from ._network import DeepARPredictionNetwork, DeepARTrainingNetwork class DeepAREstimator(GluonEstimator): """ Construct a DeepAR estimator. This implements an RNN-based model, close to the one described in [SFG17]_. *Note:* the code of this model is unrelated to the implementation behind `SageMaker's DeepAR Forecasting Algorithm <https://docs.aws.amazon.com/sagemaker/latest/dg/deepar.html>`_. Parameters ---------- freq Frequency of the data to train on and predict prediction_length Length of the prediction horizon trainer Trainer object to be used (default: Trainer()) context_length Number of steps to unroll the RNN for before computing predictions (default: None, in which case context_length = prediction_length) num_layers Number of RNN layers (default: 2) num_cells Number of RNN cells for each layer (default: 40) cell_type Type of recurrent cells to use (available: 'lstm' or 'gru'; default: 'lstm') dropout_rate Dropout regularization parameter (default: 0.1) use_feat_dynamic_real Whether to use the ``feat_dynamic_real`` field from the data (default: False) use_feat_static_cat Whether to use the ``feat_static_cat`` field from the data (default: False) use_feat_static_real Whether to use the ``feat_static_real`` field from the data (default: False) cardinality Number of values of each categorical feature. This must be set if ``use_feat_static_cat == True`` (default: None) embedding_dimension Dimension of the embeddings for categorical features (default: [min(50, (cat+1)//2) for cat in cardinality]) distr_output Distribution to use to evaluate observations and sample predictions (default: StudentTOutput()) scaling Whether to automatically scale the target values (default: true) lags_seq Indices of the lagged target values to use as inputs of the RNN (default: None, in which case these are automatically determined based on freq) time_features Time features to use as inputs of the RNN (default: None, in which case these are automatically determined based on freq) num_parallel_samples Number of evaluation samples per time series to increase parallelism during inference. This is a model optimization that does not affect the accuracy (default: 100) """ @validated() def __init__( self, freq: str, prediction_length: int, trainer: Trainer = Trainer(), context_length: Optional[int] = None, num_layers: int = 2, num_cells: int = 40, cell_type: str = "lstm", dropout_rate: float = 0.1, use_feat_dynamic_real: bool = False, use_feat_static_cat: bool = False, use_feat_static_real: bool = False, cardinality: Optional[List[int]] = None, embedding_dimension: Optional[List[int]] = None, distr_output: DistributionOutput = StudentTOutput(), scaling: bool = True, lags_seq: Optional[List[int]] = None, time_features: Optional[List[TimeFeature]] = None, num_parallel_samples: int = 100, dtype: DType = np.float32, ) -> None: super().__init__(trainer=trainer, dtype=dtype) assert ( prediction_length > 0 ), "The value of `prediction_length` should be > 0" assert ( context_length is None or context_length > 0 ), "The value of `context_length` should be > 0" assert num_layers > 0, "The value of `num_layers` should be > 0" assert num_cells > 0, "The value of `num_cells` should be > 0" assert dropout_rate >= 0, "The value of `dropout_rate` should be >= 0" assert (cardinality is not None and use_feat_static_cat) or ( cardinality is None and not use_feat_static_cat ), "You should set `cardinality` if and only if `use_feat_static_cat=True`" assert cardinality is None or all( [c > 0 for c in cardinality] ), "Elements of `cardinality` should be > 0" assert embedding_dimension is None or all( [e > 0 for e in embedding_dimension] ), "Elements of `embedding_dimension` should be > 0" assert ( num_parallel_samples > 0 ), "The value of `num_parallel_samples` should be > 0" self.freq = freq self.context_length = ( context_length if context_length is not None else prediction_length ) self.prediction_length = prediction_length self.distr_output = distr_output self.distr_output.dtype = dtype self.num_layers = num_layers self.num_cells = num_cells self.cell_type = cell_type self.dropout_rate = dropout_rate self.use_feat_dynamic_real = use_feat_dynamic_real self.use_feat_static_cat = use_feat_static_cat self.use_feat_static_real = use_feat_static_real self.cardinality = ( cardinality if cardinality and use_feat_static_cat else [1] ) self.embedding_dimension = ( embedding_dimension if embedding_dimension is not None else [min(50, (cat + 1) // 2) for cat in self.cardinality] ) self.scaling = scaling self.lags_seq = ( lags_seq if lags_seq is not None else get_lags_for_frequency(freq_str=freq) ) self.time_features = ( time_features if time_features is not None else time_features_from_frequency_str(self.freq) ) self.history_length = self.context_length + max(self.lags_seq) self.num_parallel_samples = num_parallel_samples def create_transformation(self) -> Transformation: remove_field_names = [FieldName.FEAT_DYNAMIC_CAT] if not self.use_feat_static_real: remove_field_names.append(FieldName.FEAT_STATIC_REAL) if not self.use_feat_dynamic_real: remove_field_names.append(FieldName.FEAT_DYNAMIC_REAL) return Chain( [RemoveFields(field_names=remove_field_names)] + ( [SetField(output_field=FieldName.FEAT_STATIC_CAT, value=[0.0])] if not self.use_feat_static_cat else [] ) + ( [ SetField( output_field=FieldName.FEAT_STATIC_REAL, value=[0.0] ) ] if not self.use_feat_static_real else [] ) + [ AsNumpyArray( field=FieldName.FEAT_STATIC_CAT, expected_ndim=1, dtype=self.dtype, ), AsNumpyArray( field=FieldName.FEAT_STATIC_REAL, expected_ndim=1, dtype=self.dtype, ), AsNumpyArray( field=FieldName.TARGET, # in the following line, we add 1 for the time dimension expected_ndim=1 + len(self.distr_output.event_shape), dtype=self.dtype, ), AddObservedValuesIndicator( target_field=FieldName.TARGET, output_field=FieldName.OBSERVED_VALUES, dtype=self.dtype, ), AddTimeFeatures( start_field=FieldName.START, target_field=FieldName.TARGET, output_field=FieldName.FEAT_TIME, time_features=self.time_features, pred_length=self.prediction_length, ), AddAgeFeature( target_field=FieldName.TARGET, output_field=FieldName.FEAT_AGE, pred_length=self.prediction_length, log_scale=True, dtype=self.dtype, ), VstackFeatures( output_field=FieldName.FEAT_TIME, input_fields=[FieldName.FEAT_TIME, FieldName.FEAT_AGE] + ( [FieldName.FEAT_DYNAMIC_REAL] if self.use_feat_dynamic_real else [] ), ), InstanceSplitter( target_field=FieldName.TARGET, is_pad_field=FieldName.IS_PAD, start_field=FieldName.START, forecast_start_field=FieldName.FORECAST_START, train_sampler=ExpectedNumInstanceSampler(num_instances=1), past_length=self.history_length, future_length=self.prediction_length, time_series_fields=[ FieldName.FEAT_TIME, FieldName.OBSERVED_VALUES, ], ), ] ) def create_training_network(self) -> DeepARTrainingNetwork: return DeepARTrainingNetwork( num_layers=self.num_layers, num_cells=self.num_cells, cell_type=self.cell_type, history_length=self.history_length, context_length=self.context_length, prediction_length=self.prediction_length, distr_output=self.distr_output, dropout_rate=self.dropout_rate, cardinality=self.cardinality, embedding_dimension=self.embedding_dimension, lags_seq=self.lags_seq, scaling=self.scaling, dtype=self.dtype, ) def create_predictor( self, transformation: Transformation, trained_network: HybridBlock ) -> Predictor: prediction_network = DeepARPredictionNetwork( num_parallel_samples=self.num_parallel_samples, num_layers=self.num_layers, num_cells=self.num_cells, cell_type=self.cell_type, history_length=self.history_length, context_length=self.context_length, prediction_length=self.prediction_length, distr_output=self.distr_output, dropout_rate=self.dropout_rate, cardinality=self.cardinality, embedding_dimension=self.embedding_dimension, lags_seq=self.lags_seq, scaling=self.scaling, dtype=self.dtype, ) copy_parameters(trained_network, prediction_network) return RepresentableBlockPredictor( input_transform=transformation, prediction_net=prediction_network, batch_size=self.trainer.batch_size, freq=self.freq, prediction_length=self.prediction_length, ctx=self.trainer.ctx, dtype=self.dtype, )
12,645
37.090361
94
py
rankpredictor
rankpredictor-master/sub/gluonts/model/simple_feedforward/_network.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import List # Third-party imports import mxnet as mx # First-party imports from gluonts.block.scaler import MeanScaler, NOPScaler from gluonts.core.component import validated from gluonts.distribution import Distribution, DistributionOutput from gluonts.model.common import Tensor class SimpleFeedForwardNetworkBase(mx.gluon.HybridBlock): """ Abstract base class to implement feed-forward networks for probabilistic time series prediction. This class does not implement hybrid_forward: this is delegated to the two subclasses SimpleFeedForwardTrainingNetwork and SimpleFeedForwardPredictionNetwork, that define respectively how to compute the loss and how to generate predictions. Parameters ---------- num_hidden_dimensions Number of hidden nodes in each layer. prediction_length Number of time units to predict. context_length Number of time units that condition the predictions. batch_normalization Whether to use batch normalization. mean_scaling Scale the network input by the data mean and the network output by its inverse. distr_output Distribution to fit. kwargs """ # Needs the validated decorator so that arguments types are checked and # the block can be serialized. @validated() def __init__( self, num_hidden_dimensions: List[int], prediction_length: int, context_length: int, batch_normalization: bool, mean_scaling: bool, distr_output: DistributionOutput, **kwargs, ) -> None: super().__init__(**kwargs) self.num_hidden_dimensions = num_hidden_dimensions self.prediction_length = prediction_length self.context_length = context_length self.batch_normalization = batch_normalization self.mean_scaling = mean_scaling self.distr_output = distr_output with self.name_scope(): self.distr_args_proj = self.distr_output.get_args_proj() self.mlp = mx.gluon.nn.HybridSequential() dims = self.num_hidden_dimensions for layer_no, units in enumerate(dims[:-1]): self.mlp.add(mx.gluon.nn.Dense(units=units, activation="relu")) if self.batch_normalization: self.mlp.add(mx.gluon.nn.BatchNorm()) self.mlp.add(mx.gluon.nn.Dense(units=prediction_length * dims[-1])) self.mlp.add( mx.gluon.nn.HybridLambda( lambda F, o: F.reshape( o, (-1, prediction_length, dims[-1]) ) ) ) self.scaler = MeanScaler() if mean_scaling else NOPScaler() def get_distr(self, F, past_target: Tensor) -> Distribution: """ Given past target values, applies the feed-forward network and maps the output to a probability distribution for future observations. Parameters ---------- F past_target Tensor containing past target observations. Shape: (batch_size, context_length, target_dim). Returns ------- Distribution The predicted probability distribution for future observations. """ # (batch_size, seq_len, target_dim) and (batch_size, seq_len, target_dim) scaled_target, target_scale = self.scaler( past_target, F.ones_like(past_target), # TODO: pass the actual observed here ) mlp_outputs = self.mlp(scaled_target) distr_args = self.distr_args_proj(mlp_outputs) return self.distr_output.distribution( distr_args, scale=target_scale.expand_dims(axis=1) ) class SimpleFeedForwardTrainingNetwork(SimpleFeedForwardNetworkBase): # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, past_target: Tensor, future_target: Tensor ) -> Tensor: """ Computes a probability distribution for future data given the past, and returns the loss associated with the actual future observations. Parameters ---------- F past_target Tensor with past observations. Shape: (batch_size, context_length, target_dim). future_target Tensor with future observations. Shape: (batch_size, prediction_length, target_dim). Returns ------- Tensor Loss tensor. Shape: (batch_size, ). """ distr = self.get_distr(F, past_target) # (batch_size, prediction_length, target_dim) loss = distr.loss(future_target) # (batch_size, ) return loss.mean(axis=1) class SimpleFeedForwardPredictionNetwork(SimpleFeedForwardNetworkBase): @validated() def __init__( self, num_parallel_samples: int = 100, *args, **kwargs ) -> None: super().__init__(*args, **kwargs) self.num_parallel_samples = num_parallel_samples # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward(self, F, past_target: Tensor) -> Tensor: """ Computes a probability distribution for future data given the past, and draws samples from it. Parameters ---------- F past_target Tensor with past observations. Shape: (batch_size, context_length, target_dim). Returns ------- Tensor Prediction sample. Shape: (samples, batch_size, prediction_length). """ distr = self.get_distr(F, past_target) # (num_samples, batch_size, prediction_length) samples = distr.sample(self.num_parallel_samples) # (batch_size, num_samples, prediction_length) return samples.swapaxes(0, 1)
6,491
32.989529
81
py
rankpredictor
rankpredictor-master/sub/gluonts/model/simple_feedforward/_estimator.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import List, Optional # Third-party imports from mxnet.gluon import HybridBlock # First-party imports from gluonts.core.component import validated from gluonts.dataset.field_names import FieldName from gluonts.distribution import DistributionOutput, StudentTOutput from gluonts.model.estimator import GluonEstimator from gluonts.model.predictor import Predictor, RepresentableBlockPredictor from gluonts.trainer import Trainer from gluonts.transform import ( Chain, ExpectedNumInstanceSampler, InstanceSplitter, Transformation, ) # Relative imports from ._network import ( SimpleFeedForwardPredictionNetwork, SimpleFeedForwardTrainingNetwork, ) class SimpleFeedForwardEstimator(GluonEstimator): """ SimpleFeedForwardEstimator shows how to build a simple MLP model predicting the next target time-steps given the previous ones. Given that we want to define a gluon model trainable by SGD, we inherit the parent class `GluonEstimator` that handles most of the logic for fitting a neural-network. We thus only have to define: 1. How the data is transformed before being fed to our model:: def create_transformation(self) -> Transformation 2. How the training happens:: def create_training_network(self) -> HybridBlock 3. how the predictions can be made for a batch given a trained network:: def create_predictor( self, transformation: Transformation, trained_net: HybridBlock, ) -> Predictor Parameters ---------- freq Time time granularity of the data prediction_length Length of the prediction horizon trainer Trainer object to be used (default: Trainer()) num_hidden_dimensions Number of hidden nodes in each layer (default: [40, 40]) context_length Number of time units that condition the predictions (default: None, in which case context_length = prediction_length) distr_output Distribution to fit (default: StudentTOutput()) batch_normalization Whether to use batch normalization (default: False) mean_scaling Scale the network input by the data mean and the network output by its inverse (default: True) num_parallel_samples Number of evaluation samples per time series to increase parallelism during inference. This is a model optimization that does not affect the accuracy (default: 100) """ # The validated() decorator makes sure that parameters are checked by # Pydantic and allows to serialize/print models. Note that all parameters # have defaults except for `freq` and `prediction_length`. which is # recommended in GluonTS to allow to compare models easily. @validated() def __init__( self, freq: str, prediction_length: int, trainer: Trainer = Trainer(), num_hidden_dimensions: Optional[List[int]] = None, context_length: Optional[int] = None, distr_output: DistributionOutput = StudentTOutput(), batch_normalization: bool = False, mean_scaling: bool = True, num_parallel_samples: int = 100, ) -> None: """ Defines an estimator. All parameters should be serializable. """ super().__init__(trainer=trainer) assert ( prediction_length > 0 ), "The value of `prediction_length` should be > 0" assert ( context_length is None or context_length > 0 ), "The value of `context_length` should be > 0" assert num_hidden_dimensions is None or ( [d > 0 for d in num_hidden_dimensions] ), "Elements of `num_hidden_dimensions` should be > 0" assert ( num_parallel_samples > 0 ), "The value of `num_parallel_samples` should be > 0" self.num_hidden_dimensions = ( num_hidden_dimensions if num_hidden_dimensions is not None else list([40, 40]) ) self.prediction_length = prediction_length self.context_length = ( context_length if context_length is not None else prediction_length ) self.freq = freq self.distr_output = distr_output self.batch_normalization = batch_normalization self.mean_scaling = mean_scaling self.num_parallel_samples = num_parallel_samples # here we do only a simple operation to convert the input data to a form # that can be digested by our model by only splitting the target in two, a # conditioning part and a to-predict part, for each training example. # fFr a more complex transformation example, see the `gluonts.model.deepar` # transformation that includes time features, age feature, observed values # indicator, ... def create_transformation(self) -> Transformation: return Chain( [ InstanceSplitter( target_field=FieldName.TARGET, is_pad_field=FieldName.IS_PAD, start_field=FieldName.START, forecast_start_field=FieldName.FORECAST_START, train_sampler=ExpectedNumInstanceSampler(num_instances=1), past_length=self.context_length, future_length=self.prediction_length, time_series_fields=[], # [FieldName.FEAT_DYNAMIC_REAL] ) ] ) # defines the network, we get to see one batch to initialize it. # the network should return at least one tensor that is used as a loss to minimize in the training loop. # several tensors can be returned for instance for analysis, see DeepARTrainingNetwork for an example. def create_training_network(self) -> HybridBlock: return SimpleFeedForwardTrainingNetwork( num_hidden_dimensions=self.num_hidden_dimensions, prediction_length=self.prediction_length, context_length=self.context_length, distr_output=self.distr_output, batch_normalization=self.batch_normalization, mean_scaling=self.mean_scaling, ) # we now define how the prediction happens given that we are provided a # training network. def create_predictor( self, transformation: Transformation, trained_network: HybridBlock ) -> Predictor: prediction_network = SimpleFeedForwardPredictionNetwork( num_hidden_dimensions=self.num_hidden_dimensions, prediction_length=self.prediction_length, context_length=self.context_length, distr_output=self.distr_output, batch_normalization=self.batch_normalization, mean_scaling=self.mean_scaling, params=trained_network.collect_params(), num_parallel_samples=self.num_parallel_samples, ) return RepresentableBlockPredictor( input_transform=transformation, prediction_net=prediction_network, batch_size=self.trainer.batch_size, freq=self.freq, prediction_length=self.prediction_length, ctx=self.trainer.ctx, )
7,852
37.684729
108
py
rankpredictor
rankpredictor-master/sub/gluonts/model/deepar-inuse/_network.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports import numpy as np from typing import List, Optional, Tuple # Third-party imports import mxnet as mx # First-party imports from gluonts.block.feature import FeatureEmbedder from gluonts.block.scaler import MeanScaler, NOPScaler from gluonts.core.component import DType, validated from gluonts.distribution import DistributionOutput, Distribution from gluonts.distribution.distribution import getF from gluonts.model.common import Tensor from gluonts.support.util import weighted_average def prod(xs): p = 1 for x in xs: p *= x return p class DeepARNetwork(mx.gluon.HybridBlock): @validated() def __init__( self, num_layers: int, num_cells: int, cell_type: str, history_length: int, context_length: int, prediction_length: int, distr_output: DistributionOutput, dropout_rate: float, cardinality: List[int], embedding_dimension: List[int], lags_seq: List[int], scaling: bool = True, dtype: DType = np.float32, **kwargs, ) -> None: super().__init__(**kwargs) self.num_layers = num_layers self.num_cells = num_cells self.cell_type = cell_type self.history_length = history_length self.context_length = context_length self.prediction_length = prediction_length self.dropout_rate = dropout_rate self.cardinality = cardinality self.embedding_dimension = embedding_dimension self.num_cat = len(cardinality) self.scaling = scaling self.dtype = dtype assert len(cardinality) == len( embedding_dimension ), "embedding_dimension should be a list with the same size as cardinality" assert len(set(lags_seq)) == len( lags_seq ), "no duplicated lags allowed!" lags_seq.sort() self.lags_seq = lags_seq self.distr_output = distr_output RnnCell = {"lstm": mx.gluon.rnn.LSTMCell, "gru": mx.gluon.rnn.GRUCell}[ self.cell_type ] self.target_shape = distr_output.event_shape # TODO: is the following restriction needed? assert ( len(self.target_shape) <= 1 ), "Argument `target_shape` should be a tuple with 1 element at most" with self.name_scope(): self.proj_distr_args = distr_output.get_args_proj() self.rnn = mx.gluon.rnn.HybridSequentialRNNCell() for k in range(num_layers): cell = RnnCell(hidden_size=num_cells) cell = mx.gluon.rnn.ResidualCell(cell) if k > 0 else cell cell = ( mx.gluon.rnn.ZoneoutCell(cell, zoneout_states=dropout_rate) if dropout_rate > 0.0 else cell ) self.rnn.add(cell) self.rnn.cast(dtype=dtype) self.embedder = FeatureEmbedder( cardinalities=cardinality, embedding_dims=embedding_dimension, dtype=self.dtype, ) if scaling: self.scaler = MeanScaler(keepdims=True) else: self.scaler = NOPScaler(keepdims=True) @staticmethod def get_lagged_subsequences( F, sequence: Tensor, sequence_length: int, indices: List[int], subsequences_length: int = 1, ) -> Tensor: """ Returns lagged subsequences of a given sequence. Parameters ---------- sequence : Tensor the sequence from which lagged subsequences should be extracted. Shape: (N, T, C). sequence_length : int length of sequence in the T (time) dimension (axis = 1). indices : List[int] list of lag indices to be used. subsequences_length : int length of the subsequences to be extracted. Returns -------- lagged : Tensor a tensor of shape (N, S, C, I), where S = subsequences_length and I = len(indices), containing lagged subsequences. Specifically, lagged[i, j, :, k] = sequence[i, -indices[k]-S+j, :]. """ # we must have: sequence_length - lag_index - subsequences_length >= 0 # for all lag_index, hence the following assert assert max(indices) + subsequences_length <= sequence_length, ( f"lags cannot go further than history length, " f"found lag {max(indices)} while history length is only " f"{sequence_length}" ) assert all(lag_index >= 0 for lag_index in indices) lagged_values = [] for lag_index in indices: begin_index = -lag_index - subsequences_length end_index = -lag_index if lag_index > 0 else None lagged_values.append( F.slice_axis( sequence, axis=1, begin=begin_index, end=end_index ) ) return F.stack(*lagged_values, axis=-1) def unroll_encoder( self, F, feat_static_cat: Tensor, # (batch_size, num_features) feat_static_real: Tensor, # (batch_size, num_features) past_time_feat: Tensor, # (batch_size, history_length, num_features) past_target: Tensor, # (batch_size, history_length, *target_shape) past_observed_values: Tensor, # (batch_size, history_length, *target_shape) future_time_feat: Optional[ Tensor ], # (batch_size, prediction_length, num_features) future_target: Optional[ Tensor ], # (batch_size, prediction_length, *target_shape) ) -> Tuple[Tensor, List, Tensor, Tensor]: """ Unrolls the LSTM encoder over past and, if present, future data. Returns outputs and state of the encoder, plus the scale of past_target and a vector of static features that was constructed and fed as input to the encoder. All tensor arguments should have NTC layout. """ if future_time_feat is None or future_target is None: time_feat = past_time_feat.slice_axis( axis=1, begin=self.history_length - self.context_length, end=None, ) sequence = past_target sequence_length = self.history_length subsequences_length = self.context_length else: time_feat = F.concat( past_time_feat.slice_axis( axis=1, begin=self.history_length - self.context_length, end=None, ), future_time_feat, dim=1, ) sequence = F.concat(past_target, future_target, dim=1) sequence_length = self.history_length + self.prediction_length subsequences_length = self.context_length + self.prediction_length # (batch_size, sub_seq_len, *target_shape, num_lags) lags = self.get_lagged_subsequences( F=F, sequence=sequence, sequence_length=sequence_length, indices=self.lags_seq, subsequences_length=subsequences_length, ) # scale is computed on the context length last units of the past target # scale shape is (batch_size, 1, *target_shape) _, scale = self.scaler( past_target.slice_axis( axis=1, begin=-self.context_length, end=None ), past_observed_values.slice_axis( axis=1, begin=-self.context_length, end=None ), ) # (batch_size, num_features) embedded_cat = self.embedder(feat_static_cat) # in addition to embedding features, use the log scale as it can help # prediction too # (batch_size, num_features + prod(target_shape)) static_feat = F.concat( embedded_cat, feat_static_real, F.log(scale) if len(self.target_shape) == 0 else F.log(scale.squeeze(axis=1)), dim=1, ) # (batch_size, subsequences_length, num_features + 1) repeated_static_feat = static_feat.expand_dims(axis=1).repeat( axis=1, repeats=subsequences_length ) # (batch_size, sub_seq_len, *target_shape, num_lags) lags_scaled = F.broadcast_div(lags, scale.expand_dims(axis=-1)) # from (batch_size, sub_seq_len, *target_shape, num_lags) # to (batch_size, sub_seq_len, prod(target_shape) * num_lags) input_lags = F.reshape( data=lags_scaled, shape=( -1, subsequences_length, len(self.lags_seq) * prod(self.target_shape), ), ) # (batch_size, sub_seq_len, input_dim) inputs = F.concat(input_lags, time_feat, repeated_static_feat, dim=-1) # unroll encoder outputs, state = self.rnn.unroll( inputs=inputs, length=subsequences_length, layout="NTC", merge_outputs=True, begin_state=self.rnn.begin_state( func=F.zeros, dtype=self.dtype, batch_size=inputs.shape[0] if isinstance(inputs, mx.nd.NDArray) else 0, ), ) # outputs: (batch_size, seq_len, num_cells) # state: list of (batch_size, num_cells) tensors # scale: (batch_size, 1, *target_shape) # static_feat: (batch_size, num_features + prod(target_shape)) return outputs, state, scale, static_feat class DeepARTrainingNetwork(DeepARNetwork): def distribution( self, feat_static_cat: Tensor, feat_static_real: Tensor, past_time_feat: Tensor, past_target: Tensor, past_observed_values: Tensor, future_time_feat: Tensor, future_target: Tensor, future_observed_values: Tensor, ) -> Distribution: """ Returns the distribution predicted by the model on the range of past_target and future_target. The distribution is obtained by unrolling the network with the true target, this is also the distribution that is being minimized during training. This can be used in anomaly detection, see for instance examples/anomaly_detection.py. Input arguments are the same as for the hybrid_forward method. Returns ------- Distribution a distribution object whose mean has shape: (batch_size, context_length + prediction_length). """ # unroll the decoder in "training mode" # i.e. by providing future data as well F = getF(feat_static_cat) rnn_outputs, _, scale, _ = self.unroll_encoder( F=F, feat_static_cat=feat_static_cat, feat_static_real=feat_static_real, past_time_feat=past_time_feat, past_target=past_target, past_observed_values=past_observed_values, future_time_feat=future_time_feat, future_target=future_target, ) distr_args = self.proj_distr_args(rnn_outputs) return self.distr_output.distribution(distr_args, scale=scale) # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, feat_static_cat: Tensor, feat_static_real: Tensor, past_time_feat: Tensor, past_target: Tensor, past_observed_values: Tensor, future_time_feat: Tensor, future_target: Tensor, future_observed_values: Tensor, ) -> Tensor: """ Computes the loss for training DeepAR, all inputs tensors representing time series have NTC layout. Parameters ---------- F feat_static_cat : (batch_size, num_features) feat_static_real : (batch_size, num_features) past_time_feat : (batch_size, history_length, num_features) past_target : (batch_size, history_length, *target_shape) past_observed_values : (batch_size, history_length, *target_shape, seq_len) future_time_feat : (batch_size, prediction_length, num_features) future_target : (batch_size, prediction_length, *target_shape) future_observed_values : (batch_size, prediction_length, *target_shape) Returns loss with shape (batch_size, context + prediction_length, 1) ------- """ distr = self.distribution( feat_static_cat=feat_static_cat, feat_static_real=feat_static_real, past_time_feat=past_time_feat, past_target=past_target, past_observed_values=past_observed_values, future_time_feat=future_time_feat, future_target=future_target, future_observed_values=future_observed_values, ) # put together target sequence # (batch_size, seq_len, *target_shape) target = F.concat( past_target.slice_axis( axis=1, begin=self.history_length - self.context_length, end=None, ), future_target, dim=1, ) # (batch_size, seq_len) loss = distr.loss(target) # (batch_size, seq_len, *target_shape) observed_values = F.concat( past_observed_values.slice_axis( axis=1, begin=self.history_length - self.context_length, end=self.history_length, ), future_observed_values, dim=1, ) # mask the loss at one time step iff one or more observations is missing in the target dimensions # (batch_size, seq_len) loss_weights = ( observed_values if (len(self.target_shape) == 0) else observed_values.min(axis=-1, keepdims=False) ) weighted_loss = weighted_average( F=F, x=loss, weights=loss_weights, axis=1 ) return weighted_loss, loss class DeepARPredictionNetwork(DeepARNetwork): @validated() def __init__(self, num_parallel_samples: int = 100, **kwargs) -> None: super().__init__(**kwargs) self.num_parallel_samples = num_parallel_samples # for decoding the lags are shifted by one, at the first time-step # of the decoder a lag of one corresponds to the last target value self.shifted_lags = [l - 1 for l in self.lags_seq] def sampling_decoder( self, F, static_feat: Tensor, past_target: Tensor, time_feat: Tensor, scale: Tensor, begin_states: List, ) -> Tensor: """ Computes sample paths by unrolling the LSTM starting with a initial input and state. Parameters ---------- static_feat : Tensor static features. Shape: (batch_size, num_static_features). past_target : Tensor target history. Shape: (batch_size, history_length). time_feat : Tensor time features. Shape: (batch_size, prediction_length, num_time_features). scale : Tensor tensor containing the scale of each element in the batch. Shape: (batch_size, 1, 1). begin_states : List list of initial states for the LSTM layers. the shape of each tensor of the list should be (batch_size, num_cells) Returns -------- Tensor A tensor containing sampled paths. Shape: (batch_size, num_sample_paths, prediction_length). """ # blows-up the dimension of each tensor to batch_size * self.num_parallel_samples for increasing parallelism repeated_past_target = past_target.repeat( repeats=self.num_parallel_samples, axis=0 ) repeated_time_feat = time_feat.repeat( repeats=self.num_parallel_samples, axis=0 ) repeated_static_feat = static_feat.repeat( repeats=self.num_parallel_samples, axis=0 ).expand_dims(axis=1) repeated_scale = scale.repeat( repeats=self.num_parallel_samples, axis=0 ) repeated_states = [ s.repeat(repeats=self.num_parallel_samples, axis=0) for s in begin_states ] future_samples = [] # for each future time-units we draw new samples for this time-unit and update the state for k in range(self.prediction_length): # (batch_size * num_samples, 1, *target_shape, num_lags) lags = self.get_lagged_subsequences( F=F, sequence=repeated_past_target, sequence_length=self.history_length + k, indices=self.shifted_lags, subsequences_length=1, ) # (batch_size * num_samples, 1, *target_shape, num_lags) lags_scaled = F.broadcast_div( lags, repeated_scale.expand_dims(axis=-1) ) # from (batch_size * num_samples, 1, *target_shape, num_lags) # to (batch_size * num_samples, 1, prod(target_shape) * num_lags) input_lags = F.reshape( data=lags_scaled, shape=(-1, 1, prod(self.target_shape) * len(self.lags_seq)), ) # (batch_size * num_samples, 1, prod(target_shape) * num_lags + num_time_features + num_static_features) decoder_input = F.concat( input_lags, repeated_time_feat.slice_axis(axis=1, begin=k, end=k + 1), repeated_static_feat, dim=-1, ) # output shape: (batch_size * num_samples, 1, num_cells) # state shape: (batch_size * num_samples, num_cells) rnn_outputs, repeated_states = self.rnn.unroll( inputs=decoder_input, length=1, begin_state=repeated_states, layout="NTC", merge_outputs=True, ) distr_args = self.proj_distr_args(rnn_outputs) # compute likelihood of target given the predicted parameters distr = self.distr_output.distribution( distr_args, scale=repeated_scale ) # (batch_size * num_samples, 1, *target_shape) new_samples = distr.sample(dtype=self.dtype) # (batch_size * num_samples, seq_len, *target_shape) repeated_past_target = F.concat( repeated_past_target, new_samples, dim=1 ) future_samples.append(new_samples) # (batch_size * num_samples, prediction_length, *target_shape) samples = F.concat(*future_samples, dim=1) # (batch_size, num_samples, prediction_length, *target_shape) return samples.reshape( shape=( (-1, self.num_parallel_samples) + (self.prediction_length,) + self.target_shape ) ) # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, feat_static_cat: Tensor, # (batch_size, num_features) feat_static_real: Tensor, # (batch_size, num_features) past_time_feat: Tensor, # (batch_size, history_length, num_features) past_target: Tensor, # (batch_size, history_length, *target_shape) past_observed_values: Tensor, # (batch_size, history_length, *target_shape) future_time_feat: Tensor, # (batch_size, prediction_length, num_features) ) -> Tensor: """ Predicts samples, all tensors should have NTC layout. Parameters ---------- F feat_static_cat : (batch_size, num_features) feat_static_real : (batch_size, num_features) past_time_feat : (batch_size, history_length, num_features) past_target : (batch_size, history_length, *target_shape) past_observed_values : (batch_size, history_length, *target_shape) future_time_feat : (batch_size, prediction_length, num_features) Returns ------- Tensor Predicted samples """ # unroll the decoder in "prediction mode", i.e. with past data only _, state, scale, static_feat = self.unroll_encoder( F=F, feat_static_cat=feat_static_cat, feat_static_real=feat_static_real, past_time_feat=past_time_feat, past_target=past_target, past_observed_values=past_observed_values, future_time_feat=None, future_target=None, ) return self.sampling_decoder( F=F, past_target=past_target, time_feat=future_time_feat, static_feat=static_feat, scale=scale, begin_states=state, )
21,761
34.85173
116
py
rankpredictor
rankpredictor-master/sub/gluonts/model/deepar-inuse/_estimator.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports import numpy as np from typing import List, Optional # Third-party imports from mxnet.gluon import HybridBlock # First-party imports from gluonts.core.component import DType, validated from gluonts.dataset.field_names import FieldName from gluonts.distribution import DistributionOutput, StudentTOutput from gluonts.model.estimator import GluonEstimator from gluonts.model.predictor import Predictor, RepresentableBlockPredictor from gluonts.support.util import copy_parameters from gluonts.time_feature import ( TimeFeature, time_features_from_frequency_str, get_lags_for_frequency, ) from gluonts.trainer import Trainer from gluonts.transform import ( AddAgeFeature, AddObservedValuesIndicator, AddTimeFeatures, AsNumpyArray, Chain, ExpectedNumInstanceSampler, InstanceSplitter, RemoveFields, SetField, Transformation, VstackFeatures, ) # Relative imports from ._network import DeepARPredictionNetwork, DeepARTrainingNetwork class DeepAREstimator(GluonEstimator): """ Construct a DeepAR estimator. This implements an RNN-based model, close to the one described in [SFG17]_. *Note:* the code of this model is unrelated to the implementation behind `SageMaker's DeepAR Forecasting Algorithm <https://docs.aws.amazon.com/sagemaker/latest/dg/deepar.html>`_. Parameters ---------- freq Frequency of the data to train on and predict prediction_length Length of the prediction horizon trainer Trainer object to be used (default: Trainer()) context_length Number of steps to unroll the RNN for before computing predictions (default: None, in which case context_length = prediction_length) num_layers Number of RNN layers (default: 2) num_cells Number of RNN cells for each layer (default: 40) cell_type Type of recurrent cells to use (available: 'lstm' or 'gru'; default: 'lstm') dropout_rate Dropout regularization parameter (default: 0.1) use_feat_dynamic_real Whether to use the ``feat_dynamic_real`` field from the data (default: False) use_feat_static_cat Whether to use the ``feat_static_cat`` field from the data (default: False) use_feat_static_real Whether to use the ``feat_static_real`` field from the data (default: False) cardinality Number of values of each categorical feature. This must be set if ``use_feat_static_cat == True`` (default: None) embedding_dimension Dimension of the embeddings for categorical features (default: [min(50, (cat+1)//2) for cat in cardinality]) distr_output Distribution to use to evaluate observations and sample predictions (default: StudentTOutput()) scaling Whether to automatically scale the target values (default: true) lags_seq Indices of the lagged target values to use as inputs of the RNN (default: None, in which case these are automatically determined based on freq) time_features Time features to use as inputs of the RNN (default: None, in which case these are automatically determined based on freq) num_parallel_samples Number of evaluation samples per time series to increase parallelism during inference. This is a model optimization that does not affect the accuracy (default: 100) """ @validated() def __init__( self, freq: str, prediction_length: int, trainer: Trainer = Trainer(), context_length: Optional[int] = None, num_layers: int = 2, num_cells: int = 40, cell_type: str = "lstm", dropout_rate: float = 0.1, use_feat_dynamic_real: bool = False, use_feat_static_cat: bool = False, use_feat_static_real: bool = False, cardinality: Optional[List[int]] = None, embedding_dimension: Optional[List[int]] = None, distr_output: DistributionOutput = StudentTOutput(), scaling: bool = True, lags_seq: Optional[List[int]] = None, time_features: Optional[List[TimeFeature]] = None, num_parallel_samples: int = 100, dtype: DType = np.float32, ) -> None: super().__init__(trainer=trainer, dtype=dtype) assert ( prediction_length > 0 ), "The value of `prediction_length` should be > 0" assert ( context_length is None or context_length > 0 ), "The value of `context_length` should be > 0" assert num_layers > 0, "The value of `num_layers` should be > 0" assert num_cells > 0, "The value of `num_cells` should be > 0" assert dropout_rate >= 0, "The value of `dropout_rate` should be >= 0" assert (cardinality is not None and use_feat_static_cat) or ( cardinality is None and not use_feat_static_cat ), "You should set `cardinality` if and only if `use_feat_static_cat=True`" assert cardinality is None or all( [c > 0 for c in cardinality] ), "Elements of `cardinality` should be > 0" assert embedding_dimension is None or all( [e > 0 for e in embedding_dimension] ), "Elements of `embedding_dimension` should be > 0" assert ( num_parallel_samples > 0 ), "The value of `num_parallel_samples` should be > 0" self.freq = freq self.context_length = ( context_length if context_length is not None else prediction_length ) self.prediction_length = prediction_length self.distr_output = distr_output self.distr_output.dtype = dtype self.num_layers = num_layers self.num_cells = num_cells self.cell_type = cell_type self.dropout_rate = dropout_rate self.use_feat_dynamic_real = use_feat_dynamic_real self.use_feat_static_cat = use_feat_static_cat self.use_feat_static_real = use_feat_static_real self.cardinality = ( cardinality if cardinality and use_feat_static_cat else [1] ) self.embedding_dimension = ( embedding_dimension if embedding_dimension is not None else [min(50, (cat + 1) // 2) for cat in self.cardinality] ) self.scaling = scaling self.lags_seq = ( lags_seq if lags_seq is not None else get_lags_for_frequency(freq_str=freq) ) self.time_features = ( time_features if time_features is not None else time_features_from_frequency_str(self.freq) ) self.history_length = self.context_length + max(self.lags_seq) self.num_parallel_samples = num_parallel_samples def create_transformation(self) -> Transformation: remove_field_names = [FieldName.FEAT_DYNAMIC_CAT] if not self.use_feat_static_real: remove_field_names.append(FieldName.FEAT_STATIC_REAL) if not self.use_feat_dynamic_real: remove_field_names.append(FieldName.FEAT_DYNAMIC_REAL) return Chain( [RemoveFields(field_names=remove_field_names)] + ( [SetField(output_field=FieldName.FEAT_STATIC_CAT, value=[0.0])] if not self.use_feat_static_cat else [] ) + ( [ SetField( output_field=FieldName.FEAT_STATIC_REAL, value=[0.0] ) ] if not self.use_feat_static_real else [] ) + [ AsNumpyArray( field=FieldName.FEAT_STATIC_CAT, expected_ndim=1, dtype=self.dtype, ), AsNumpyArray( field=FieldName.FEAT_STATIC_REAL, expected_ndim=1, dtype=self.dtype, ), AsNumpyArray( field=FieldName.TARGET, # in the following line, we add 1 for the time dimension expected_ndim=1 + len(self.distr_output.event_shape), dtype=self.dtype, ), AddObservedValuesIndicator( target_field=FieldName.TARGET, output_field=FieldName.OBSERVED_VALUES, dtype=self.dtype, ), AddTimeFeatures( start_field=FieldName.START, target_field=FieldName.TARGET, output_field=FieldName.FEAT_TIME, time_features=self.time_features, pred_length=self.prediction_length, ), AddAgeFeature( target_field=FieldName.TARGET, output_field=FieldName.FEAT_AGE, pred_length=self.prediction_length, log_scale=True, dtype=self.dtype, ), VstackFeatures( output_field=FieldName.FEAT_TIME, input_fields=[FieldName.FEAT_TIME, FieldName.FEAT_AGE] + ( [FieldName.FEAT_DYNAMIC_REAL] if self.use_feat_dynamic_real else [] ), ), InstanceSplitter( target_field=FieldName.TARGET, is_pad_field=FieldName.IS_PAD, start_field=FieldName.START, forecast_start_field=FieldName.FORECAST_START, train_sampler=ExpectedNumInstanceSampler(num_instances=1), past_length=self.history_length, future_length=self.prediction_length, time_series_fields=[ FieldName.FEAT_TIME, FieldName.OBSERVED_VALUES, ], ), ] ) def create_training_network(self) -> DeepARTrainingNetwork: return DeepARTrainingNetwork( num_layers=self.num_layers, num_cells=self.num_cells, cell_type=self.cell_type, history_length=self.history_length, context_length=self.context_length, prediction_length=self.prediction_length, distr_output=self.distr_output, dropout_rate=self.dropout_rate, cardinality=self.cardinality, embedding_dimension=self.embedding_dimension, lags_seq=self.lags_seq, scaling=self.scaling, dtype=self.dtype, ) def create_predictor( self, transformation: Transformation, trained_network: HybridBlock ) -> Predictor: prediction_network = DeepARPredictionNetwork( num_parallel_samples=self.num_parallel_samples, num_layers=self.num_layers, num_cells=self.num_cells, cell_type=self.cell_type, history_length=self.history_length, context_length=self.context_length, prediction_length=self.prediction_length, distr_output=self.distr_output, dropout_rate=self.dropout_rate, cardinality=self.cardinality, embedding_dimension=self.embedding_dimension, lags_seq=self.lags_seq, scaling=self.scaling, dtype=self.dtype, ) copy_parameters(trained_network, prediction_network) return RepresentableBlockPredictor( input_transform=transformation, prediction_net=prediction_network, batch_size=self.trainer.batch_size, freq=self.freq, prediction_length=self.prediction_length, ctx=self.trainer.ctx, dtype=self.dtype, )
12,645
37.090361
94
py
rankpredictor
rankpredictor-master/sub/gluonts/model/deepar-savedata/predictor.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports import functools import itertools import logging import multiprocessing as mp import sys import traceback from pathlib import Path from pydoc import locate from tempfile import TemporaryDirectory import json from typing import ( TYPE_CHECKING, Tuple, Union, Any, Callable, Dict, Iterator, List, Optional, Type, ) # Third-party imports import mxnet as mx import numpy as np # First-party imports import gluonts from gluonts.distribution import Distribution, DistributionOutput from gluonts.core.component import ( DType, equals, from_hyperparameters, get_mxnet_context, validated, ) from gluonts.core.exception import GluonTSException from gluonts.core.serde import dump_json, fqname_for, load_json from gluonts.dataset.common import DataEntry, Dataset, ListDataset from .forecast_generator import ForecastGenerator, SampleForecastGenerator from gluonts.dataset.loader import DataBatch, InferenceDataLoader from gluonts.model.forecast import Forecast from gluonts.support.util import ( export_repr_block, export_symb_block, get_hybrid_forward_input_names, hybrid_block_to_symbol_block, import_repr_block, import_symb_block, ) from gluonts.transform import Transformation if TYPE_CHECKING: # avoid circular import from gluonts.model.estimator import Estimator # noqa OutputTransform = Callable[[DataEntry, np.ndarray], np.ndarray] class Predictor: """ Abstract class representing predictor objects. Parameters ---------- prediction_length Prediction horizon. freq Frequency of the predicted data. """ __version__: str = gluonts.__version__ def __init__(self, prediction_length: int, freq: str) -> None: assert ( prediction_length > 0 ), "The value of `prediction_length` should be > 0" self.prediction_length = prediction_length self.freq = freq def predict(self, dataset: Dataset, **kwargs) -> Iterator[Forecast]: """ Compute forecasts for the time series in the provided dataset. This method is not implemented in this abstract class; please use one of the subclasses. Parameters ---------- dataset The dataset containing the time series to predict. Returns ------- Iterator[Forecast] Iterator over the forecasts, in the same order as the dataset iterable was provided. """ raise NotImplementedError def serialize(self, path: Path) -> None: # serialize Predictor type with (path / "type.txt").open("w") as fp: fp.write(fqname_for(self.__class__)) with (path / "version.json").open("w") as fp: json.dump( {"model": self.__version__, "gluonts": gluonts.__version__}, fp ) @classmethod def deserialize( cls, path: Path, ctx: Optional[mx.Context] = None ) -> "Predictor": """ Load a serialized predictor from the given path Parameters ---------- path Path to the serialized files predictor. ctx Optional mxnet context to be used with the predictor. If nothing is passed will use the GPU if available and CPU otherwise. """ # deserialize Predictor type with (path / "type.txt").open("r") as fp: tpe = locate(fp.readline()) # ensure that predictor_cls is a subtype of Predictor if not issubclass(tpe, Predictor): raise IOError( f"Class {fqname_for(tpe)} is not " f"a subclass of {fqname_for(Predictor)}" ) # call deserialize() for the concrete Predictor type return tpe.deserialize(path, ctx) @classmethod def from_hyperparameters(cls, **hyperparameters): return from_hyperparameters(cls, **hyperparameters) class RepresentablePredictor(Predictor): """ An abstract predictor that can be subclassed by models that are not based on Gluon. Subclasses should have @validated() constructors. (De)serialization and value equality are all implemented on top of the @validated() logic. Parameters ---------- prediction_length Prediction horizon. freq Frequency of the predicted data. """ @validated() def __init__(self, prediction_length: int, freq: str) -> None: super().__init__(prediction_length, freq) def predict(self, dataset: Dataset, **kwargs) -> Iterator[Forecast]: for item in dataset: yield self.predict_item(item) def predict_item(self, item: DataEntry) -> Forecast: raise NotImplementedError def __eq__(self, that): """ Two RepresentablePredictor instances are considered equal if they have the same constructor arguments. """ return equals(self, that) def serialize(self, path: Path) -> None: # call Predictor.serialize() in order to serialize the class name super().serialize(path) with (path / "predictor.json").open("w") as fp: print(dump_json(self), file=fp) @classmethod def deserialize( cls, path: Path, ctx: Optional[mx.Context] = None ) -> "RepresentablePredictor": with (path / "predictor.json").open("r") as fp: return load_json(fp.read()) class GluonPredictor(Predictor): """ Base predictor type for Gluon-based models. Parameters ---------- input_names Input tensor names for the graph prediction_net Network that will be called for prediction batch_size Number of time series to predict in a single batch prediction_length Number of time steps to predict freq Frequency of the input data input_transform Input transformation pipeline output_transform Output transformation ctx MXNet context to use for computation forecast_generator Class to generate forecasts from network ouputs """ BlockType = mx.gluon.Block def __init__( self, input_names: List[str], prediction_net: BlockType, batch_size: int, prediction_length: int, freq: str, ctx: mx.Context, input_transform: Transformation, forecast_generator: ForecastGenerator = SampleForecastGenerator(), output_transform: Optional[OutputTransform] = None, dtype: DType = np.float32, ) -> None: super().__init__(prediction_length, freq) self.input_names = input_names self.prediction_net = prediction_net #self.batch_size = batch_size self.batch_size = 1 self.input_transform = input_transform self.forecast_generator = forecast_generator self.output_transform = output_transform self.ctx = ctx self.dtype = dtype def hybridize(self, batch: DataBatch) -> None: """ Hybridizes the underlying prediction network. Parameters ---------- batch A batch of data to use for the required forward pass after the `hybridize()` call. """ self.prediction_net.hybridize(active=True) self.prediction_net(*[batch[k] for k in self.input_names]) def as_symbol_block_predictor( self, batch: DataBatch ) -> "SymbolBlockPredictor": """ Returns a variant of the current :class:`GluonPredictor` backed by a Gluon `SymbolBlock`. If the current predictor is already a :class:`SymbolBlockPredictor`, it just returns itself. Parameters ---------- batch A batch of data to use for the required forward pass after the `hybridize()` call of the underlying network. Returns ------- SymbolBlockPredictor A predictor derived from the current one backed by a `SymbolBlock`. """ raise NotImplementedError def predict( self, dataset: Dataset, num_samples: Optional[int] = None ) -> Iterator[Forecast]: #print('predict') inference_data_loader = InferenceDataLoader( dataset, self.input_transform, #self.batch_size, 1, ctx=self.ctx, dtype=self.dtype, ) yield from self.forecast_generator( inference_data_loader=inference_data_loader, prediction_net=self.prediction_net, input_names=self.input_names, freq=self.freq, output_transform=self.output_transform, num_samples=num_samples, ) def __eq__(self, that): if type(self) != type(that): return False # TODO: also consider equality of the pipelines # if not equals(self.input_transform, that.input_transform): # return False return equals( self.prediction_net.collect_params(), that.prediction_net.collect_params(), ) def serialize(self, path: Path) -> None: # call Predictor.serialize() in order to serialize the class name super().serialize(path) # serialize every GluonPredictor-specific parameters # serialize the prediction network self.serialize_prediction_net(path) # serialize transformation chain with (path / "input_transform.json").open("w") as fp: print(dump_json(self.input_transform), file=fp) # FIXME: also needs to serialize the output_transform # serialize all remaining constructor parameters with (path / "parameters.json").open("w") as fp: parameters = dict( batch_size=self.batch_size, prediction_length=self.prediction_length, freq=self.freq, ctx=self.ctx, dtype=self.dtype, forecast_generator=self.forecast_generator, input_names=self.input_names, ) print(dump_json(parameters), file=fp) def serialize_prediction_net(self, path: Path) -> None: raise NotImplementedError() class SymbolBlockPredictor(GluonPredictor): """ A predictor which serializes the network structure as an MXNet symbolic graph. Should be used for models deployed in production in order to ensure forward-compatibility as GluonTS models evolve. Used by the training shell if training is invoked with a hyperparameter `use_symbol_block_predictor = True`. """ BlockType = mx.gluon.SymbolBlock def as_symbol_block_predictor( self, batch: DataBatch ) -> "SymbolBlockPredictor": return self def serialize_prediction_net(self, path: Path) -> None: export_symb_block(self.prediction_net, path, "prediction_net") @classmethod def deserialize( cls, path: Path, ctx: Optional[mx.Context] = None ) -> "SymbolBlockPredictor": ctx = ctx if ctx is not None else get_mxnet_context() with mx.Context(ctx): # deserialize constructor parameters with (path / "parameters.json").open("r") as fp: parameters = load_json(fp.read()) parameters["ctx"] = ctx # deserialize transformation chain with (path / "input_transform.json").open("r") as fp: transform = load_json(fp.read()) # deserialize prediction network num_inputs = len(parameters["input_names"]) prediction_net = import_symb_block( num_inputs, path, "prediction_net" ) return SymbolBlockPredictor( input_transform=transform, prediction_net=prediction_net, **parameters, ) class RepresentableBlockPredictor(GluonPredictor): """ A predictor which serializes the network structure using the JSON-serialization methods located in `gluonts.core.serde`. Use the following logic to create a `RepresentableBlockPredictor` from a trained prediction network. >>> def create_representable_block_predictor( ... prediction_network: mx.gluon.HybridBlock, ... **kwargs ... ) -> RepresentableBlockPredictor: ... return RepresentableBlockPredictor( ... prediction_net=prediction_network, ... **kwargs ... ) """ BlockType = mx.gluon.HybridBlock def __init__( self, prediction_net: BlockType, batch_size: int, prediction_length: int, freq: str, ctx: mx.Context, input_transform: Transformation, forecast_generator: ForecastGenerator = SampleForecastGenerator(), output_transform: Optional[ Callable[[DataEntry, np.ndarray], np.ndarray] ] = None, dtype: DType = np.float32, ) -> None: super().__init__( input_names=get_hybrid_forward_input_names(prediction_net), prediction_net=prediction_net, batch_size=batch_size, prediction_length=prediction_length, freq=freq, ctx=ctx, input_transform=input_transform, forecast_generator=forecast_generator, output_transform=output_transform, dtype=dtype, ) def as_symbol_block_predictor( self, batch: DataBatch ) -> SymbolBlockPredictor: symbol_block_net = hybrid_block_to_symbol_block( hb=self.prediction_net, data_batch=[batch[k] for k in self.input_names], ) return SymbolBlockPredictor( input_names=self.input_names, prediction_net=symbol_block_net, batch_size=self.batch_size, prediction_length=self.prediction_length, freq=self.freq, ctx=self.ctx, input_transform=self.input_transform, forecast_generator=self.forecast_generator, output_transform=self.output_transform, dtype=self.dtype, ) def serialize(self, path: Path) -> None: logging.warning( "Serializing RepresentableBlockPredictor instances does not save " "the prediction network structure in a backwards-compatible " "manner. Be careful not to use this method in production." ) super().serialize(path) def serialize_prediction_net(self, path: Path) -> None: export_repr_block(self.prediction_net, path, "prediction_net") @classmethod def deserialize( cls, path: Path, ctx: Optional[mx.Context] = None ) -> "RepresentableBlockPredictor": ctx = ctx if ctx is not None else get_mxnet_context() with mx.Context(ctx): # deserialize constructor parameters with (path / "parameters.json").open("r") as fp: parameters = load_json(fp.read()) # deserialize transformation chain with (path / "input_transform.json").open("r") as fp: transform = load_json(fp.read()) # deserialize prediction network prediction_net = import_repr_block(path, "prediction_net") # input_names is derived from the prediction_net if "input_names" in parameters: del parameters["input_names"] parameters["ctx"] = ctx return RepresentableBlockPredictor( input_transform=transform, prediction_net=prediction_net, **parameters, ) class WorkerError: def __init__(self, msg): self.msg = msg def _worker_loop( predictor_path: Path, input_queue: mp.Queue, output_queue: mp.Queue, worker_id, **kwargs, ): """ Worker loop for multiprocessing Predictor. Loads the predictor serialized in predictor_path reads inputs from input_queue and writes forecasts to output_queue """ predictor = Predictor.deserialize(predictor_path) while True: idx, data_chunk = input_queue.get() if idx is None: output_queue.put((None, None, None)) break try: result = list(predictor.predict(data_chunk, **kwargs)) except Exception: we = WorkerError( "".join(traceback.format_exception(*sys.exc_info())) ) output_queue.put((we, None, None)) break output_queue.put((idx, worker_id, result)) class ParallelizedPredictor(Predictor): """ Runs multiple instances (workers) of a predictor in parallel. Exceptions are propagated from the workers. Note: That there is currently an issue with tqdm that will cause things to hang if the ParallelizedPredictor is used with tqdm and an exception occurs during prediction. https://github.com/tqdm/tqdm/issues/548 Parameters ---------- base_predictor A representable predictor that will be used num_workers Number of workers (processes) to use. If set to None, one worker per CPU will be used. chunk_size Number of items to pass per call """ def __init__( self, base_predictor: Predictor, num_workers: Optional[int] = None, chunk_size=1, ) -> None: super().__init__(base_predictor.prediction_length, base_predictor.freq) self._base_predictor = base_predictor self._num_workers = ( num_workers if num_workers is not None else mp.cpu_count() ) self._chunk_size = chunk_size self._num_running_workers = 0 self._input_queues = [] self._output_queue = None def _grouper(self, iterable, n): iterator = iter(iterable) group = tuple(itertools.islice(iterator, n)) while group: yield group group = tuple(itertools.islice(iterator, n)) def terminate(self): for q in self._input_queues: q.put((None, None)) for w in self._workers: w.terminate() for i, w in enumerate(self._workers): w.join() def predict(self, dataset: Dataset, **kwargs) -> Iterator[Forecast]: with TemporaryDirectory() as tempdir: predictor_path = Path(tempdir) self._base_predictor.serialize(predictor_path) # TODO: Consider using shared memory for the data transfer. self._input_queues = [mp.Queue() for _ in range(self._num_workers)] self._output_queue = mp.Queue() workers = [] for worker_id, in_q in enumerate(self._input_queues): worker = mp.Process( target=_worker_loop, args=(predictor_path, in_q, self._output_queue, worker_id), kwargs=kwargs, ) worker.daemon = True worker.start() workers.append(worker) self._num_running_workers += 1 self._workers = workers chunked_data = self._grouper(dataset, self._chunk_size) self._send_idx = 0 self._next_idx = 0 self._data_buffer = {} worker_ids = list(range(self._num_workers)) def receive(): idx, worker_id, result = self._output_queue.get() if isinstance(idx, WorkerError): self._num_running_workers -= 1 self.terminate() raise Exception(idx.msg) if idx is not None: self._data_buffer[idx] = result return idx, worker_id, result def get_next_from_buffer(): while self._next_idx in self._data_buffer: result_batch = self._data_buffer.pop(self._next_idx) self._next_idx += 1 for result in result_batch: yield result def send(worker_id, chunk): q = self._input_queues[worker_id] q.put((self._send_idx, chunk)) self._send_idx += 1 try: # prime the queues for wid in worker_ids: chunk = next(chunked_data) send(wid, chunk) while True: idx, wid, result = receive() for res in get_next_from_buffer(): yield res chunk = next(chunked_data) send(wid, chunk) except StopIteration: # signal workers end of data for q in self._input_queues: q.put((None, None)) # collect any outstanding results while self._num_running_workers > 0: idx, worker_id, result = receive() if idx is None: self._num_running_workers -= 1 continue for res in get_next_from_buffer(): yield res assert len(self._data_buffer) == 0 assert self._send_idx == self._next_idx class Localizer(Predictor): """ A Predictor that uses an estimator to train a local model per time series and immediatly calls this to predict. Parameters ---------- estimator The estimator object to train on each dataset entry at prediction time. """ def __init__(self, estimator: "Estimator"): super().__init__(estimator.prediction_length, estimator.freq) self.estimator = estimator def predict(self, dataset: Dataset, **kwargs) -> Iterator[Forecast]: logger = logging.getLogger(__name__) for i, ts in enumerate(dataset, start=1): logger.info(f"training for time series {i} / {len(dataset)}") local_ds = ListDataset([ts], freq=self.freq) trained_pred = self.estimator.train(local_ds) logger.info(f"predicting for time series {i} / {len(dataset)}") predictions = trained_pred.predict(local_ds, **kwargs) for pred in predictions: yield pred class FallbackPredictor(Predictor): @classmethod def from_predictor( cls, base: RepresentablePredictor, **overrides ) -> Predictor: # Create predictor based on an existing predictor. # This let's us create a MeanPredictor as a fallback on the fly. return cls.from_hyperparameters( **getattr(base, "__init_args__"), **overrides ) def fallback(fallback_cls: Type[FallbackPredictor]): def decorator(predict_item): @functools.wraps(predict_item) def fallback_predict(self, item: DataEntry) -> Forecast: try: return predict_item(self, item) except GluonTSException: raise except Exception: logging.warning( f"Base predictor failed with: {traceback.format_exc()}" ) fallback_predictor = fallback_cls.from_predictor(self) return fallback_predictor.predict_item(item) return fallback_predict return decorator
24,040
30.969415
81
py
rankpredictor
rankpredictor-master/sub/gluonts/model/deepar-savedata/_network.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports import numpy as np from typing import List, Optional, Tuple # Third-party imports import mxnet as mx # First-party imports from gluonts.block.feature import FeatureEmbedder from gluonts.block.scaler import MeanScaler, NOPScaler from gluonts.core.component import DType, validated from gluonts.distribution import DistributionOutput, Distribution from gluonts.distribution.distribution import getF from gluonts.model.common import Tensor from gluonts.support.util import weighted_average def prod(xs): p = 1 for x in xs: p *= x return p class DeepARNetwork(mx.gluon.HybridBlock): @validated() def __init__( self, num_layers: int, num_cells: int, cell_type: str, history_length: int, context_length: int, prediction_length: int, distr_output: DistributionOutput, dropout_rate: float, cardinality: List[int], embedding_dimension: List[int], lags_seq: List[int], scaling: bool = True, dtype: DType = np.float32, **kwargs, ) -> None: super().__init__(**kwargs) self.num_layers = num_layers self.num_cells = num_cells self.cell_type = cell_type self.history_length = history_length self.context_length = context_length self.prediction_length = prediction_length self.dropout_rate = dropout_rate self.cardinality = cardinality self.embedding_dimension = embedding_dimension self.num_cat = len(cardinality) self.scaling = scaling self.dtype = dtype assert len(cardinality) == len( embedding_dimension ), "embedding_dimension should be a list with the same size as cardinality" assert len(set(lags_seq)) == len( lags_seq ), "no duplicated lags allowed!" lags_seq.sort() self.lags_seq = lags_seq self.distr_output = distr_output RnnCell = {"lstm": mx.gluon.rnn.LSTMCell, "gru": mx.gluon.rnn.GRUCell}[ self.cell_type ] self.target_shape = distr_output.event_shape # TODO: is the following restriction needed? assert ( len(self.target_shape) <= 1 ), "Argument `target_shape` should be a tuple with 1 element at most" with self.name_scope(): self.proj_distr_args = distr_output.get_args_proj() self.rnn = mx.gluon.rnn.HybridSequentialRNNCell() for k in range(num_layers): cell = RnnCell(hidden_size=num_cells) cell = mx.gluon.rnn.ResidualCell(cell) if k > 0 else cell cell = ( mx.gluon.rnn.ZoneoutCell(cell, zoneout_states=dropout_rate) if dropout_rate > 0.0 else cell ) self.rnn.add(cell) self.rnn.cast(dtype=dtype) self.embedder = FeatureEmbedder( cardinalities=cardinality, embedding_dims=embedding_dimension, dtype=self.dtype, ) if scaling: self.scaler = MeanScaler(keepdims=True) else: self.scaler = NOPScaler(keepdims=True) #save data self.reset_savedata() def reset_savedata(self): self.savedata = {} self.savedata['input'] = [] self.savedata['target'] = [] self.savedata['lags'] = [] self.savedata['theta'] = [] self.savedata['hstate'] = [] self.savedata['rnnoutput'] = [] @staticmethod def get_lagged_subsequences( F, sequence: Tensor, sequence_length: int, indices: List[int], subsequences_length: int = 1, ) -> Tensor: """ Returns lagged subsequences of a given sequence. Parameters ---------- sequence : Tensor the sequence from which lagged subsequences should be extracted. Shape: (N, T, C). sequence_length : int length of sequence in the T (time) dimension (axis = 1). indices : List[int] list of lag indices to be used. subsequences_length : int length of the subsequences to be extracted. Returns -------- lagged : Tensor a tensor of shape (N, S, C, I), where S = subsequences_length and I = len(indices), containing lagged subsequences. Specifically, lagged[i, j, :, k] = sequence[i, -indices[k]-S+j, :]. """ # we must have: sequence_length - lag_index - subsequences_length >= 0 # for all lag_index, hence the following assert assert max(indices) + subsequences_length <= sequence_length, ( f"lags cannot go further than history length, " f"found lag {max(indices)} while history length is only " f"{sequence_length}" ) assert all(lag_index >= 0 for lag_index in indices) lagged_values = [] for lag_index in indices: begin_index = -lag_index - subsequences_length end_index = -lag_index if lag_index > 0 else None lagged_values.append( F.slice_axis( sequence, axis=1, begin=begin_index, end=end_index ) ) return F.stack(*lagged_values, axis=-1) def unroll_encoder( self, F, feat_static_cat: Tensor, # (batch_size, num_features) feat_static_real: Tensor, # (batch_size, num_features) past_time_feat: Tensor, # (batch_size, history_length, num_features) past_target: Tensor, # (batch_size, history_length, *target_shape) past_observed_values: Tensor, # (batch_size, history_length, *target_shape) future_time_feat: Optional[ Tensor ], # (batch_size, prediction_length, num_features) future_target: Optional[ Tensor ], # (batch_size, prediction_length, *target_shape) ) -> Tuple[Tensor, List, Tensor, Tensor]: """ Unrolls the LSTM encoder over past and, if present, future data. Returns outputs and state of the encoder, plus the scale of past_target and a vector of static features that was constructed and fed as input to the encoder. All tensor arguments should have NTC layout. """ if future_time_feat is None or future_target is None: time_feat = past_time_feat.slice_axis( axis=1, begin=self.history_length - self.context_length, end=None, ) sequence = past_target sequence_length = self.history_length subsequences_length = self.context_length else: time_feat = F.concat( past_time_feat.slice_axis( axis=1, begin=self.history_length - self.context_length, end=None, ), future_time_feat, dim=1, ) sequence = F.concat(past_target, future_target, dim=1) sequence_length = self.history_length + self.prediction_length subsequences_length = self.context_length + self.prediction_length # (batch_size, sub_seq_len, *target_shape, num_lags) lags = self.get_lagged_subsequences( F=F, sequence=sequence, sequence_length=sequence_length, indices=self.lags_seq, subsequences_length=subsequences_length, ) # scale is computed on the context length last units of the past target # scale shape is (batch_size, 1, *target_shape) _, scale = self.scaler( past_target.slice_axis( axis=1, begin=-self.context_length, end=None ), past_observed_values.slice_axis( axis=1, begin=-self.context_length, end=None ), ) # (batch_size, num_features) embedded_cat = self.embedder(feat_static_cat) # in addition to embedding features, use the log scale as it can help # prediction too # (batch_size, num_features + prod(target_shape)) static_feat = F.concat( embedded_cat, feat_static_real, F.log(scale) if len(self.target_shape) == 0 else F.log(scale.squeeze(axis=1)), dim=1, ) # (batch_size, subsequences_length, num_features + 1) repeated_static_feat = static_feat.expand_dims(axis=1).repeat( axis=1, repeats=subsequences_length ) # (batch_size, sub_seq_len, *target_shape, num_lags) lags_scaled = F.broadcast_div(lags, scale.expand_dims(axis=-1)) # from (batch_size, sub_seq_len, *target_shape, num_lags) # to (batch_size, sub_seq_len, prod(target_shape) * num_lags) input_lags = F.reshape( data=lags_scaled, shape=( -1, subsequences_length, len(self.lags_seq) * prod(self.target_shape), ), ) # (batch_size, sub_seq_len, input_dim) inputs = F.concat(input_lags, time_feat, repeated_static_feat, dim=-1) #save data here self.savedata['input'].append(inputs.asnumpy().copy()) #self.savedata.append(inputs) self.savedata['lags'].append(lags.asnumpy().copy()) #print(self.lags_seq) # unroll encoder outputs, state = self.rnn.unroll( inputs=inputs, length=subsequences_length, layout="NTC", merge_outputs=True, begin_state=self.rnn.begin_state( func=F.zeros, dtype=self.dtype, batch_size=inputs.shape[0] if isinstance(inputs, mx.nd.NDArray) else 0, ), ) # outputs: (batch_size, seq_len, num_cells) # state: list of (batch_size, num_cells) tensors # scale: (batch_size, 1, *target_shape) # static_feat: (batch_size, num_features + prod(target_shape)) return outputs, state, scale, static_feat class DeepARTrainingNetwork(DeepARNetwork): def distribution( self, feat_static_cat: Tensor, feat_static_real: Tensor, past_time_feat: Tensor, past_target: Tensor, past_observed_values: Tensor, future_time_feat: Tensor, future_target: Tensor, future_observed_values: Tensor, ) -> Distribution: """ Returns the distribution predicted by the model on the range of past_target and future_target. The distribution is obtained by unrolling the network with the true target, this is also the distribution that is being minimized during training. This can be used in anomaly detection, see for instance examples/anomaly_detection.py. Input arguments are the same as for the hybrid_forward method. Returns ------- Distribution a distribution object whose mean has shape: (batch_size, context_length + prediction_length). """ # unroll the decoder in "training mode" # i.e. by providing future data as well F = getF(feat_static_cat) rnn_outputs, _, scale, _ = self.unroll_encoder( F=F, feat_static_cat=feat_static_cat, feat_static_real=feat_static_real, past_time_feat=past_time_feat, past_target=past_target, past_observed_values=past_observed_values, future_time_feat=future_time_feat, future_target=future_target, ) distr_args = self.proj_distr_args(rnn_outputs) return self.distr_output.distribution(distr_args, scale=scale) # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, feat_static_cat: Tensor, feat_static_real: Tensor, past_time_feat: Tensor, past_target: Tensor, past_observed_values: Tensor, future_time_feat: Tensor, future_target: Tensor, future_observed_values: Tensor, ) -> Tensor: """ Computes the loss for training DeepAR, all inputs tensors representing time series have NTC layout. Parameters ---------- F feat_static_cat : (batch_size, num_features) feat_static_real : (batch_size, num_features) past_time_feat : (batch_size, history_length, num_features) past_target : (batch_size, history_length, *target_shape) past_observed_values : (batch_size, history_length, *target_shape, seq_len) future_time_feat : (batch_size, prediction_length, num_features) future_target : (batch_size, prediction_length, *target_shape) future_observed_values : (batch_size, prediction_length, *target_shape) Returns loss with shape (batch_size, context + prediction_length, 1) ------- """ distr = self.distribution( feat_static_cat=feat_static_cat, feat_static_real=feat_static_real, past_time_feat=past_time_feat, past_target=past_target, past_observed_values=past_observed_values, future_time_feat=future_time_feat, future_target=future_target, future_observed_values=future_observed_values, ) # put together target sequence # (batch_size, seq_len, *target_shape) target = F.concat( past_target.slice_axis( axis=1, begin=self.history_length - self.context_length, end=None, ), future_target, dim=1, ) # (batch_size, seq_len) loss = distr.loss(target) #save target in training self.savedata['target'].append(target.asnumpy().copy()) # (batch_size, seq_len, *target_shape) observed_values = F.concat( past_observed_values.slice_axis( axis=1, begin=self.history_length - self.context_length, end=self.history_length, ), future_observed_values, dim=1, ) # mask the loss at one time step iff one or more observations is missing in the target dimensions # (batch_size, seq_len) loss_weights = ( observed_values if (len(self.target_shape) == 0) else observed_values.min(axis=-1, keepdims=False) ) weighted_loss = weighted_average( F=F, x=loss, weights=loss_weights, axis=1 ) return weighted_loss, loss class DeepARPredictionNetwork(DeepARNetwork): @validated() #def __init__(self, num_parallel_samples: int = 100, **kwargs) -> None: def __init__(self, num_parallel_samples: int = 1, **kwargs) -> None: super().__init__(**kwargs) #self.num_parallel_samples = num_parallel_samples self.num_parallel_samples = 1 # for decoding the lags are shifted by one, at the first time-step # of the decoder a lag of one corresponds to the last target value self.shifted_lags = [l - 1 for l in self.lags_seq] def sampling_decoder( self, F, static_feat: Tensor, past_target: Tensor, time_feat: Tensor, scale: Tensor, begin_states: List, ) -> Tensor: """ Computes sample paths by unrolling the LSTM starting with a initial input and state. Parameters ---------- static_feat : Tensor static features. Shape: (batch_size, num_static_features). past_target : Tensor target history. Shape: (batch_size, history_length). time_feat : Tensor time features. Shape: (batch_size, prediction_length, num_time_features). scale : Tensor tensor containing the scale of each element in the batch. Shape: (batch_size, 1, 1). begin_states : List list of initial states for the LSTM layers. the shape of each tensor of the list should be (batch_size, num_cells) Returns -------- Tensor A tensor containing sampled paths. Shape: (batch_size, num_sample_paths, prediction_length). """ # blows-up the dimension of each tensor to batch_size * self.num_parallel_samples for increasing parallelism repeated_past_target = past_target.repeat( repeats=self.num_parallel_samples, axis=0 ) repeated_time_feat = time_feat.repeat( repeats=self.num_parallel_samples, axis=0 ) repeated_static_feat = static_feat.repeat( repeats=self.num_parallel_samples, axis=0 ).expand_dims(axis=1) repeated_scale = scale.repeat( repeats=self.num_parallel_samples, axis=0 ) repeated_states = [ s.repeat(repeats=self.num_parallel_samples, axis=0) for s in begin_states ] future_samples = [] # for each future time-units we draw new samples for this time-unit and update the state for k in range(self.prediction_length): # (batch_size * num_samples, 1, *target_shape, num_lags) lags = self.get_lagged_subsequences( F=F, sequence=repeated_past_target, sequence_length=self.history_length + k, indices=self.shifted_lags, subsequences_length=1, ) # (batch_size * num_samples, 1, *target_shape, num_lags) lags_scaled = F.broadcast_div( lags, repeated_scale.expand_dims(axis=-1) ) # from (batch_size * num_samples, 1, *target_shape, num_lags) # to (batch_size * num_samples, 1, prod(target_shape) * num_lags) input_lags = F.reshape( data=lags_scaled, shape=(-1, 1, prod(self.target_shape) * len(self.lags_seq)), ) # (batch_size * num_samples, 1, prod(target_shape) * num_lags + num_time_features + num_static_features) decoder_input = F.concat( input_lags, repeated_time_feat.slice_axis(axis=1, begin=k, end=k + 1), repeated_static_feat, dim=-1, ) # output shape: (batch_size * num_samples, 1, num_cells) # state shape: (batch_size * num_samples, num_cells) rnn_outputs, repeated_states = self.rnn.unroll( inputs=decoder_input, length=1, begin_state=repeated_states, layout="NTC", merge_outputs=True, ) distr_args = self.proj_distr_args(rnn_outputs) # compute likelihood of target given the predicted parameters distr = self.distr_output.distribution( distr_args, scale=repeated_scale ) # (batch_size * num_samples, 1, *target_shape) new_samples = distr.sample(dtype=self.dtype) # (batch_size * num_samples, seq_len, *target_shape) repeated_past_target = F.concat( repeated_past_target, new_samples, dim=1 ) future_samples.append(new_samples) #save only the last output if k == self.prediction_length -1: self.savedata['hstate'].append(repeated_states) self.savedata['rnnoutput'].append(rnn_outputs.asnumpy().copy()) self.savedata['theta'].append(distr_args) self.savedata['target'].append(new_samples.asnumpy().copy()) # (batch_size * num_samples, prediction_length, *target_shape) samples = F.concat(*future_samples, dim=1) # (batch_size, num_samples, prediction_length, *target_shape) return samples.reshape( shape=( (-1, self.num_parallel_samples) + (self.prediction_length,) + self.target_shape ) ) # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, feat_static_cat: Tensor, # (batch_size, num_features) feat_static_real: Tensor, # (batch_size, num_features) past_time_feat: Tensor, # (batch_size, history_length, num_features) past_target: Tensor, # (batch_size, history_length, *target_shape) past_observed_values: Tensor, # (batch_size, history_length, *target_shape) future_time_feat: Tensor, # (batch_size, prediction_length, num_features) ) -> Tensor: """ Predicts samples, all tensors should have NTC layout. Parameters ---------- F feat_static_cat : (batch_size, num_features) feat_static_real : (batch_size, num_features) past_time_feat : (batch_size, history_length, num_features) past_target : (batch_size, history_length, *target_shape) past_observed_values : (batch_size, history_length, *target_shape) future_time_feat : (batch_size, prediction_length, num_features) Returns ------- Tensor Predicted samples """ # unroll the decoder in "prediction mode", i.e. with past data only _, state, scale, static_feat = self.unroll_encoder( F=F, feat_static_cat=feat_static_cat, feat_static_real=feat_static_real, past_time_feat=past_time_feat, past_target=past_target, past_observed_values=past_observed_values, future_time_feat=None, future_target=None, ) return self.sampling_decoder( F=F, past_target=past_target, time_feat=future_time_feat, static_feat=static_feat, scale=scale, begin_states=state, )
22,891
34.601866
116
py
rankpredictor
rankpredictor-master/sub/gluonts/model/deepar-savedata/_estimator.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports import numpy as np from typing import List, Optional # Third-party imports from mxnet.gluon import HybridBlock # First-party imports from gluonts.core.component import DType, validated from gluonts.dataset.field_names import FieldName from gluonts.distribution import DistributionOutput, StudentTOutput from gluonts.model.estimator import GluonEstimator from gluonts.model.predictor import Predictor, RepresentableBlockPredictor from gluonts.support.util import copy_parameters from gluonts.time_feature import ( TimeFeature, time_features_from_frequency_str, get_lags_for_frequency, ) from gluonts.trainer import Trainer from gluonts.transform import ( AddAgeFeature, AddObservedValuesIndicator, AddTimeFeatures, AsNumpyArray, Chain, ExpectedNumInstanceSampler, InstanceSplitter, RemoveFields, SetField, Transformation, VstackFeatures, ) # Relative imports from ._network import DeepARPredictionNetwork, DeepARTrainingNetwork class DeepAREstimator(GluonEstimator): """ Construct a DeepAR estimator. This implements an RNN-based model, close to the one described in [SFG17]_. *Note:* the code of this model is unrelated to the implementation behind `SageMaker's DeepAR Forecasting Algorithm <https://docs.aws.amazon.com/sagemaker/latest/dg/deepar.html>`_. Parameters ---------- freq Frequency of the data to train on and predict prediction_length Length of the prediction horizon trainer Trainer object to be used (default: Trainer()) context_length Number of steps to unroll the RNN for before computing predictions (default: None, in which case context_length = prediction_length) num_layers Number of RNN layers (default: 2) num_cells Number of RNN cells for each layer (default: 40) cell_type Type of recurrent cells to use (available: 'lstm' or 'gru'; default: 'lstm') dropout_rate Dropout regularization parameter (default: 0.1) use_feat_dynamic_real Whether to use the ``feat_dynamic_real`` field from the data (default: False) use_feat_static_cat Whether to use the ``feat_static_cat`` field from the data (default: False) use_feat_static_real Whether to use the ``feat_static_real`` field from the data (default: False) cardinality Number of values of each categorical feature. This must be set if ``use_feat_static_cat == True`` (default: None) embedding_dimension Dimension of the embeddings for categorical features (default: [min(50, (cat+1)//2) for cat in cardinality]) distr_output Distribution to use to evaluate observations and sample predictions (default: StudentTOutput()) scaling Whether to automatically scale the target values (default: true) lags_seq Indices of the lagged target values to use as inputs of the RNN (default: None, in which case these are automatically determined based on freq) time_features Time features to use as inputs of the RNN (default: None, in which case these are automatically determined based on freq) num_parallel_samples Number of evaluation samples per time series to increase parallelism during inference. This is a model optimization that does not affect the accuracy (default: 100) """ @validated() def __init__( self, freq: str, prediction_length: int, trainer: Trainer = Trainer(), context_length: Optional[int] = None, num_layers: int = 2, num_cells: int = 40, cell_type: str = "lstm", dropout_rate: float = 0.1, use_feat_dynamic_real: bool = False, use_feat_static_cat: bool = False, use_feat_static_real: bool = False, cardinality: Optional[List[int]] = None, embedding_dimension: Optional[List[int]] = None, distr_output: DistributionOutput = StudentTOutput(), scaling: bool = True, lags_seq: Optional[List[int]] = None, time_features: Optional[List[TimeFeature]] = None, num_parallel_samples: int = 100, dtype: DType = np.float32, ) -> None: super().__init__(trainer=trainer, dtype=dtype) assert ( prediction_length > 0 ), "The value of `prediction_length` should be > 0" assert ( context_length is None or context_length > 0 ), "The value of `context_length` should be > 0" assert num_layers > 0, "The value of `num_layers` should be > 0" assert num_cells > 0, "The value of `num_cells` should be > 0" assert dropout_rate >= 0, "The value of `dropout_rate` should be >= 0" assert (cardinality is not None and use_feat_static_cat) or ( cardinality is None and not use_feat_static_cat ), "You should set `cardinality` if and only if `use_feat_static_cat=True`" assert cardinality is None or all( [c > 0 for c in cardinality] ), "Elements of `cardinality` should be > 0" assert embedding_dimension is None or all( [e > 0 for e in embedding_dimension] ), "Elements of `embedding_dimension` should be > 0" assert ( num_parallel_samples > 0 ), "The value of `num_parallel_samples` should be > 0" self.freq = freq self.context_length = ( context_length if context_length is not None else prediction_length ) self.prediction_length = prediction_length self.distr_output = distr_output self.distr_output.dtype = dtype self.num_layers = num_layers self.num_cells = num_cells self.cell_type = cell_type self.dropout_rate = dropout_rate self.use_feat_dynamic_real = use_feat_dynamic_real self.use_feat_static_cat = use_feat_static_cat self.use_feat_static_real = use_feat_static_real self.cardinality = ( cardinality if cardinality and use_feat_static_cat else [1] ) self.embedding_dimension = ( embedding_dimension if embedding_dimension is not None else [min(50, (cat + 1) // 2) for cat in self.cardinality] ) self.scaling = scaling self.lags_seq = ( lags_seq if lags_seq is not None else get_lags_for_frequency(freq_str=freq) ) self.time_features = ( time_features if time_features is not None else time_features_from_frequency_str(self.freq) ) self.history_length = self.context_length + max(self.lags_seq) self.num_parallel_samples = num_parallel_samples def create_transformation(self) -> Transformation: remove_field_names = [FieldName.FEAT_DYNAMIC_CAT] if not self.use_feat_static_real: remove_field_names.append(FieldName.FEAT_STATIC_REAL) if not self.use_feat_dynamic_real: remove_field_names.append(FieldName.FEAT_DYNAMIC_REAL) return Chain( [RemoveFields(field_names=remove_field_names)] + ( [SetField(output_field=FieldName.FEAT_STATIC_CAT, value=[0.0])] if not self.use_feat_static_cat else [] ) + ( [ SetField( output_field=FieldName.FEAT_STATIC_REAL, value=[0.0] ) ] if not self.use_feat_static_real else [] ) + [ AsNumpyArray( field=FieldName.FEAT_STATIC_CAT, expected_ndim=1, dtype=self.dtype, ), AsNumpyArray( field=FieldName.FEAT_STATIC_REAL, expected_ndim=1, dtype=self.dtype, ), AsNumpyArray( field=FieldName.TARGET, # in the following line, we add 1 for the time dimension expected_ndim=1 + len(self.distr_output.event_shape), dtype=self.dtype, ), AddObservedValuesIndicator( target_field=FieldName.TARGET, output_field=FieldName.OBSERVED_VALUES, dtype=self.dtype, ), AddTimeFeatures( start_field=FieldName.START, target_field=FieldName.TARGET, output_field=FieldName.FEAT_TIME, time_features=self.time_features, pred_length=self.prediction_length, ), AddAgeFeature( target_field=FieldName.TARGET, output_field=FieldName.FEAT_AGE, pred_length=self.prediction_length, log_scale=True, dtype=self.dtype, ), VstackFeatures( output_field=FieldName.FEAT_TIME, input_fields=[FieldName.FEAT_TIME, FieldName.FEAT_AGE] + ( [FieldName.FEAT_DYNAMIC_REAL] if self.use_feat_dynamic_real else [] ), ), InstanceSplitter( target_field=FieldName.TARGET, is_pad_field=FieldName.IS_PAD, start_field=FieldName.START, forecast_start_field=FieldName.FORECAST_START, train_sampler=ExpectedNumInstanceSampler(num_instances=1), past_length=self.history_length, future_length=self.prediction_length, time_series_fields=[ FieldName.FEAT_TIME, FieldName.OBSERVED_VALUES, ], ), ] ) def create_training_network(self) -> DeepARTrainingNetwork: return DeepARTrainingNetwork( num_layers=self.num_layers, num_cells=self.num_cells, cell_type=self.cell_type, history_length=self.history_length, context_length=self.context_length, prediction_length=self.prediction_length, distr_output=self.distr_output, dropout_rate=self.dropout_rate, cardinality=self.cardinality, embedding_dimension=self.embedding_dimension, lags_seq=self.lags_seq, scaling=self.scaling, dtype=self.dtype, ) def create_predictor( self, transformation: Transformation, trained_network: HybridBlock ) -> Predictor: prediction_network = DeepARPredictionNetwork( num_parallel_samples=self.num_parallel_samples, num_layers=self.num_layers, num_cells=self.num_cells, cell_type=self.cell_type, history_length=self.history_length, context_length=self.context_length, prediction_length=self.prediction_length, distr_output=self.distr_output, dropout_rate=self.dropout_rate, cardinality=self.cardinality, embedding_dimension=self.embedding_dimension, lags_seq=self.lags_seq, scaling=self.scaling, dtype=self.dtype, ) copy_parameters(trained_network, prediction_network) return RepresentableBlockPredictor( input_transform=transformation, prediction_net=prediction_network, batch_size=self.trainer.batch_size, freq=self.freq, prediction_length=self.prediction_length, ctx=self.trainer.ctx, dtype=self.dtype, )
12,645
37.090361
94
py
rankpredictor
rankpredictor-master/sub/gluonts/model/wavenet/_network.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports import math from typing import List # Third-party imports import mxnet as mx from mxnet import gluon from mxnet.gluon import nn # First-party imports from gluonts.block.feature import FeatureEmbedder from gluonts.model.common import Tensor from gluonts.core.component import validated class LookupValues(gluon.HybridBlock): def __init__(self, values: mx.nd.NDArray, **kwargs): super().__init__(**kwargs) with self.name_scope(): self.bin_values = self.params.get_constant("bin_values", values) def hybrid_forward(self, F, indices, bin_values): return F.take(bin_values, indices) def conv1d(channels, kernel_size, in_channels, use_bias=True, **kwargs): """ Conv1D with better default initialization. """ n = in_channels kernel_size = ( kernel_size if isinstance(kernel_size, list) else [kernel_size] ) for k in kernel_size: n *= k stdv = 1.0 / math.sqrt(n) winit = mx.initializer.Uniform(stdv) if use_bias: binit = mx.initializer.Uniform(stdv) else: binit = "zeros" return nn.Conv1D( channels=channels, kernel_size=kernel_size, in_channels=in_channels, use_bias=use_bias, weight_initializer=winit, bias_initializer=binit, **kwargs, ) class CausalDilatedResidue(nn.HybridBlock): def __init__( self, n_residue, n_skip, dilation, return_dense_out, kernel_size, **kwargs, ): super().__init__(**kwargs) self.n_residue = n_residue self.n_skip = n_skip self.dilation = dilation self.kernel_size = kernel_size self.return_dense_out = return_dense_out with self.name_scope(): self.conv_sigmoid = conv1d( in_channels=n_residue, channels=n_residue, kernel_size=kernel_size, dilation=dilation, activation="sigmoid", ) self.conv_tanh = conv1d( in_channels=n_residue, channels=n_residue, kernel_size=kernel_size, dilation=dilation, activation="tanh", ) self.skip = conv1d( in_channels=n_residue, channels=n_skip, kernel_size=1 ) self.residue = ( conv1d( in_channels=n_residue, channels=n_residue, kernel_size=1 ) if self.return_dense_out else None ) def hybrid_forward(self, F, x): u = self.conv_sigmoid(x) * self.conv_tanh(x) s = self.skip(u) if not self.return_dense_out: return s, F.zeros(shape=(1,)) output = self.residue(u) output = output + F.slice_axis( x, begin=(self.kernel_size - 1) * self.dilation, end=None, axis=-1 ) return s, output class WaveNet(nn.HybridBlock): def __init__( self, bin_values: List[float], n_residue: int, n_skip: int, dilation_depth: int, n_stacks: int, act_type: str, cardinality: List[int], embedding_dimension: int, pred_length: int, **kwargs, ): super().__init__(**kwargs) self.dilation_depth = dilation_depth self.pred_length = pred_length self.mu = len(bin_values) self.dilations = WaveNet._get_dilations( dilation_depth=dilation_depth, n_stacks=n_stacks ) self.receptive_field = WaveNet.get_receptive_field( dilation_depth=dilation_depth, n_stacks=n_stacks ) self.trim_lengths = [ sum(self.dilations) - sum(self.dilations[: i + 1]) for i, _ in enumerate(self.dilations) ] with self.name_scope(): self.feature_embedder = FeatureEmbedder( cardinalities=cardinality, embedding_dims=[embedding_dimension for _ in cardinality], ) # self.post_transform = LookupValues(mx.nd.array(bin_values)) self.target_embed = nn.Embedding( input_dim=self.mu, output_dim=n_residue ) self.residuals = nn.HybridSequential() for i, d in enumerate(self.dilations): is_not_last = i + 1 < len(self.dilations) self.residuals.add( CausalDilatedResidue( n_residue=n_residue, n_skip=n_skip, dilation=d, return_dense_out=is_not_last, kernel_size=2, ) ) std = 1.0 / math.sqrt(n_residue) self.conv_project = nn.Conv1D( channels=n_residue, kernel_size=1, use_bias=True, weight_initializer=mx.init.Uniform(std), bias_initializer="zero", ) self.conv1 = conv1d( in_channels=n_skip, channels=n_skip, kernel_size=1 ) self.conv2 = conv1d( in_channels=n_skip, channels=self.mu, kernel_size=1 ) self.output_act = ( nn.ELU() if act_type == "elu" else nn.Activation(activation=act_type) ) self.cross_entropy_loss = gluon.loss.SoftmaxCrossEntropyLoss() @staticmethod def _get_dilations(dilation_depth, n_stacks): return [2 ** i for i in range(dilation_depth)] * n_stacks @staticmethod def get_receptive_field(dilation_depth, n_stacks): """ Return the length of the receptive field """ dilations = WaveNet._get_dilations( dilation_depth=dilation_depth, n_stacks=n_stacks ) return sum(dilations) + 1 def hybrid_forward( self, F, feat_static_cat: Tensor, past_target: Tensor, past_observed_values: Tensor, past_time_feat: Tensor, future_time_feat: Tensor, future_target: Tensor, future_observed_values: Tensor, scale: Tensor, ) -> Tensor: embedded_cat = self.feature_embedder(feat_static_cat) static_feat = F.concat(embedded_cat, F.log(scale + 1.0), dim=1) full_target = F.concat(past_target, future_target, dim=-1).astype( "int32" ) full_observed = F.expand_dims( F.concat(past_observed_values, future_observed_values, dim=-1), axis=1, ) full_time_features = F.concat(past_time_feat, future_time_feat, dim=-1) repeated_static_feat = F.repeat( F.expand_dims(static_feat, axis=-1), repeats=self.pred_length + self.receptive_field, axis=-1, ) full_features = F.concat( full_time_features, full_observed, repeated_static_feat, dim=1 ) # (batch_size, embed_dim, sequence_length) o = self.target_embed( F.slice_axis(full_target, begin=0, end=-1, axis=-1) ).swapaxes(1, 2) o = F.concat( o, F.slice_axis(full_features, begin=1, end=None, axis=-1), dim=1 ) o = self.conv_project(o) skip_outs = [] for i, d in enumerate(self.dilations): skip, o = self.residuals[i](o) skip_trim = F.slice_axis( skip, begin=self.trim_lengths[i], end=None, axis=-1 ) skip_outs.append(skip_trim) y = sum(skip_outs) y = self.output_act(y) y = self.conv1(y) y = self.output_act(y) y = self.conv2(y) unnormalized_output = y.swapaxes(1, 2) label = F.slice_axis( full_target, begin=self.receptive_field, end=None, axis=-1 ) loss_weight = F.slice_axis( full_observed, begin=self.receptive_field, end=None, axis=-1 ) loss_weight = F.expand_dims(loss_weight, axis=2) loss = self.cross_entropy_loss(unnormalized_output, label, loss_weight) return loss class WaveNetSampler(WaveNet): """ Runs Wavenet generation in an auto-regressive manner using caching for speedup [PKC+16]_. Same arguments as WaveNet. In addition Parameters ---------- pred_length Length of the prediction horizon num_samples Number of sample paths to generate in parallel in the graph temperature If set to 1.0 (default), sample according to estimated probabilities, if set to 0.0 most likely sample at each step is chosen. post_transform An optional post transform that will be applied to the samples """ @validated() def __init__( self, bin_values: List[float], num_samples: int, temperature: float = 1.0, **kwargs, ): """ Same arguments as WaveNet. In addition :param pred_length: prediction length :param num_samples: number of sample paths to generate in parallel in the graph :param temperature: if set to 1.0 (default), sample according to estimated probabilities - if set to 0.0 most likely sample at each step is chosen. :param post_transform: An optional post transform that will be applied to the samples. """ super().__init__(bin_values=bin_values, **kwargs) self.num_samples = num_samples self.temperature = temperature with self.name_scope(): self.post_transform = LookupValues(mx.nd.array(bin_values)) def hybrid_forward( self, F, feat_static_cat: Tensor, past_target: Tensor, past_observed_values: Tensor, past_time_feat: Tensor, future_time_feat: Tensor, scale: Tensor, ) -> Tensor: embedded_cat = self.feature_embedder(feat_static_cat) static_feat = F.concat(embedded_cat, F.log(scale + 1.0), dim=1) past_target = past_target.astype("int32") def blow_up(u): """ Expand to (batch_size x num_samples) """ return F.repeat(u, repeats=self.num_samples, axis=0) def is_last_layer(i): return i + 1 == len(self.dilations) queues = [] full_time_features = F.concat(past_time_feat, future_time_feat, dim=-1) future_observed_values = F.slice_axis( future_time_feat, begin=0, end=1, axis=1 ).ones_like() full_observed = F.concat( F.expand_dims(past_observed_values, axis=1), future_observed_values, dim=-1, ) repeated_static_feat = F.repeat( F.expand_dims(static_feat, axis=-1), repeats=self.pred_length + self.receptive_field, axis=-1, ) full_features = F.concat( full_time_features, full_observed, repeated_static_feat, dim=1 ) feature_slice = F.slice_axis( full_features, begin=-self.pred_length - self.receptive_field + 1, end=None, axis=-1, ) tmp = F.slice_axis( past_target, begin=-self.receptive_field, end=None, axis=-1 ) o = self.target_embed(tmp).swapaxes(1, 2) o = F.concat( o, F.slice_axis( feature_slice, begin=-self.receptive_field, end=None, axis=-1 ), dim=1, ) o = self.conv_project(o) for i, d in enumerate(self.dilations): sz = 1 if d == 2 ** (self.dilation_depth - 1) else d * 2 _, o = self.residuals[i](o) if not is_last_layer(i): o_chunk = F.slice_axis(o, begin=-sz - 1, end=-1, axis=-1) else: o_chunk = o queues.append(blow_up(o_chunk)) res = F.slice_axis(past_target, begin=-2, end=None, axis=-1) res = blow_up(res) for n in range(self.pred_length): queues_next = [] o = self.target_embed( F.slice_axis(res, begin=-2, end=None, axis=-1) ).swapaxes(1, 2) b = F.slice_axis( full_features, begin=self.receptive_field + n - 1, end=self.receptive_field + n + 1, axis=-1, ) b = blow_up(b) o = F.concat(o, b, dim=1) o = self.conv_project(o) skip_outs = [] for i, d in enumerate(self.dilations): skip, o = self.residuals[i](o) skip_outs.append(skip) if not is_last_layer(i): q = queues[i] o = F.concat(q, o, num_args=2, dim=-1) queues_next.append( F.slice_axis(o, begin=1, end=None, axis=-1) ) queues = queues_next y = sum(skip_outs) y = self.output_act(y) y = self.conv1(y) y = self.output_act(y) unnormalized_outputs = self.conv2(y) if self.temperature > 0: probs = F.softmax( unnormalized_outputs / self.temperature, axis=1 ) y = F.sample_multinomial(probs.swapaxes(1, 2)) else: y = F.argmax(unnormalized_outputs, axis=1) y = y.astype("int32") res = F.concat(res, y, num_args=2, dim=-1) samples = F.slice_axis(res, begin=-self.pred_length, end=None, axis=-1) samples = samples.reshape( shape=(-1, self.num_samples, self.pred_length) ) samples = self.post_transform(samples) samples = F.broadcast_mul(scale.expand_dims(axis=1), samples) return samples
14,582
31.121145
96
py
rankpredictor
rankpredictor-master/sub/gluonts/model/wavenet/_estimator.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports import logging import re from typing import Dict, List, Optional # Third-party imports import mxnet as mx import numpy as np # First-party imports from gluonts import transform from gluonts.core.component import validated from gluonts.dataset.common import Dataset from gluonts.dataset.field_names import FieldName from gluonts.dataset.loader import TrainDataLoader, ValidationDataLoader from gluonts.model.estimator import GluonEstimator from gluonts.model.predictor import Predictor, RepresentableBlockPredictor from gluonts.model.wavenet._network import WaveNet, WaveNetSampler from gluonts.support.util import ( copy_parameters, get_hybrid_forward_input_names, ) from gluonts.time_feature import time_features_from_frequency_str from gluonts.trainer import Trainer from gluonts.transform import ( AddAgeFeature, AddObservedValuesIndicator, AddTimeFeatures, AsNumpyArray, Chain, DataEntry, ExpectedNumInstanceSampler, InstanceSplitter, SetFieldIfNotPresent, SimpleTransformation, VstackFeatures, ) class QuantizeScaled(SimpleTransformation): """ Rescale and quantize the target variable. Requires past_target and future_target fields. The mean absolute value of the past_target is used to rescale past_target and future_target. Then the bin_edges are used to quantize the rescaled target. The calculated scale is included as a new field "scale" """ @validated() def __init__( self, bin_edges: List[float], past_target: str, future_target: str, scale: str = "scale", ): self.bin_edges = np.array(bin_edges) self.future_target = future_target self.past_target = past_target self.scale = scale def transform(self, data: DataEntry) -> DataEntry: p = data[self.past_target] m = np.mean(np.abs(p)) scale = m if m > 0 else 1.0 data[self.future_target] = np.digitize( data[self.future_target] / scale, bins=self.bin_edges, right=False ) data[self.past_target] = np.digitize( data[self.past_target] / scale, bins=self.bin_edges, right=False ) data[self.scale] = np.array([scale]) return data def _get_seasonality(freq: str, seasonality_dict: Dict) -> int: match = re.match(r"(\d*)(\w+)", freq) assert match, "Cannot match freq regex" multiple, base_freq = match.groups() multiple = int(multiple) if multiple else 1 seasonality = seasonality_dict[base_freq] if seasonality % multiple != 0: logging.warning( f"multiple {multiple} does not divide base seasonality {seasonality}." f"Falling back to seasonality 1" ) return 1 return seasonality // multiple class WaveNetEstimator(GluonEstimator): """ Model with Wavenet architecture and quantized target. Parameters ---------- freq Frequency of the data to train on and predict prediction_length Length of the prediction horizon trainer Trainer object to be used (default: Trainer()) cardinality Number of values of the each categorical feature (default: [1]) embedding_dimension Dimension of the embeddings for categorical features (the same dimension is used for all embeddings, default: 5) num_bins Number of bins used for quantization of signal (default: 1024) hybridize_prediction_net Boolean (default: False) n_residue Number of residual channels in wavenet architecture (default: 24) n_skip Number of skip channels in wavenet architecture (default: 32) dilation_depth Number of dilation layers in wavenet architecture. If set to None (default), dialation_depth is set such that the receptive length is at least as long as typical seasonality for the frequency and at least 2 * prediction_length. n_stacks Number of dilation stacks in wavenet architecture (default: 1) temperature Temparature used for sampling from softmax distribution. For temperature = 1.0 (default) sampling is according to estimated probability. act_type Activation type used after before output layer (default: "elu"). Can be any of 'elu', 'relu', 'sigmoid', 'tanh', 'softrelu', 'softsign'. num_parallel_samples Number of evaluation samples per time series to increase parallelism during inference. This is a model optimization that does not affect the accuracy (default: 200) """ @validated() def __init__( self, freq: str, prediction_length: int, trainer: Trainer = Trainer( learning_rate=0.01, epochs=200, num_batches_per_epoch=50, hybridize=False, ), cardinality: List[int] = [1], seasonality: Optional[int] = None, embedding_dimension: int = 5, num_bins: int = 1024, hybridize_prediction_net: bool = False, n_residue=24, n_skip=32, dilation_depth: Optional[int] = None, n_stacks: int = 1, train_window_length: Optional[int] = None, temperature: float = 1.0, act_type: str = "elu", num_parallel_samples: int = 200, ) -> None: """ Model with Wavenet architecture and quantized target. :param freq: :param prediction_length: :param trainer: :param num_eval_samples: :param cardinality: :param embedding_dimension: :param num_bins: Number of bins used for quantization of signal :param hybridize_prediction_net: :param n_residue: Number of residual channels in wavenet architecture :param n_skip: Number of skip channels in wavenet architecture :param dilation_depth: number of dilation layers in wavenet architecture. If set to None, dialation_depth is set such that the receptive length is at least as long as 2 * seasonality for the frequency and at least 2 * prediction_length. :param n_stacks: Number of dilation stacks in wavenet architecture :param train_window_length: Length of windows used for training. This should be longer than context + prediction length. Larger values result in more efficient reuse of computations for convolutions. :param temperature: Temparature used for sampling from softmax distribution. For temperature = 1.0 sampling is according to estimated probability. :param act_type: Activation type used after before output layer. Can be any of 'elu', 'relu', 'sigmoid', 'tanh', 'softrelu', 'softsign' """ super().__init__(trainer=trainer) self.freq = freq self.prediction_length = prediction_length self.cardinality = cardinality self.embedding_dimension = embedding_dimension self.num_bins = num_bins self.hybridize_prediction_net = hybridize_prediction_net self.n_residue = n_residue self.n_skip = n_skip self.n_stacks = n_stacks self.train_window_length = ( train_window_length if train_window_length is not None else prediction_length ) self.temperature = temperature self.act_type = act_type self.num_parallel_samples = num_parallel_samples seasonality = ( _get_seasonality( self.freq, { "H": 7 * 24, "D": 7, "W": 52, "M": 12, "B": 7 * 5, "min": 24 * 60, }, ) if seasonality is None else seasonality ) goal_receptive_length = max( 2 * seasonality, 2 * self.prediction_length ) if dilation_depth is None: d = 1 while ( WaveNet.get_receptive_field( dilation_depth=d, n_stacks=n_stacks ) < goal_receptive_length ): d += 1 self.dilation_depth = d else: self.dilation_depth = dilation_depth self.context_length = WaveNet.get_receptive_field( dilation_depth=self.dilation_depth, n_stacks=n_stacks ) self.logger = logging.getLogger(__name__) self.logger.info( f"Using dilation depth {self.dilation_depth} and receptive field length {self.context_length}" ) def train( self, training_data: Dataset, validation_data: Optional[Dataset] = None ) -> Predictor: has_negative_data = any(np.any(d["target"] < 0) for d in training_data) low = -10.0 if has_negative_data else 0 high = 10.0 bin_centers = np.linspace(low, high, self.num_bins) bin_edges = np.concatenate( [[-1e20], (bin_centers[1:] + bin_centers[:-1]) / 2.0, [1e20]] ) logging.info( f"using training windows of length = {self.train_window_length}" ) transformation = self.create_transformation( bin_edges, pred_length=self.train_window_length ) transformation.estimate(iter(training_data)) training_data_loader = TrainDataLoader( dataset=training_data, transform=transformation, batch_size=self.trainer.batch_size, num_batches_per_epoch=self.trainer.num_batches_per_epoch, ctx=self.trainer.ctx, ) validation_data_loader = None if validation_data is not None: validation_data_loader = ValidationDataLoader( dataset=validation_data, transform=transformation, batch_size=self.trainer.batch_size, ctx=self.trainer.ctx, dtype=self.dtype, ) # ensure that the training network is created within the same MXNet # context as the one that will be used during training with self.trainer.ctx: params = self._get_wavenet_args(bin_centers) params.update(pred_length=self.train_window_length) trained_net = WaveNet(**params) self.trainer( net=trained_net, input_names=get_hybrid_forward_input_names(trained_net), train_iter=training_data_loader, validation_iter=validation_data_loader, ) # ensure that the prediction network is created within the same MXNet # context as the one that was used during training with self.trainer.ctx: return self.create_predictor( transformation, trained_net, bin_centers ) def create_transformation( self, bin_edges: np.ndarray, pred_length: int ) -> transform.Transformation: return Chain( [ AsNumpyArray(field=FieldName.TARGET, expected_ndim=1), AddObservedValuesIndicator( target_field=FieldName.TARGET, output_field=FieldName.OBSERVED_VALUES, ), AddTimeFeatures( start_field=FieldName.START, target_field=FieldName.TARGET, output_field=FieldName.FEAT_TIME, time_features=time_features_from_frequency_str(self.freq), pred_length=self.prediction_length, ), AddAgeFeature( target_field=FieldName.TARGET, output_field=FieldName.FEAT_AGE, pred_length=self.prediction_length, ), VstackFeatures( output_field=FieldName.FEAT_TIME, input_fields=[FieldName.FEAT_TIME, FieldName.FEAT_AGE], ), SetFieldIfNotPresent( field=FieldName.FEAT_STATIC_CAT, value=[0.0] ), AsNumpyArray(field=FieldName.FEAT_STATIC_CAT, expected_ndim=1), InstanceSplitter( target_field=FieldName.TARGET, is_pad_field=FieldName.IS_PAD, start_field=FieldName.START, forecast_start_field=FieldName.FORECAST_START, train_sampler=ExpectedNumInstanceSampler(num_instances=1), past_length=self.context_length, future_length=pred_length, output_NTC=False, time_series_fields=[ FieldName.FEAT_TIME, FieldName.OBSERVED_VALUES, ], ), QuantizeScaled( bin_edges=bin_edges.tolist(), future_target="future_target", past_target="past_target", ), ] ) def _get_wavenet_args(self, bin_centers): return dict( n_residue=self.n_residue, n_skip=self.n_skip, dilation_depth=self.dilation_depth, n_stacks=self.n_stacks, act_type=self.act_type, cardinality=self.cardinality, embedding_dimension=self.embedding_dimension, bin_values=bin_centers.tolist(), pred_length=self.prediction_length, ) def create_predictor( self, transformation: transform.Transformation, trained_network: mx.gluon.HybridBlock, bin_values: np.ndarray, ) -> Predictor: prediction_network = WaveNetSampler( num_samples=self.num_parallel_samples, temperature=self.temperature, **self._get_wavenet_args(bin_values), ) # The lookup layer is specific to the sampling network here # we make sure it is initialized. prediction_network.initialize() copy_parameters( net_source=trained_network, net_dest=prediction_network, allow_missing=True, ) return RepresentableBlockPredictor( input_transform=transformation, prediction_net=prediction_network, batch_size=self.trainer.batch_size, freq=self.freq, prediction_length=self.prediction_length, ctx=self.trainer.ctx, )
15,319
35.563246
106
py
rankpredictor
rankpredictor-master/sub/gluonts/model/deepstate/_network.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import List, Optional # Third-party imports import mxnet as mx # First-party imports from gluonts.block.feature import FeatureEmbedder from gluonts.block.scaler import NOPScaler, MeanScaler from gluonts.core.component import validated from gluonts.distribution.lds import ParameterBounds, LDS, LDSArgsProj from gluonts.model.deepstate.issm import ISSM from gluonts.model.common import Tensor from gluonts.support.util import weighted_average, make_nd_diag class DeepStateNetwork(mx.gluon.HybridBlock): @validated() def __init__( self, num_layers: int, num_cells: int, cell_type: str, past_length: int, prediction_length: int, issm: ISSM, dropout_rate: float, cardinality: List[int], embedding_dimension: List[int], scaling: bool = True, noise_std_bounds: ParameterBounds = ParameterBounds(1e-6, 1.0), prior_cov_bounds: ParameterBounds = ParameterBounds(1e-6, 1.0), innovation_bounds: ParameterBounds = ParameterBounds(1e-6, 0.01), **kwargs, ) -> None: super().__init__(**kwargs) self.num_layers = num_layers self.num_cells = num_cells self.cell_type = cell_type self.past_length = past_length self.prediction_length = prediction_length self.issm = issm self.dropout_rate = dropout_rate self.cardinality = cardinality self.embedding_dimension = embedding_dimension self.num_cat = len(cardinality) self.scaling = scaling assert len(cardinality) == len( embedding_dimension ), "embedding_dimension should be a list with the same size as cardinality" self.univariate = self.issm.output_dim() == 1 self.noise_std_bounds = noise_std_bounds self.prior_cov_bounds = prior_cov_bounds self.innovation_bounds = innovation_bounds with self.name_scope(): self.prior_mean_model = mx.gluon.nn.Dense( units=self.issm.latent_dim(), flatten=False ) self.prior_cov_diag_model = mx.gluon.nn.Dense( units=self.issm.latent_dim(), activation="sigmoid", flatten=False, ) self.lstm = mx.gluon.rnn.HybridSequentialRNNCell() self.lds_proj = LDSArgsProj( output_dim=self.issm.output_dim(), noise_std_bounds=self.noise_std_bounds, innovation_bounds=self.innovation_bounds, ) for k in range(num_layers): cell = mx.gluon.rnn.LSTMCell(hidden_size=num_cells) cell = mx.gluon.rnn.ResidualCell(cell) if k > 0 else cell cell = ( mx.gluon.rnn.ZoneoutCell(cell, zoneout_states=dropout_rate) if dropout_rate > 0.0 else cell ) self.lstm.add(cell) self.embedder = FeatureEmbedder( cardinalities=cardinality, embedding_dims=embedding_dimension ) if scaling: self.scaler = MeanScaler(keepdims=False) else: self.scaler = NOPScaler(keepdims=False) def compute_lds( self, F, feat_static_cat: Tensor, seasonal_indicators: Tensor, time_feat: Tensor, length: int, prior_mean: Optional[Tensor] = None, prior_cov: Optional[Tensor] = None, lstm_begin_state: Optional[List[Tensor]] = None, ): # embed categorical features and expand along time axis embedded_cat = self.embedder(feat_static_cat) repeated_static_features = embedded_cat.expand_dims(axis=1).repeat( axis=1, repeats=length ) # construct big features tensor (context) features = F.concat(time_feat, repeated_static_features, dim=2) output, lstm_final_state = self.lstm.unroll( inputs=features, begin_state=lstm_begin_state, length=length, merge_outputs=True, ) if prior_mean is None: prior_input = F.slice_axis(output, axis=1, begin=0, end=1).squeeze( axis=1 ) prior_mean = self.prior_mean_model(prior_input) prior_cov_diag = ( self.prior_cov_diag_model(prior_input) * (self.prior_cov_bounds.upper - self.prior_cov_bounds.lower) + self.prior_cov_bounds.lower ) prior_cov = make_nd_diag(F, prior_cov_diag, self.issm.latent_dim()) ( emission_coeff, transition_coeff, innovation_coeff, ) = self.issm.get_issm_coeff(seasonal_indicators) noise_std, innovation, residuals = self.lds_proj(output) lds = LDS( emission_coeff=emission_coeff, transition_coeff=transition_coeff, innovation_coeff=F.broadcast_mul(innovation, innovation_coeff), noise_std=noise_std, residuals=residuals, prior_mean=prior_mean, prior_cov=prior_cov, latent_dim=self.issm.latent_dim(), output_dim=self.issm.output_dim(), seq_length=length, ) return lds, lstm_final_state class DeepStateTrainingNetwork(DeepStateNetwork): # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, feat_static_cat: Tensor, past_observed_values: Tensor, past_seasonal_indicators: Tensor, past_time_feat: Tensor, past_target: Tensor, ) -> Tensor: lds, _ = self.compute_lds( F, feat_static_cat=feat_static_cat, seasonal_indicators=past_seasonal_indicators.slice_axis( axis=1, begin=-self.past_length, end=None ), time_feat=past_time_feat.slice_axis( axis=1, begin=-self.past_length, end=None ), length=self.past_length, ) _, scale = self.scaler(past_target, past_observed_values) observed_context = past_observed_values.slice_axis( axis=1, begin=-self.past_length, end=None ) ll, _, _ = lds.log_prob( x=past_target.slice_axis( axis=1, begin=-self.past_length, end=None ), observed=observed_context.min(axis=-1, keepdims=False), scale=scale, ) return weighted_average( F=F, x=-ll, axis=1, weights=observed_context.squeeze(axis=-1) ) class DeepStatePredictionNetwork(DeepStateNetwork): @validated() def __init__(self, num_parallel_samples: int, *args, **kwargs) -> None: super().__init__(*args, **kwargs) self.num_parallel_samples = num_parallel_samples # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, feat_static_cat: Tensor, past_observed_values: Tensor, past_seasonal_indicators: Tensor, past_time_feat: Tensor, past_target: Tensor, future_seasonal_indicators: Tensor, future_time_feat: Tensor, ) -> Tensor: lds, lstm_state = self.compute_lds( F, feat_static_cat=feat_static_cat, seasonal_indicators=past_seasonal_indicators.slice_axis( axis=1, begin=-self.past_length, end=None ), time_feat=past_time_feat.slice_axis( axis=1, begin=-self.past_length, end=None ), length=self.past_length, ) _, scale = self.scaler(past_target, past_observed_values) observed_context = past_observed_values.slice_axis( axis=1, begin=-self.past_length, end=None ) _, final_mean, final_cov = lds.log_prob( x=past_target.slice_axis( axis=1, begin=-self.past_length, end=None ), observed=observed_context.min(axis=-1, keepdims=False), scale=scale, ) lds_prediction, _ = self.compute_lds( F, feat_static_cat=feat_static_cat, seasonal_indicators=future_seasonal_indicators, time_feat=future_time_feat, length=self.prediction_length, lstm_begin_state=lstm_state, prior_mean=final_mean, prior_cov=final_cov, ) samples = lds_prediction.sample( num_samples=self.num_parallel_samples, scale=scale ) # convert samples from # (num_samples, batch_size, prediction_length, target_dim) # to # (batch_size, num_samples, prediction_length, target_dim) # and squeeze last axis in the univariate case if self.univariate: return samples.transpose(axes=(1, 0, 2, 3)).squeeze(axis=3) else: return samples.transpose(axes=(1, 0, 2, 3))
9,652
33.723022
83
py
rankpredictor
rankpredictor-master/sub/gluonts/model/deepstate/_estimator.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports import numpy as np from typing import List, Optional # Third-party imports from mxnet.gluon import HybridBlock from pandas.tseries.frequencies import to_offset # First-party imports from gluonts.core.component import validated from gluonts.dataset.field_names import FieldName from gluonts.distribution.lds import ParameterBounds from gluonts.model.deepstate.issm import ISSM, CompositeISSM from gluonts.model.estimator import GluonEstimator from gluonts.model.predictor import Predictor, RepresentableBlockPredictor from gluonts.support.util import copy_parameters from gluonts.time_feature import TimeFeature, time_features_from_frequency_str from gluonts.trainer import Trainer from gluonts.transform import ( AddObservedValuesIndicator, AddAgeFeature, AddTimeFeatures, AsNumpyArray, Chain, CanonicalInstanceSplitter, ExpandDimArray, RemoveFields, SetField, TestSplitSampler, Transformation, VstackFeatures, ) # Relative imports from ._network import DeepStateTrainingNetwork, DeepStatePredictionNetwork SEASON_INDICATORS_FIELD = "seasonal_indicators" # A dictionary mapping granularity to the period length of the longest season # one can expect given the granularity of the time series. # This is similar to the frequency value in the R forecast package: # https://stats.stackexchange.com/questions/120806/frequency-value-for-seconds-minutes-intervals-data-in-r # This is useful for setting default values for past/context length for models # that do not do data augmentation and uses a single training example per time series in the dataset. FREQ_LONGEST_PERIOD_DICT = { "M": 12, # yearly seasonality "W-SUN": 52, # yearly seasonality "D": 31, # monthly seasonality "B": 22, # monthly seasonality "H": 168, # weekly seasonality "T": 1440, # daily seasonality } def longest_period_from_frequency_str(freq_str: str) -> int: offset = to_offset(freq_str) return FREQ_LONGEST_PERIOD_DICT[offset.name] // offset.n class DeepStateEstimator(GluonEstimator): """ Construct a DeepState estimator. This implements the deep state space model described in [RSG+18]_. Parameters ---------- freq Frequency of the data to train on and predict prediction_length Length of the prediction horizon cardinality Number of values of each categorical feature. This must be set by default unless ``use_feat_static_cat`` is set to `False` explicitly (which is NOT recommended). add_trend Flag to indicate whether to include trend component in the state space model past_length This is the length of the training time series; i.e., number of steps to unroll the RNN for before computing predictions. Set this to (at most) the length of the shortest time series in the dataset. (default: None, in which case the training length is set such that at least `num_seasons_to_train` seasons are included in the training. See `num_seasons_to_train`) num_periods_to_train (Used only when `past_length` is not set) Number of periods to include in the training time series. (default: 4) Here period corresponds to the longest cycle one can expect given the granularity of the time series. See: https://stats.stackexchange.com/questions/120806/frequency -value-for-seconds-minutes-intervals-data-in-r trainer Trainer object to be used (default: Trainer()) num_layers Number of RNN layers (default: 2) num_cells Number of RNN cells for each layer (default: 40) cell_type Type of recurrent cells to use (available: 'lstm' or 'gru'; default: 'lstm') num_parallel_samples Number of evaluation samples per time series to increase parallelism during inference. This is a model optimization that does not affect the accuracy ( default: 100). dropout_rate Dropout regularization parameter (default: 0.1) use_feat_dynamic_real Whether to use the ``feat_dynamic_real`` field from the data (default: False) use_feat_static_cat Whether to use the ``feat_static_cat`` field from the data (default: True) embedding_dimension Dimension of the embeddings for categorical features (default: [min(50, (cat+1)//2) for cat in cardinality]) scaling Whether to automatically scale the target values (default: true) time_features Time features to use as inputs of the RNN (default: None, in which case these are automatically determined based on freq) noise_std_bounds Lower and upper bounds for the standard deviation of the observation noise prior_cov_bounds Lower and upper bounds for the diagonal of the prior covariance matrix innovation_bounds Lower and upper bounds for the standard deviation of the observation noise """ @validated() def __init__( self, freq: str, prediction_length: int, cardinality: List[int], add_trend: bool = False, past_length: Optional[int] = None, num_periods_to_train: int = 4, trainer: Trainer = Trainer( epochs=100, num_batches_per_epoch=50, hybridize=False ), num_layers: int = 2, num_cells: int = 40, cell_type: str = "lstm", num_parallel_samples: int = 100, dropout_rate: float = 0.1, use_feat_dynamic_real: bool = False, use_feat_static_cat: bool = True, embedding_dimension: Optional[List[int]] = None, issm: Optional[ISSM] = None, scaling: bool = True, time_features: Optional[List[TimeFeature]] = None, noise_std_bounds: ParameterBounds = ParameterBounds(1e-6, 1.0), prior_cov_bounds: ParameterBounds = ParameterBounds(1e-6, 1.0), innovation_bounds: ParameterBounds = ParameterBounds(1e-6, 0.01), ) -> None: super().__init__(trainer=trainer) assert ( prediction_length > 0 ), "The value of `prediction_length` should be > 0" assert ( past_length is None or past_length > 0 ), "The value of `past_length` should be > 0" assert num_layers > 0, "The value of `num_layers` should be > 0" assert num_cells > 0, "The value of `num_cells` should be > 0" assert ( num_parallel_samples > 0 ), "The value of `num_parallel_samples` should be > 0" assert dropout_rate >= 0, "The value of `dropout_rate` should be >= 0" assert not use_feat_static_cat or any(c > 1 for c in cardinality), ( f"Cardinality of at least one static categorical feature must be larger than 1 " f"if `use_feat_static_cat=True`. But cardinality provided is: {cardinality}" ) assert embedding_dimension is None or all( e > 0 for e in embedding_dimension ), "Elements of `embedding_dimension` should be > 0" assert all( np.isfinite(p.lower) and np.isfinite(p.upper) and p.lower > 0 for p in [noise_std_bounds, prior_cov_bounds, innovation_bounds] ), "All parameter bounds should be finite, and lower bounds should be positive" self.freq = freq self.past_length = ( past_length if past_length is not None else num_periods_to_train * longest_period_from_frequency_str(freq) ) self.prediction_length = prediction_length self.add_trend = add_trend self.num_layers = num_layers self.num_cells = num_cells self.cell_type = cell_type self.num_parallel_samples = num_parallel_samples self.scaling = scaling self.dropout_rate = dropout_rate self.use_feat_dynamic_real = use_feat_dynamic_real self.use_feat_static_cat = use_feat_static_cat self.cardinality = ( cardinality if cardinality and use_feat_static_cat else [1] ) self.embedding_dimension = ( embedding_dimension if embedding_dimension is not None else [min(50, (cat + 1) // 2) for cat in self.cardinality] ) self.issm = ( issm if issm is not None else CompositeISSM.get_from_freq(freq, add_trend) ) self.time_features = ( time_features if time_features is not None else time_features_from_frequency_str(self.freq) ) self.noise_std_bounds = noise_std_bounds self.prior_cov_bounds = prior_cov_bounds self.innovation_bounds = innovation_bounds def create_transformation(self) -> Transformation: remove_field_names = [ FieldName.FEAT_DYNAMIC_CAT, FieldName.FEAT_STATIC_REAL, ] if not self.use_feat_dynamic_real: remove_field_names.append(FieldName.FEAT_DYNAMIC_REAL) return Chain( [RemoveFields(field_names=remove_field_names)] + ( [SetField(output_field=FieldName.FEAT_STATIC_CAT, value=[0.0])] if not self.use_feat_static_cat else [] ) + [ AsNumpyArray(field=FieldName.FEAT_STATIC_CAT, expected_ndim=1), AsNumpyArray(field=FieldName.TARGET, expected_ndim=1), # gives target the (1, T) layout ExpandDimArray(field=FieldName.TARGET, axis=0), AddObservedValuesIndicator( target_field=FieldName.TARGET, output_field=FieldName.OBSERVED_VALUES, ), # Unnormalized seasonal features AddTimeFeatures( time_features=CompositeISSM.seasonal_features(self.freq), pred_length=self.prediction_length, start_field=FieldName.START, target_field=FieldName.TARGET, output_field=SEASON_INDICATORS_FIELD, ), AddTimeFeatures( start_field=FieldName.START, target_field=FieldName.TARGET, output_field=FieldName.FEAT_TIME, time_features=self.time_features, pred_length=self.prediction_length, ), AddAgeFeature( target_field=FieldName.TARGET, output_field=FieldName.FEAT_AGE, pred_length=self.prediction_length, log_scale=True, ), VstackFeatures( output_field=FieldName.FEAT_TIME, input_fields=[FieldName.FEAT_TIME, FieldName.FEAT_AGE] + ( [FieldName.FEAT_DYNAMIC_REAL] if self.use_feat_dynamic_real else [] ), ), CanonicalInstanceSplitter( target_field=FieldName.TARGET, is_pad_field=FieldName.IS_PAD, start_field=FieldName.START, forecast_start_field=FieldName.FORECAST_START, instance_sampler=TestSplitSampler(), time_series_fields=[ FieldName.FEAT_TIME, SEASON_INDICATORS_FIELD, FieldName.OBSERVED_VALUES, ], allow_target_padding=True, instance_length=self.past_length, use_prediction_features=True, prediction_length=self.prediction_length, ), ] ) def create_training_network(self) -> DeepStateTrainingNetwork: return DeepStateTrainingNetwork( num_layers=self.num_layers, num_cells=self.num_cells, cell_type=self.cell_type, past_length=self.past_length, prediction_length=self.prediction_length, issm=self.issm, dropout_rate=self.dropout_rate, cardinality=self.cardinality, embedding_dimension=self.embedding_dimension, scaling=self.scaling, noise_std_bounds=self.noise_std_bounds, prior_cov_bounds=self.prior_cov_bounds, innovation_bounds=self.innovation_bounds, ) def create_predictor( self, transformation: Transformation, trained_network: HybridBlock ) -> Predictor: prediction_network = DeepStatePredictionNetwork( num_layers=self.num_layers, num_cells=self.num_cells, cell_type=self.cell_type, past_length=self.past_length, prediction_length=self.prediction_length, issm=self.issm, dropout_rate=self.dropout_rate, cardinality=self.cardinality, embedding_dimension=self.embedding_dimension, scaling=self.scaling, num_parallel_samples=self.num_parallel_samples, noise_std_bounds=self.noise_std_bounds, prior_cov_bounds=self.prior_cov_bounds, innovation_bounds=self.innovation_bounds, params=trained_network.collect_params(), ) copy_parameters(trained_network, prediction_network) return RepresentableBlockPredictor( input_transform=transformation, prediction_net=prediction_network, batch_size=self.trainer.batch_size, freq=self.freq, prediction_length=self.prediction_length, ctx=self.trainer.ctx, )
14,462
38.408719
106
py
rankpredictor
rankpredictor-master/sub/gluonts/model/deep_factor/RNNModel.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Third-party imports from mxnet.gluon import HybridBlock, nn # First-party imports from gluonts.block.rnn import RNN from gluonts.core.component import validated class RNNModel(HybridBlock): @validated() def __init__( self, mode, num_hidden, num_layers, num_output, bidirectional=False, **kwargs, ): super(RNNModel, self).__init__(**kwargs) self.num_output = num_output with self.name_scope(): self.rnn = RNN( mode=mode, num_hidden=num_hidden, num_layers=num_layers, bidirectional=bidirectional, ) self.decoder = nn.Dense( num_output, in_units=num_hidden, flatten=False ) def hybrid_forward(self, F, inputs): return self.decoder(self.rnn(inputs))
1,462
28.26
75
py
rankpredictor
rankpredictor-master/sub/gluonts/model/deep_factor/_network.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. import math from mxnet.gluon import HybridBlock from mxnet.gluon import nn # First-party imports from gluonts.block.feature import FeatureEmbedder from gluonts.core.component import validated from gluonts.model.common import Tensor class DeepFactorNetworkBase(HybridBlock): def __init__( self, global_model: HybridBlock, local_model: HybridBlock, embedder: FeatureEmbedder, **kwargs, ) -> None: super().__init__(**kwargs) self.global_model = global_model self.local_model = local_model self.embedder = embedder with self.name_scope(): self.loading = nn.Dense( units=global_model.num_output, use_bias=False ) def assemble_features( self, F, feat_static_cat: Tensor, # (batch_size, 1) time_feat: Tensor, # (batch_size, history_length, num_features) ) -> Tensor: # (batch_size, history_length, num_features) # todo: this is shared by more than one places, and should be a general routine embedded_cat = self.embedder( feat_static_cat ) # (batch_size, num_features * embedding_size) # a workaround when you wish to repeat without knowing the number # of repeats helper_ones = F.ones_like( F.slice_axis(time_feat, axis=2, begin=-1, end=None) ) # (batch_size, history_length, num_features * embedding_size) repeated_cat = F.batch_dot( helper_ones, F.expand_dims(embedded_cat, axis=1) ) # putting together all the features input_feat = F.concat(repeated_cat, time_feat, dim=2) return embedded_cat, input_feat def compute_global_local( self, F, feat_static_cat: Tensor, # (batch_size, 1) time_feat: Tensor, # (batch_size, history_length, num_features) ) -> (Tensor, Tensor): # both of size (batch_size, history_length, 1) cat, local_input = self.assemble_features( F, feat_static_cat, time_feat ) loadings = self.loading(cat) # (batch_size, num_factors) global_factors = self.global_model( time_feat ) # (batch_size, history_length, num_factors) fixed_effect = F.batch_dot( global_factors, loadings.expand_dims(axis=2) ) # (batch_size, history_length, 1) random_effect = F.log( F.exp(self.local_model(local_input)) + 1.0 ) # (batch_size, history_length, 1) return F.exp(fixed_effect), random_effect def hybrid_forward(self, F, x, *args, **kwargs): raise NotImplementedError def negative_normal_likelihood(self, F, y, mu, sigma): return ( F.log(sigma) + 0.5 * math.log(2 * math.pi) + 0.5 * F.square((y - mu) / sigma) ) class DeepFactorTrainingNetwork(DeepFactorNetworkBase): def hybrid_forward( self, F, feat_static_cat: Tensor, # (batch_size, 1) past_time_feat: Tensor, # (batch_size, history_length, num_features) past_target: Tensor, # (batch_size, history_length) ) -> Tensor: """ Parameters ---------- F Function space feat_static_cat Shape: (batch_size, 1) past_time_feat Shape: (batch_size, history_length, num_features) past_target Shape: (batch_size, history_length) Returns ------- Tensor A batch of negative log likelihoods. """ fixed_effect, random_effect = self.compute_global_local( F, feat_static_cat, past_time_feat ) loss = self.negative_normal_likelihood( F, past_target.expand_dims(axis=2), fixed_effect, random_effect ) return loss class DeepFactorPredictionNetwork(DeepFactorNetworkBase): @validated() def __init__( self, prediction_len: int, num_parallel_samples: int, **kwargs ) -> None: super().__init__(**kwargs) self.prediction_len = prediction_len self.num_parallel_samples = num_parallel_samples def hybrid_forward( self, F, feat_static_cat: Tensor, past_time_feat: Tensor, future_time_feat: Tensor, past_target: Tensor, ) -> Tensor: """ Parameters ---------- F Function space feat_static_cat Shape: (batch_size, 1) past_time_feat Shape: (batch_size, history_length, num_features) future_time_feat Shape: (batch_size, prediction_length, num_features) past_target Shape: (batch_size, history_length) Returns ------- Tensor Samples of shape (batch_size, num_samples, prediction_length). """ time_feat = F.concat(past_time_feat, future_time_feat, dim=1) fixed_effect, random_effect = self.compute_global_local( F, feat_static_cat, time_feat ) samples = F.concat( *[ F.sample_normal(fixed_effect, random_effect) for _ in range(self.num_parallel_samples) ], dim=2, ) # (batch_size, train_len + prediction_len, num_samples) pred_samples = F.slice_axis( samples, axis=1, begin=-self.prediction_len, end=None ) # (batch_size, prediction_len, num_samples) return pred_samples.swapaxes(1, 2)
6,153
30.88601
87
py
rankpredictor
rankpredictor-master/sub/gluonts/model/seq2seq/_seq2seq_estimator.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import List, Optional # Third-party imports import mxnet as mx # First-party imports from gluonts import transform from gluonts.block.decoder import OneShotDecoder from gluonts.block.enc2dec import PassThroughEnc2Dec from gluonts.block.encoder import ( HierarchicalCausalConv1DEncoder, RNNCovariateEncoder, MLPEncoder, Seq2SeqEncoder, ) from gluonts.block.feature import FeatureEmbedder from gluonts.block.quantile_output import QuantileOutput from gluonts.block.scaler import NOPScaler, Scaler from gluonts.core.component import validated from gluonts.dataset.field_names import FieldName from gluonts.model.estimator import GluonEstimator from gluonts.model.forecast import QuantileForecast, Quantile from gluonts.model.predictor import Predictor, RepresentableBlockPredictor from gluonts.support.util import copy_parameters from gluonts.time_feature import time_features_from_frequency_str from gluonts.trainer import Trainer from gluonts.transform import ExpectedNumInstanceSampler from gluonts.model.forecast_generator import QuantileForecastGenerator # Relative imports from ._seq2seq_network import Seq2SeqPredictionNetwork, Seq2SeqTrainingNetwork class Seq2SeqEstimator(GluonEstimator): """ Quantile-Regression Sequence-to-Sequence Estimator """ @validated() def __init__( self, freq: str, prediction_length: int, cardinality: List[int], embedding_dimension: int, encoder: Seq2SeqEncoder, decoder_mlp_layer: List[int], decoder_mlp_static_dim: int, scaler: Scaler = NOPScaler(), context_length: Optional[int] = None, quantiles: List[float] = [0.1, 0.5, 0.9], trainer: Trainer = Trainer(), num_parallel_samples: int = 100, ) -> None: assert ( prediction_length > 0 ), "The value of `prediction_length` should be > 0" assert ( context_length is None or context_length > 0 ), "The value of `context_length` should be > 0" super().__init__(trainer=trainer) self.context_length = ( context_length if context_length is not None else prediction_length ) self.prediction_length = prediction_length self.freq = freq self.quantiles = quantiles self.encoder = encoder self.decoder_mlp_layer = decoder_mlp_layer self.decoder_mlp_static_dim = decoder_mlp_static_dim self.scaler = scaler self.embedder = FeatureEmbedder( cardinalities=cardinality, embedding_dims=[embedding_dimension for _ in cardinality], ) self.num_parallel_samples = num_parallel_samples def create_transformation(self) -> transform.Transformation: return transform.Chain( trans=[ transform.AsNumpyArray( field=FieldName.TARGET, expected_ndim=1 ), transform.AddTimeFeatures( start_field=FieldName.START, target_field=FieldName.TARGET, output_field=FieldName.FEAT_TIME, time_features=time_features_from_frequency_str(self.freq), pred_length=self.prediction_length, ), transform.VstackFeatures( output_field=FieldName.FEAT_DYNAMIC_REAL, input_fields=[FieldName.FEAT_TIME], ), transform.SetFieldIfNotPresent( field=FieldName.FEAT_STATIC_CAT, value=[0.0] ), transform.AsNumpyArray( field=FieldName.FEAT_STATIC_CAT, expected_ndim=1 ), transform.InstanceSplitter( target_field=FieldName.TARGET, is_pad_field=FieldName.IS_PAD, start_field=FieldName.START, forecast_start_field=FieldName.FORECAST_START, train_sampler=ExpectedNumInstanceSampler(num_instances=1), past_length=self.context_length, future_length=self.prediction_length, time_series_fields=[FieldName.FEAT_DYNAMIC_REAL], ), ] ) def create_training_network(self) -> mx.gluon.HybridBlock: distribution = QuantileOutput(self.quantiles) enc2dec = PassThroughEnc2Dec() decoder = OneShotDecoder( decoder_length=self.prediction_length, layer_sizes=self.decoder_mlp_layer, static_outputs_per_time_step=self.decoder_mlp_static_dim, ) training_network = Seq2SeqTrainingNetwork( embedder=self.embedder, scaler=self.scaler, encoder=self.encoder, enc2dec=enc2dec, decoder=decoder, quantile_output=distribution, ) return training_network def create_predictor( self, transformation: transform.Transformation, trained_network: Seq2SeqTrainingNetwork, ) -> Predictor: # todo: this is specific to quantile output quantile_strs = [ Quantile.from_float(quantile).name for quantile in self.quantiles ] prediction_network = Seq2SeqPredictionNetwork( embedder=trained_network.embedder, scaler=trained_network.scaler, encoder=trained_network.encoder, enc2dec=trained_network.enc2dec, decoder=trained_network.decoder, quantile_output=trained_network.quantile_output, ) copy_parameters(trained_network, prediction_network) return RepresentableBlockPredictor( input_transform=transformation, prediction_net=prediction_network, batch_size=self.trainer.batch_size, freq=self.freq, prediction_length=self.prediction_length, ctx=self.trainer.ctx, forecast_generator=QuantileForecastGenerator(quantile_strs), ) class MLP2QRForecaster(Seq2SeqEstimator): @validated() def __init__( self, freq: str, prediction_length: int, cardinality: List[int], embedding_dimension: int, encoder_mlp_layer: List[int], decoder_mlp_layer: List[int], decoder_mlp_static_dim: int, scaler: Scaler = NOPScaler, context_length: Optional[int] = None, quantiles: List[float] = list([0.1, 0.5, 0.9]), trainer: Trainer = Trainer(), num_parallel_samples: int = 100, ) -> None: encoder = MLPEncoder(layer_sizes=encoder_mlp_layer) super(MLP2QRForecaster, self).__init__( freq=freq, prediction_length=prediction_length, encoder=encoder, cardinality=cardinality, embedding_dimension=embedding_dimension, decoder_mlp_layer=decoder_mlp_layer, decoder_mlp_static_dim=decoder_mlp_static_dim, context_length=context_length, scaler=scaler, quantiles=quantiles, trainer=trainer, num_parallel_samples=num_parallel_samples, ) class RNN2QRForecaster(Seq2SeqEstimator): @validated() def __init__( self, freq: str, prediction_length: int, cardinality: List[int], embedding_dimension: int, encoder_rnn_layer: int, encoder_rnn_num_hidden: int, decoder_mlp_layer: List[int], decoder_mlp_static_dim: int, encoder_rnn_model: str = "lstm", encoder_rnn_bidirectional: bool = True, scaler: Scaler = NOPScaler, context_length: Optional[int] = None, quantiles: List[float] = list([0.1, 0.5, 0.9]), trainer: Trainer = Trainer(), num_parallel_samples: int = 100, ) -> None: encoder = RNNCovariateEncoder( mode=encoder_rnn_model, hidden_size=encoder_rnn_num_hidden, num_layers=encoder_rnn_layer, bidirectional=encoder_rnn_bidirectional, ) super(RNN2QRForecaster, self).__init__( freq=freq, prediction_length=prediction_length, encoder=encoder, cardinality=cardinality, embedding_dimension=embedding_dimension, decoder_mlp_layer=decoder_mlp_layer, decoder_mlp_static_dim=decoder_mlp_static_dim, context_length=context_length, scaler=scaler, quantiles=quantiles, trainer=trainer, num_parallel_samples=num_parallel_samples, ) class CNN2QRForecaster(Seq2SeqEstimator): @validated() def __init__( self, freq: str, prediction_length: int, cardinality: List[int], embedding_dimension: int, decoder_mlp_layer: List[int], decoder_mlp_static_dim: int, scaler: Scaler = NOPScaler, context_length: Optional[int] = None, quantiles: List[float] = list([0.1, 0.5, 0.9]), trainer: Trainer = Trainer(), num_parallel_samples: int = 100, ) -> None: encoder = HierarchicalCausalConv1DEncoder( dilation_seq=[1, 3, 9], kernel_size_seq=([3] * len([30, 30, 30])), channels_seq=[30, 30, 30], use_residual=True, use_covariates=True, ) super(CNN2QRForecaster, self).__init__( freq=freq, prediction_length=prediction_length, encoder=encoder, cardinality=cardinality, embedding_dimension=embedding_dimension, decoder_mlp_layer=decoder_mlp_layer, decoder_mlp_static_dim=decoder_mlp_static_dim, context_length=context_length, scaler=scaler, quantiles=quantiles, trainer=trainer, num_parallel_samples=num_parallel_samples, )
10,621
34.644295
79
py
rankpredictor
rankpredictor-master/sub/gluonts/model/seq2seq/_forking_network.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Third-party imports from mxnet import gluon, nd # First-party imports from gluonts.block.decoder import Seq2SeqDecoder from gluonts.block.enc2dec import Seq2SeqEnc2Dec from gluonts.block.encoder import Seq2SeqEncoder from gluonts.block.quantile_output import QuantileOutput from gluonts.core.component import validated from gluonts.model.common import Tensor nd_None = nd.array([]) class ForkingSeq2SeqNetworkBase(gluon.HybridBlock): """ Base network for the :class:`ForkingSeq2SeqEstimator`. Parameters ---------- encoder: Seq2SeqEncoder encoder block enc2dec: Seq2SeqEnc2Dec encoder to decoder mapping block decoder: Seq2SeqDecoder decoder block quantile_output: QuantileOutput quantile output block kwargs: dict dictionary of Gluon HybridBlock parameters """ @validated() def __init__( self, encoder: Seq2SeqEncoder, enc2dec: Seq2SeqEnc2Dec, decoder: Seq2SeqDecoder, quantile_output: QuantileOutput, **kwargs, ) -> None: super().__init__(**kwargs) self.encoder = encoder self.enc2dec = enc2dec self.decoder = decoder self.quantile_output = quantile_output with self.name_scope(): self.quantile_proj = quantile_output.get_quantile_proj() self.loss = quantile_output.get_loss() class ForkingSeq2SeqTrainingNetwork(ForkingSeq2SeqNetworkBase): # noinspection PyMethodOverriding def hybrid_forward( self, F, past_target: Tensor, future_target: Tensor ) -> Tensor: """ Parameters ---------- F: mx.symbol or mx.ndarray Gluon function space past_target: Tensor FIXME future_target: Tensor shape (num_ts, encoder_length, 1) FIXME Returns ------- loss with shape (FIXME, FIXME) """ # FIXME: can we factor out a common prefix in the base network? feat_static_real = nd_None past_feat_dynamic_real = nd_None future_feat_dynamic_real = nd_None enc_output_static, enc_output_dynamic = self.encoder( past_target, feat_static_real, past_feat_dynamic_real ) dec_input_static, dec_input_dynamic, _ = self.enc2dec( enc_output_static, enc_output_dynamic, future_feat_dynamic_real ) dec_output = self.decoder(dec_input_dynamic, dec_input_static) dec_dist_output = self.quantile_proj(dec_output) loss = self.loss(future_target, dec_dist_output) return loss.mean(axis=1) class ForkingSeq2SeqPredictionNetwork(ForkingSeq2SeqNetworkBase): # noinspection PyMethodOverriding def hybrid_forward(self, F, past_target: Tensor) -> Tensor: """ Parameters ---------- F: mx.symbol or mx.ndarray Gluon function space past_target: Tensor FIXME Returns ------- prediction tensor with shape (FIXME, FIXME) """ # FIXME: can we factor out a common prefix in the base network? feat_static_real = nd_None past_feat_dynamic_real = nd_None future_feat_dynamic_real = nd_None enc_output_static, enc_output_dynamic = self.encoder( past_target, feat_static_real, past_feat_dynamic_real ) enc_output_static = ( nd_None if enc_output_static is None else enc_output_static ) dec_inp_static, dec_inp_dynamic, _ = self.enc2dec( enc_output_static, enc_output_dynamic, future_feat_dynamic_real ) dec_output = self.decoder(dec_inp_dynamic, dec_inp_static) fcst_output = F.slice_axis(dec_output, axis=1, begin=-1, end=None) fcst_output = F.squeeze(fcst_output, axis=1) predictions = self.quantile_proj(fcst_output).swapaxes(2, 1) return predictions
4,527
30.013699
75
py
rankpredictor
rankpredictor-master/sub/gluonts/model/seq2seq/_seq2seq_network.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Third-party imports import mxnet as mx # First-party imports from gluonts.block.decoder import Seq2SeqDecoder from gluonts.block.enc2dec import Seq2SeqEnc2Dec from gluonts.block.encoder import Seq2SeqEncoder from gluonts.block.feature import FeatureEmbedder from gluonts.block.quantile_output import QuantileOutput from gluonts.block.scaler import Scaler from gluonts.core.component import validated from gluonts.model.common import Tensor class Seq2SeqNetworkBase(mx.gluon.HybridBlock): """ Base network for the :class:`Seq2SeqEstimator`. Parameters ---------- scaler : Scaler scale of the target time series, both as input or in the output distributions encoder : encoder see encoder.py for possible choices enc2dec : encoder to decoder see enc2dec.py for possible choices decoder : decoder see decoder.py for possible choices quantile_output : QuantileOutput quantile regression output kwargs : dict a dict of parameters to be passed to the parent initializer """ @validated() def __init__( self, embedder: FeatureEmbedder, scaler: Scaler, encoder: Seq2SeqEncoder, enc2dec: Seq2SeqEnc2Dec, decoder: Seq2SeqDecoder, quantile_output: QuantileOutput, **kwargs, ) -> None: super().__init__(**kwargs) self.embedder = embedder self.scaler = scaler self.encoder = encoder self.enc2dec = enc2dec self.decoder = decoder self.quantile_output = quantile_output with self.name_scope(): self.quantile_proj = quantile_output.get_quantile_proj() self.loss = quantile_output.get_loss() def compute_decoder_outputs( self, F, past_target: Tensor, feat_static_cat: Tensor, past_feat_dynamic_real: Tensor, future_feat_dynamic_real: Tensor, ) -> Tensor: scaled_target, scale = self.scaler( past_target, F.ones_like(past_target) ) embedded_cat = self.embedder( feat_static_cat ) # (batch_size, num_features * embedding_size) encoder_output_static, encoder_output_dynamic = self.encoder( scaled_target, embedded_cat, past_feat_dynamic_real ) decoder_input_static, _, decoder_input_dynamic = self.enc2dec( encoder_output_static, encoder_output_dynamic, future_feat_dynamic_real, ) decoder_output = self.decoder( decoder_input_static, decoder_input_dynamic ) scaled_decoder_output = F.broadcast_mul( decoder_output, scale.expand_dims(-1).expand_dims(-1) ) return scaled_decoder_output class Seq2SeqTrainingNetwork(Seq2SeqNetworkBase): # noinspection PyMethodOverriding def hybrid_forward( self, F, past_target: Tensor, future_target: Tensor, feat_static_cat: Tensor, past_feat_dynamic_real: Tensor, future_feat_dynamic_real: Tensor, ) -> Tensor: """ Parameters ---------- F: mx.symbol or mx.ndarray Gluon function space past_target: mx.nd.NDArray or mx.sym.Symbol past target future_target: mx.nd.NDArray or mx.sym.Symbol future target feat_static_cat: mx.nd.NDArray or mx.sym.Symbol static categorical features past_feat_dynamic_real: mx.nd.NDArray or mx.sym.Symbol past dynamic real-valued features future_feat_dynamic_real: mx.nd.NDArray or mx.sym.Symbol future dynamic real-valued features Returns ------- mx.nd.NDArray or mx.sym.Symbol the computed loss """ scaled_decoder_output = self.compute_decoder_outputs( F, past_target=past_target, feat_static_cat=feat_static_cat, past_feat_dynamic_real=past_feat_dynamic_real, future_feat_dynamic_real=future_feat_dynamic_real, ) projected = self.quantile_proj(scaled_decoder_output) loss = self.loss(future_target, projected) # TODO: there used to be "nansum" here, to be fully equivalent we # TODO: should have a "nanmean" here # TODO: shouldn't we sum and divide by the number of observed values # TODO: here? return loss class Seq2SeqPredictionNetwork(Seq2SeqNetworkBase): # noinspection PyMethodOverriding def hybrid_forward( self, F, past_target: Tensor, feat_static_cat: Tensor, past_feat_dynamic_real: Tensor, future_feat_dynamic_real: Tensor, ) -> Tensor: """ Parameters ---------- F: mx.symbol or mx.ndarray Gluon function space past_target: mx.nd.NDArray or mx.sym.Symbol past target feat_static_cat: mx.nd.NDArray or mx.sym.Symbol static categorical features past_feat_dynamic_real: mx.nd.NDArray or mx.sym.Symbol past dynamic real-valued features future_feat_dynamic_real: mx.nd.NDArray or mx.sym.Symbol future dynamic real-valued features Returns ------- mx.nd.NDArray or mx.sym.Symbol the predicted sequence """ scaled_decoder_output = self.compute_decoder_outputs( F, past_target=past_target, feat_static_cat=feat_static_cat, past_feat_dynamic_real=past_feat_dynamic_real, future_feat_dynamic_real=future_feat_dynamic_real, ) predictions = self.quantile_proj(scaled_decoder_output).swapaxes(2, 1) return predictions
6,383
31.738462
78
py
rankpredictor
rankpredictor-master/sub/gluonts/model/transformer/trans_encoder.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import Dict # Third-party imports from mxnet.gluon import HybridBlock # First-party imports from gluonts.core.component import validated from gluonts.model.common import Tensor from gluonts.model.transformer.layers import ( TransformerProcessBlock, TransformerFeedForward, MultiHeadSelfAttention, InputLayer, ) class TransformerEncoder(HybridBlock): @validated() def __init__(self, encoder_length: int, config: Dict, **kwargs) -> None: super().__init__(**kwargs) self.encoder_length = encoder_length with self.name_scope(): self.enc_input_layer = InputLayer(model_size=config["model_dim"]) self.enc_pre_self_att = TransformerProcessBlock( sequence=config["pre_seq"], dropout=config["dropout_rate"], prefix="pretransformerprocessblock_", ) self.enc_self_att = MultiHeadSelfAttention( att_dim_in=config["model_dim"], heads=config["num_heads"], att_dim_out=config["model_dim"], dropout=config["dropout_rate"], prefix="multiheadselfattention_", ) self.enc_post_self_att = TransformerProcessBlock( sequence=config["post_seq"], dropout=config["dropout_rate"], prefix="postselfatttransformerprocessblock_", ) self.enc_ff = TransformerFeedForward( inner_dim=config["model_dim"] * config["inner_ff_dim_scale"], out_dim=config["model_dim"], act_type=config["act_type"], dropout=config["dropout_rate"], prefix="transformerfeedforward_", ) self.enc_post_ff = TransformerProcessBlock( sequence=config["post_seq"], dropout=config["dropout_rate"], prefix="postfftransformerprocessblock_", ) # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward(self, F, data: Tensor) -> Tensor: """ A transformer encoder block consists of a self-attention and a feed-forward layer with pre/post process blocks in between. """ # input layer inputs = self.enc_input_layer(data) # self-attention data_self_att, _ = self.enc_self_att( self.enc_pre_self_att(inputs, None) ) data = self.enc_post_self_att(data_self_att, inputs) # feed-forward data_ff = self.enc_ff(data) data = self.enc_post_ff(data_ff, data) return data
3,242
33.5
118
py
rankpredictor
rankpredictor-master/sub/gluonts/model/transformer/layers.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import Dict, Optional, Tuple # Third-party imports import mxnet as mx from mxnet.gluon import HybridBlock # First-party imports from gluonts.model.common import Tensor def split_heads(F, x: Tensor, dim_per_head: int, heads: int) -> Tensor: r""" Returns a tensor with head dimension folded into batch and last dimension divided by the number of heads. Parameters ---------- x Tensor of shape (batch_size, time_length, dim). dim_per_head Dimension per head heads Number of heads Returns ------- Tensor of shape (batch_size * heads, time_length, dim_per_head). """ # (batch_size, time_length, heads, dim_per_head) x = F.reshape(data=x, shape=(0, -1, heads, dim_per_head)) # (batch_size, heads, time_length, dim/heads) x = F.transpose(data=x, axes=(0, 2, 1, 3)) # (batch_size * heads, time_length, dim/heads) return F.reshape(data=x, shape=(-3, -1, dim_per_head)) def dot_attention( F, queries: Tensor, keys: Tensor, values: Tensor, mask: Optional[Tensor] = None, dropout: float = 0.0, ) -> Tensor: r""" Parameters ---------- queries Attention queries of shape (n, lq, d) keys Attention keys of shape (n, lk, d) values Attention values of shape (n, lk, dv) mask Optional mask tensor dropout Dropout rate Returns ------- 'Context' vectors for each query of shape (n, lq, dv) """ # (n, lq, lk) logits = F.batch_dot(lhs=queries, rhs=keys, transpose_b=True) if mask is not None: logits = F.broadcast_add(logits, mask) probs = F.softmax(logits, axis=-1) probs = F.Dropout(probs, p=dropout) if dropout > 0.0 else probs # (n, lq, lk) x (n, lk, dv) -> (n, lq, dv) return F.batch_dot(lhs=probs, rhs=values) def combine_heads(F, x: Tensor, dim_per_head: int, heads: int) -> Tensor: r""" Parameters ---------- x Tensor of shape (batch_size * heads, time_length, dim_per_head) dim_per_head Dimension per head heads Number of heads Returns ------- Tensor of shape (batch_size, time_length, dim) """ # (batch_size, heads, time_length, dim_per_head) x = F.reshape(data=x, shape=(-4, -1, heads, 0, dim_per_head)) # (batch_size, time_length, heads, dim_per_head) x = F.transpose(x, axes=(0, 2, 1, 3)) # (batch_size, time_length, dim) return F.reshape(x, shape=(-1, 0, dim_per_head * heads)) class LayerNormalization(HybridBlock): """ Implements layer normalization as proposed in [BKH16]_. """ def __init__( self, scale_init: str = "ones", shift_init: str = "zeros", eps: float = 1e-06, **kwargs, ) -> None: super().__init__(**kwargs) self.scale_init = scale_init self.shift_init = shift_init with self.name_scope(): self.lnorm = mx.gluon.nn.LayerNorm( axis=-1, gamma_initializer=self.scale_init, beta_initializer=self.shift_init, epsilon=eps, ) # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward(self, F, data: Tensor) -> Tensor: r""" Normalizes hidden units of data as follows: data = scale * (data - mean) / sqrt(var + eps) + shift Normalization is performed over the last dimension of the input data. Parameters ---------- data Data to normalize of shape (d0, ..., dn, num_hidden) Returns ------- Normalized inputs of shape: (d0, ..., dn, num_hidden) """ return self.lnorm(data) class InputLayer(HybridBlock): r""" Transforms the input vector to model_size with an one-layer MPL, i.e., (batch_size, time_length, input_dim) -> (batch_size, time_length, model_size) """ def __init__(self, model_size: int = 64, **kwargs) -> None: super().__init__(**kwargs) self.model_size = model_size with self.name_scope(): self.net = mx.gluon.nn.Dense(units=self.model_size, flatten=False) def hybrid_forward(self, F, data: Tensor, *args): return self.net(data) class MultiHeadAttentionBase(HybridBlock): """ Base class for Multi-head attention. Parameters ---------- att_dim_in Attention dimension (number of hidden units) heads Number of attention heads att_dim_out Output dimension (number of output units) dropout Dropout rate on attention scores """ def __init__( self, att_dim_in: int = 32, heads: int = 8, att_dim_out: int = 32, dropout: float = 0.0, **kwargs, ) -> None: super().__init__(**kwargs) assert ( att_dim_in % heads == 0 ), "Number of heads {} must divide attention att_dim_in {}".format( heads, att_dim_in ) self.att_dim_in = att_dim_in self.heads = heads self.att_dim_out = att_dim_out self.dropout = dropout self.dim_per_head = self.att_dim_in // self.heads with self.name_scope(): self.dense_att = mx.gluon.nn.Dense( units=self.att_dim_out, flatten=False ) def _attend( self, F, queries: Tensor, keys: Tensor, values: Tensor, mask: Optional[Tensor] = None, ) -> Tensor: r""" Returns context vectors of multi-head dot attention. Parameters ---------- queries Queries tensor of shape (batch_size, query_max_length, dim) keys Keys tensor of shape (batch_size, memory_max_length, dim) values Values tensor of shape (batch_size, memory_max_length, dim) mask Returns ------- Context vectors of shape (batch_size, query_max_length, att_dim_out) """ # scale by 1/sqrt(dim_per_head) queries = queries * (self.dim_per_head ** -0.5) # (batch_size * heads, length, dim/heads) queries = split_heads(F, queries, self.dim_per_head, self.heads) keys = split_heads(F, keys, self.dim_per_head, self.heads) values = split_heads(F, values, self.dim_per_head, self.heads) # (batch_size * heads, query_max_length, dim_per_head) contexts = dot_attention( F, queries, keys, values, mask=mask, dropout=self.dropout ) # (batch_size, query_max_length, input_dim) contexts = combine_heads(F, contexts, self.dim_per_head, self.heads) # contexts: (batch_size, query_max_length, output_dim) contexts = self.dense_att(contexts) return contexts def hybrid_forward(self, F, *args, **kwargs): raise NotImplementedError class MultiHeadSelfAttention(MultiHeadAttentionBase): r""" Multi-head self-attention. Independent linear projections of inputs serve as queries, keys, and values for the attention. Parameters ---------- att_dim_in Attention dimension (number of hidden units) heads Number of attention heads att_dim_out Output dimension (number of output units) dropout Dropout rate on attention scores """ def __init__( self, att_dim_in: int = 32, heads: int = 8, att_dim_out: int = 32, dropout: float = 0.0, **kwargs, ) -> None: super().__init__(att_dim_in, heads, att_dim_out, dropout, **kwargs) with self.name_scope(): self.dense_pre_satt = mx.gluon.nn.Dense( units=self.att_dim_in * 3, flatten=False ) # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, inputs: Tensor, mask: Optional[Tensor] = None, cache: Optional[Dict[str, Optional[Tensor]]] = None, ) -> Tuple[Tensor, Optional[Dict]]: r""" Computes multi-head attention on a set of inputs, serving as queries, keys, and values. If sequence lengths are provided, they will be used to mask the attention scores. May also use a cache of previously computed inputs. Parameters ---------- inputs Input data of shape (batch_size, max_length, att_dim_in) mask Optional tensor to mask attention scores cache Optional dictionary of previously computed keys and values Returns ------- Tensor A tensor of shape (batch_size, max_length, att_dim_out) """ # Q = K = V -> Q * W_q, K * W_k, V * W_v # combined: (batch_size, max_length, att_dim_in * 3) combined = self.dense_pre_satt(inputs) # split into queries, keys and values # (batch_size, max_length, att_dim_in) queries, keys, values = F.split(data=combined, num_outputs=3, axis=2) if cache is not None: # append new keys and values to cache, update the cache keys = cache["k"] = ( keys if "k" not in cache.keys() else F.concat(cache["k"], keys, dim=1) ) values = cache["v"] = ( values if "v" not in cache.keys() else F.concat(cache["v"], values, dim=1) ) return self._attend(F, queries, keys, values, mask), cache class MultiHeadAttention(MultiHeadAttentionBase): r""" Multi-head attention layer for queries independent from keys/values. Parameters ---------- att_dim_in Attention dimension (number of hidden units) heads Number of attention heads att_dim_out Output dimension (number of output units) dropout Dropout rate on attention scores """ def __init__( self, att_dim_in: int = 32, heads: int = 8, att_dim_out: int = 32, dropout: float = 0.0, **kwargs, ) -> None: super().__init__(att_dim_in, heads, att_dim_out, dropout, **kwargs) with self.name_scope(): self.dense_pre_att_q = mx.gluon.nn.Dense( units=self.att_dim_in, flatten=False ) self.dense_pre_att_k = mx.gluon.nn.Dense( units=self.att_dim_in, flatten=False ) self.dense_pre_att_v = mx.gluon.nn.Dense( units=self.att_dim_in, flatten=False ) # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, queries: Tensor, memory: Tensor, mask: Optional[Tensor] = None ) -> Tensor: r""" Computes multi-head attention for queries given a memory tensor. If sequence lengths are provided, they will be used to mask the attention scores. A mask tensor may also be used to mask the attention scores. Returns a tensor of shape (batch_size, max_length, att_dim_out). Parameters ---------- queries Queries tensor of shape (batch_size, query_max_length, att_dim_in) memory Memory tensor to attend to of shape (batch_size, memory_max_length, att_dim_in) mask Optional tensor to mask attention scores Returns ------- Tensor of shape (batch_size, query_seq_len, att_dim_out) """ # Q -> Q * W_q # K = V -> K * W_k, V * W_v # (batch, query_max_length, att_dim_in) queries = self.dense_pre_att_q(queries) # (batch, memory_max_length, att_dim_in) keys = self.dense_pre_att_k(memory) # (batch, memory_max_length, att_dim_in) values = self.dense_pre_att_v(memory) return self._attend(F, queries, keys, values, mask=mask) class TransformerFeedForward(HybridBlock): r""" Position-wise feed-forward network with activation. .. math:: activation(XW_1 + b_1)W_2 + b_2 :math:`W_1`: (batch_size, d, inner_dim) :math:`W_2`: (batch_size, inner_dim, out_dim) """ def __init__( self, inner_dim: int = 32, # W1: (batch_size, d, inner_dim) out_dim: int = 32, # W2: (batch_size, inner_dim, out_dim) act_type: str = "softrelu", dropout: float = 0.0, **kwargs, ) -> None: super().__init__(**kwargs) self.inner_dim = inner_dim self.out_dim = out_dim self.dropout = dropout self.act_type = act_type with self.name_scope(): self.mlp = mx.gluon.nn.HybridSequential() self.mlp.add( mx.gluon.nn.Dense( units=self.inner_dim, use_bias=True, activation=self.act_type, flatten=False, ) ) if self.dropout > 0.0: self.mlp.add(mx.gluon.nn.Dropout(self.dropout)) self.mlp.add( mx.gluon.nn.Dense(units=out_dim, use_bias=True, flatten=False) ) # no activation def hybrid_forward(self, F, x: Tensor, *args) -> Tensor: r""" Position-wise feed-forward network with activation. Parameters ---------- x Tensor of shape (batch_size, d, in_dim) Returns ------- Tensor of shape (batch_size, d1, out_dim) """ return self.mlp(x) class TransformerProcessBlock(HybridBlock): r""" Block to perform pre/post processing on layer inputs. The processing steps are determined by the sequence argument, which can contain one of the three operations: n: layer normalization r: residual connection d: dropout """ def __init__(self, sequence: str, dropout: float, **kwargs) -> None: super().__init__(**kwargs) self.sequence = sequence self.dropout = dropout self.layer_norm = None if "n" in sequence: self.layer_norm = LayerNormalization() # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, data: Tensor, prev: Optional[Tensor] = None ) -> Tensor: r""" Apply processing sequence to data with optional previous input. Parameters ---------- data Input data of shape: (batch_size, length, num_hidden) prev Previous data of shape (batch_size, length, num_hidden) Returns ------- Processed data of shape (batch_size, length, num_hidden). """ if not self.sequence: return data if prev is None: assert ( "r" not in self.sequence ), "Residual connection not allowed if no previous value given." for step in self.sequence: if step == "r": data = F.broadcast_add(data, prev) elif step == "n": data = self.layer_norm(data) elif step == "d": if self.dropout > 0.0: data = F.Dropout(data, p=self.dropout) else: raise ValueError("Unknown step in sequence: %s" % step) return data
15,991
27.506239
112
py
rankpredictor
rankpredictor-master/sub/gluonts/model/transformer/trans_decoder.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import Dict, Optional # Third-party imports from mxnet.gluon import HybridBlock # First-party imports from gluonts.core.component import validated from gluonts.model.common import Tensor from gluonts.model.transformer.layers import ( TransformerProcessBlock, TransformerFeedForward, MultiHeadSelfAttention, MultiHeadAttention, InputLayer, ) class TransformerDecoder(HybridBlock): @validated() def __init__(self, decoder_length: int, config: Dict, **kwargs) -> None: super().__init__(**kwargs) self.decoder_length = decoder_length self.cache = {} with self.name_scope(): self.enc_input_layer = InputLayer(model_size=config["model_dim"]) self.dec_pre_self_att = TransformerProcessBlock( sequence=config["pre_seq"], dropout=config["dropout_rate"], prefix="pretransformerprocessblock_", ) self.dec_self_att = MultiHeadSelfAttention( att_dim_in=config["model_dim"], heads=config["num_heads"], att_dim_out=config["model_dim"], dropout=config["dropout_rate"], prefix="multiheadselfattention_", ) self.dec_post_self_att = TransformerProcessBlock( sequence=config["post_seq"], dropout=config["dropout_rate"], prefix="postselfatttransformerprocessblock_", ) self.dec_enc_att = MultiHeadAttention( att_dim_in=config["model_dim"], heads=config["num_heads"], att_dim_out=config["model_dim"], dropout=config["dropout_rate"], prefix="multiheadattention_", ) self.dec_post_att = TransformerProcessBlock( sequence=config["post_seq"], dropout=config["dropout_rate"], prefix="postatttransformerprocessblock_", ) self.dec_ff = TransformerFeedForward( inner_dim=config["model_dim"] * config["inner_ff_dim_scale"], out_dim=config["model_dim"], act_type=config["act_type"], dropout=config["dropout_rate"], prefix="transformerfeedforward_", ) self.dec_post_ff = TransformerProcessBlock( sequence=config["post_seq"], dropout=config["dropout_rate"], prefix="postffransformerprocessblock_", ) def cache_reset(self): self.cache = {} # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, data: Tensor, enc_out: Tensor, mask: Optional[Tensor] = None, is_train: bool = True, ) -> Tensor: """ A transformer encoder block consists of a self-attention and a feed-forward layer with pre/post process blocks in between. """ # embedding inputs = self.enc_input_layer(data) # self-attention data_att, cache = self.dec_self_att( self.dec_pre_self_att(inputs, None), mask, self.cache.copy() if not is_train else None, ) data = self.dec_post_self_att(data_att, inputs) # encoder attention data_att = self.dec_enc_att(data, enc_out) data = self.dec_post_att(data_att, data) # feed-forward data_ff = self.dec_ff(data) data = self.dec_post_ff(data_ff, data) if not is_train: self.cache = cache.copy() return data
4,259
32.543307
118
py
rankpredictor
rankpredictor-master/sub/gluonts/model/transformer/_network.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import Tuple, List, Optional # Third-party imports import mxnet as mx # First-party imports from gluonts.block.scaler import NOPScaler, MeanScaler from gluonts.block.feature import FeatureEmbedder from gluonts.core.component import validated from gluonts.distribution import DistributionOutput from gluonts.model.common import Tensor from gluonts.model.transformer.trans_encoder import TransformerEncoder from gluonts.model.transformer.trans_decoder import TransformerDecoder LARGE_NEGATIVE_VALUE = -99999999 def prod(xs): p = 1 for x in xs: p *= x return p class TransformerNetwork(mx.gluon.HybridBlock): @validated() def __init__( self, encoder: TransformerEncoder, decoder: TransformerDecoder, history_length: int, context_length: int, prediction_length: int, distr_output: DistributionOutput, cardinality: List[int], embedding_dimension: int, lags_seq: List[int], scaling: bool = True, **kwargs, ) -> None: super().__init__(**kwargs) self.history_length = history_length self.context_length = context_length self.prediction_length = prediction_length self.scaling = scaling self.cardinality = cardinality self.embedding_dimension = embedding_dimension self.distr_output = distr_output assert len(set(lags_seq)) == len( lags_seq ), "no duplicated lags allowed!" lags_seq.sort() self.lags_seq = lags_seq self.target_shape = distr_output.event_shape with self.name_scope(): self.proj_dist_args = distr_output.get_args_proj() self.encoder = encoder self.decoder = decoder self.embedder = FeatureEmbedder( cardinalities=cardinality, embedding_dims=[embedding_dimension for _ in cardinality], ) if scaling: self.scaler = MeanScaler(keepdims=True) else: self.scaler = NOPScaler(keepdims=True) @staticmethod def get_lagged_subsequences( F, sequence: Tensor, sequence_length: int, indices: List[int], subsequences_length: int = 1, ) -> Tensor: """ Returns lagged subsequences of a given sequence. Parameters ---------- sequence : Tensor the sequence from which lagged subsequences should be extracted. Shape: (N, T, C). sequence_length : int length of sequence in the T (time) dimension (axis = 1). indices : List[int] list of lag indices to be used. subsequences_length : int length of the subsequences to be extracted. Returns -------- lagged : Tensor a tensor of shape (N, S, C, I), where S = subsequences_length and I = len(indices), containing lagged subsequences. Specifically, lagged[i, j, :, k] = sequence[i, -indices[k]-S+j, :]. """ # we must have: sequence_length - lag_index - subsequences_length >= 0 # for all lag_index, hence the following assert assert max(indices) + subsequences_length <= sequence_length, ( f"lags cannot go further than history length, found lag {max(indices)} " f"while history length is only {sequence_length}" ) assert all(lag_index >= 0 for lag_index in indices) lagged_values = [] for lag_index in indices: begin_index = -lag_index - subsequences_length end_index = -lag_index if lag_index > 0 else None lagged_values.append( F.slice_axis( sequence, axis=1, begin=begin_index, end=end_index ) ) return F.stack(*lagged_values, axis=-1) def create_network_input( self, F, feat_static_cat: Tensor, # (batch_size, num_features) past_time_feat: Tensor, # (batch_size, num_features, history_length) past_target: Tensor, # (batch_size, history_length, 1) past_observed_values: Tensor, # (batch_size, history_length) future_time_feat: Optional[ Tensor ], # (batch_size, num_features, prediction_length) future_target: Optional[Tensor], # (batch_size, prediction_length) ) -> Tuple[Tensor, Tensor, Tensor]: """ Creates inputs for the transformer network. All tensor arguments should have NTC layout. """ if future_time_feat is None or future_target is None: time_feat = past_time_feat.slice_axis( axis=1, begin=self.history_length - self.context_length, end=None, ) sequence = past_target sequence_length = self.history_length subsequences_length = self.context_length else: time_feat = F.concat( past_time_feat.slice_axis( axis=1, begin=self.history_length - self.context_length, end=None, ), future_time_feat, dim=1, ) sequence = F.concat(past_target, future_target, dim=1) sequence_length = self.history_length + self.prediction_length subsequences_length = self.context_length + self.prediction_length # (batch_size, sub_seq_len, *target_shape, num_lags) lags = self.get_lagged_subsequences( F=F, sequence=sequence, sequence_length=sequence_length, indices=self.lags_seq, subsequences_length=subsequences_length, ) # scale is computed on the context length last units of the past target # scale shape is (batch_size, 1, *target_shape) _, scale = self.scaler( past_target.slice_axis( axis=1, begin=-self.context_length, end=None ), past_observed_values.slice_axis( axis=1, begin=-self.context_length, end=None ), ) embedded_cat = self.embedder(feat_static_cat) # in addition to embedding features, use the log scale as it can help prediction too # (batch_size, num_features + prod(target_shape)) static_feat = F.concat( embedded_cat, F.log(scale) if len(self.target_shape) == 0 else F.log(scale.squeeze(axis=1)), dim=1, ) repeated_static_feat = static_feat.expand_dims(axis=1).repeat( axis=1, repeats=subsequences_length ) # (batch_size, sub_seq_len, *target_shape, num_lags) lags_scaled = F.broadcast_div(lags, scale.expand_dims(axis=-1)) # from (batch_size, sub_seq_len, *target_shape, num_lags) # to (batch_size, sub_seq_len, prod(target_shape) * num_lags) input_lags = F.reshape( data=lags_scaled, shape=( -1, subsequences_length, len(self.lags_seq) * prod(self.target_shape), ), ) # (batch_size, sub_seq_len, input_dim) inputs = F.concat(input_lags, time_feat, repeated_static_feat, dim=-1) return inputs, scale, static_feat @staticmethod def upper_triangular_mask(F, d): mask = F.zeros_like(F.eye(d)) for k in range(d - 1): mask = mask + F.eye(d, d, k + 1) return mask * LARGE_NEGATIVE_VALUE def hybrid_forward(self, F, x, *args, **kwargs): raise NotImplementedError class TransformerTrainingNetwork(TransformerNetwork): # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, feat_static_cat: Tensor, past_time_feat: Tensor, past_target: Tensor, past_observed_values: Tensor, future_time_feat: Tensor, future_target: Tensor, ) -> Tensor: """ Computes the loss for training Transformer, all inputs tensors representing time series have NTC layout. Parameters ---------- F feat_static_cat : (batch_size, num_features) past_time_feat : (batch_size, history_length, num_features) past_target : (batch_size, history_length, *target_shape) past_observed_values : (batch_size, history_length, *target_shape, seq_len) future_time_feat : (batch_size, prediction_length, num_features) future_target : (batch_size, prediction_length, *target_shape) Returns ------- Loss with shape (batch_size, context + prediction_length, 1) """ # create the inputs for the encoder inputs, scale, _ = self.create_network_input( F=F, feat_static_cat=feat_static_cat, past_time_feat=past_time_feat, past_target=past_target, past_observed_values=past_observed_values, future_time_feat=future_time_feat, future_target=future_target, ) enc_input = F.slice_axis( inputs, axis=1, begin=0, end=self.context_length ) dec_input = F.slice_axis( inputs, axis=1, begin=self.context_length, end=None ) # pass through encoder enc_out = self.encoder(enc_input) # input to decoder dec_output = self.decoder( dec_input, enc_out, self.upper_triangular_mask(F, self.prediction_length), ) # compute loss distr_args = self.proj_dist_args(dec_output) distr = self.distr_output.distribution(distr_args, scale=scale) loss = distr.loss(future_target) return loss.mean() class TransformerPredictionNetwork(TransformerNetwork): @validated() def __init__(self, num_parallel_samples: int = 100, **kwargs) -> None: super().__init__(**kwargs) self.num_parallel_samples = num_parallel_samples # for decoding the lags are shifted by one, # at the first time-step of the decoder a lag of one corresponds to the last target value self.shifted_lags = [l - 1 for l in self.lags_seq] def sampling_decoder( self, F, static_feat: Tensor, past_target: Tensor, time_feat: Tensor, scale: Tensor, enc_out: Tensor, ) -> Tensor: """ Computes sample paths by unrolling the LSTM starting with a initial input and state. Parameters ---------- static_feat : Tensor static features. Shape: (batch_size, num_static_features). past_target : Tensor target history. Shape: (batch_size, history_length, 1). time_feat : Tensor time features. Shape: (batch_size, prediction_length, num_time_features). scale : Tensor tensor containing the scale of each element in the batch. Shape: (batch_size, ). enc_out: Tensor output of the encoder. Shape: (batch_size, num_cells) Returns -------- sample_paths : Tensor a tensor containing sampled paths. Shape: (batch_size, num_sample_paths, prediction_length). """ # blows-up the dimension of each tensor to batch_size * self.num_parallel_samples for increasing parallelism repeated_past_target = past_target.repeat( repeats=self.num_parallel_samples, axis=0 ) repeated_time_feat = time_feat.repeat( repeats=self.num_parallel_samples, axis=0 ) repeated_static_feat = static_feat.repeat( repeats=self.num_parallel_samples, axis=0 ).expand_dims(axis=1) repeated_enc_out = enc_out.repeat( repeats=self.num_parallel_samples, axis=0 ).expand_dims(axis=1) repeated_scale = scale.repeat( repeats=self.num_parallel_samples, axis=0 ) future_samples = [] # for each future time-units we draw new samples for this time-unit and update the state for k in range(self.prediction_length): lags = self.get_lagged_subsequences( F=F, sequence=repeated_past_target, sequence_length=self.history_length + k, indices=self.shifted_lags, subsequences_length=1, ) # (batch_size * num_samples, 1, *target_shape, num_lags) lags_scaled = F.broadcast_div( lags, repeated_scale.expand_dims(axis=-1) ) # from (batch_size * num_samples, 1, *target_shape, num_lags) # to (batch_size * num_samples, 1, prod(target_shape) * num_lags) input_lags = F.reshape( data=lags_scaled, shape=(-1, 1, prod(self.target_shape) * len(self.lags_seq)), ) # (batch_size * num_samples, 1, prod(target_shape) * num_lags + num_time_features + num_static_features) dec_input = F.concat( input_lags, repeated_time_feat.slice_axis(axis=1, begin=k, end=k + 1), repeated_static_feat, dim=-1, ) dec_output = self.decoder(dec_input, repeated_enc_out, None, False) distr_args = self.proj_dist_args(dec_output) # compute likelihood of target given the predicted parameters distr = self.distr_output.distribution( distr_args, scale=repeated_scale ) # (batch_size * num_samples, 1, *target_shape) new_samples = distr.sample() # (batch_size * num_samples, seq_len, *target_shape) repeated_past_target = F.concat( repeated_past_target, new_samples, dim=1 ) future_samples.append(new_samples) # reset cache of the decoder self.decoder.cache_reset() # (batch_size * num_samples, prediction_length, *target_shape) samples = F.concat(*future_samples, dim=1) # (batch_size, num_samples, *target_shape, prediction_length) return samples.reshape( shape=( (-1, self.num_parallel_samples) + self.target_shape + (self.prediction_length,) ) ) # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, feat_static_cat: Tensor, past_time_feat: Tensor, past_target: Tensor, past_observed_values: Tensor, future_time_feat: Tensor, ) -> Tensor: """ Predicts samples, all tensors should have NTC layout. Parameters ---------- F feat_static_cat : (batch_size, num_features) past_time_feat : (batch_size, history_length, num_features) past_target : (batch_size, history_length, *target_shape) past_observed_values : (batch_size, history_length, *target_shape) future_time_feat : (batch_size, prediction_length, num_features) Returns predicted samples ------- """ # create the inputs for the encoder inputs, scale, static_feat = self.create_network_input( F=F, feat_static_cat=feat_static_cat, past_time_feat=past_time_feat, past_target=past_target, past_observed_values=past_observed_values, future_time_feat=None, future_target=None, ) # pass through encoder enc_out = self.encoder(inputs) return self.sampling_decoder( F=F, past_target=past_target, time_feat=future_time_feat, static_feat=static_feat, scale=scale, enc_out=enc_out, )
16,590
33.564583
116
py
rankpredictor
rankpredictor-master/sub/gluonts/model/transformer/_estimator.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import List, Optional # Third-party imports from mxnet.gluon import HybridBlock # First-party imports from gluonts.core.component import validated from gluonts.dataset.field_names import FieldName from gluonts.distribution import StudentTOutput, DistributionOutput from gluonts.model.estimator import GluonEstimator from gluonts.model.predictor import Predictor, RepresentableBlockPredictor from gluonts.support.util import copy_parameters from gluonts.time_feature import ( TimeFeature, get_lags_for_frequency, time_features_from_frequency_str, ) from gluonts.trainer import Trainer from gluonts.transform import ( AddAgeFeature, AddObservedValuesIndicator, AddTimeFeatures, AsNumpyArray, Chain, ExpectedNumInstanceSampler, InstanceSplitter, RemoveFields, SetField, Transformation, VstackFeatures, ) # Relative imports from gluonts.model.transformer._network import ( TransformerPredictionNetwork, TransformerTrainingNetwork, ) from gluonts.model.transformer.trans_encoder import TransformerEncoder from gluonts.model.transformer.trans_decoder import TransformerDecoder class TransformerEstimator(GluonEstimator): """ Construct a Transformer estimator. This implements a Transformer model, close to the one described in [Vaswani2017]_. .. [Vaswani2017] Vaswani, Ashish, et al. "Attention is all you need." Advances in neural information processing systems. 2017. Parameters ---------- freq Frequency of the data to train on and predict prediction_length Length of the prediction horizon context_length Number of steps to unroll the RNN for before computing predictions (default: None, in which case context_length = prediction_length) trainer Trainer object to be used (default: Trainer()) dropout_rate Dropout regularization parameter (default: 0.1) cardinality Number of values of the each categorical feature (default: [1]) embedding_dimension Dimension of the embeddings for categorical features (the same dimension is used for all embeddings, default: 5) distr_output Distribution to use to evaluate observations and sample predictions (default: StudentTOutput()) model_dim Dimension of the transformer network, i.e., embedding dimension of the input (default: 32) inner_ff_dim_scale Dimension scale of the inner hidden layer of the transformer's feedforward network (default: 4) pre_seq Sequence that defined operations of the processing block before the main transformer network. Available operations: 'd' for dropout, 'r' for residual connections and 'n' for normalization (default: 'dn') post_seq seq Sequence that defined operations of the processing block in and after the main transformer network. Available operations: 'd' for dropout, 'r' for residual connections and 'n' for normalization (default: 'drn'). act_type Activation type of the transformer network (default: 'softrelu') num_heads Number of heads in the multi-head attention (default: 8) scaling Whether to automatically scale the target values (default: true) lags_seq Indices of the lagged target values to use as inputs of the RNN (default: None, in which case these are automatically determined based on freq) time_features Time features to use as inputs of the RNN (default: None, in which case these are automatically determined based on freq) num_parallel_samples Number of evaluation samples per time series to increase parallelism during inference. This is a model optimization that does not affect the accuracy (default: 100) """ @validated() def __init__( self, freq: str, prediction_length: int, context_length: Optional[int] = None, trainer: Trainer = Trainer(), dropout_rate: float = 0.1, cardinality: Optional[List[int]] = None, embedding_dimension: int = 20, distr_output: DistributionOutput = StudentTOutput(), model_dim: int = 32, inner_ff_dim_scale: int = 4, pre_seq: str = "dn", post_seq: str = "drn", act_type: str = "softrelu", num_heads: int = 8, scaling: bool = True, lags_seq: Optional[List[int]] = None, time_features: Optional[List[TimeFeature]] = None, use_feat_dynamic_real: bool = False, use_feat_static_cat: bool = False, num_parallel_samples: int = 100, ) -> None: super().__init__(trainer=trainer) assert ( prediction_length > 0 ), "The value of `prediction_length` should be > 0" assert ( context_length is None or context_length > 0 ), "The value of `context_length` should be > 0" assert dropout_rate >= 0, "The value of `dropout_rate` should be >= 0" assert ( cardinality is not None or not use_feat_static_cat ), "You must set `cardinality` if `use_feat_static_cat=True`" assert cardinality is None or all( [c > 0 for c in cardinality] ), "Elements of `cardinality` should be > 0" assert ( embedding_dimension > 0 ), "The value of `embedding_dimension` should be > 0" assert ( num_parallel_samples > 0 ), "The value of `num_parallel_samples` should be > 0" self.freq = freq self.prediction_length = prediction_length self.context_length = ( context_length if context_length is not None else prediction_length ) self.distr_output = distr_output self.dropout_rate = dropout_rate self.use_feat_dynamic_real = use_feat_dynamic_real self.use_feat_static_cat = use_feat_static_cat self.cardinality = cardinality if use_feat_static_cat else [1] self.embedding_dimension = embedding_dimension self.num_parallel_samples = num_parallel_samples self.lags_seq = ( lags_seq if lags_seq is not None else get_lags_for_frequency(freq_str=freq) ) self.time_features = ( time_features if time_features is not None else time_features_from_frequency_str(self.freq) ) self.history_length = self.context_length + max(self.lags_seq) self.scaling = scaling self.config = { "model_dim": model_dim, "pre_seq": pre_seq, "post_seq": post_seq, "dropout_rate": dropout_rate, "inner_ff_dim_scale": inner_ff_dim_scale, "act_type": act_type, "num_heads": num_heads, } self.encoder = TransformerEncoder( self.context_length, self.config, prefix="enc_" ) self.decoder = TransformerDecoder( self.prediction_length, self.config, prefix="dec_" ) def create_transformation(self) -> Transformation: remove_field_names = [ FieldName.FEAT_DYNAMIC_CAT, FieldName.FEAT_STATIC_REAL, ] if not self.use_feat_dynamic_real: remove_field_names.append(FieldName.FEAT_DYNAMIC_REAL) return Chain( [RemoveFields(field_names=remove_field_names)] + ( [SetField(output_field=FieldName.FEAT_STATIC_CAT, value=[0.0])] if not self.use_feat_static_cat else [] ) + [ AsNumpyArray(field=FieldName.FEAT_STATIC_CAT, expected_ndim=1), AsNumpyArray( field=FieldName.TARGET, # in the following line, we add 1 for the time dimension expected_ndim=1 + len(self.distr_output.event_shape), ), AddObservedValuesIndicator( target_field=FieldName.TARGET, output_field=FieldName.OBSERVED_VALUES, ), AddTimeFeatures( start_field=FieldName.START, target_field=FieldName.TARGET, output_field=FieldName.FEAT_TIME, time_features=self.time_features, pred_length=self.prediction_length, ), AddAgeFeature( target_field=FieldName.TARGET, output_field=FieldName.FEAT_AGE, pred_length=self.prediction_length, log_scale=True, ), VstackFeatures( output_field=FieldName.FEAT_TIME, input_fields=[FieldName.FEAT_TIME, FieldName.FEAT_AGE] + ( [FieldName.FEAT_DYNAMIC_REAL] if self.use_feat_dynamic_real else [] ), ), InstanceSplitter( target_field=FieldName.TARGET, is_pad_field=FieldName.IS_PAD, start_field=FieldName.START, forecast_start_field=FieldName.FORECAST_START, train_sampler=ExpectedNumInstanceSampler(num_instances=1), past_length=self.history_length, future_length=self.prediction_length, time_series_fields=[ FieldName.FEAT_TIME, FieldName.OBSERVED_VALUES, ], ), ] ) def create_training_network(self) -> TransformerTrainingNetwork: training_network = TransformerTrainingNetwork( encoder=self.encoder, decoder=self.decoder, history_length=self.history_length, context_length=self.context_length, prediction_length=self.prediction_length, distr_output=self.distr_output, cardinality=self.cardinality, embedding_dimension=self.embedding_dimension, lags_seq=self.lags_seq, scaling=True, ) return training_network def create_predictor( self, transformation: Transformation, trained_network: HybridBlock ) -> Predictor: prediction_network = TransformerPredictionNetwork( encoder=self.encoder, decoder=self.decoder, history_length=self.history_length, context_length=self.context_length, prediction_length=self.prediction_length, distr_output=self.distr_output, cardinality=self.cardinality, embedding_dimension=self.embedding_dimension, lags_seq=self.lags_seq, scaling=True, num_parallel_samples=self.num_parallel_samples, ) copy_parameters(trained_network, prediction_network) return RepresentableBlockPredictor( input_transform=transformation, prediction_net=prediction_network, batch_size=self.trainer.batch_size, freq=self.freq, prediction_length=self.prediction_length, ctx=self.trainer.ctx, )
12,230
37.583596
100
py
rankpredictor
rankpredictor-master/sub/gluonts/model/gp_forecaster/_network.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import Tuple # Third-party imports import mxnet as mx # First-party imports from gluonts.core.component import DType, validated from gluonts.distribution.distribution import softplus from gluonts.gp import GaussianProcess from gluonts.kernels import KernelOutputDict from gluonts.model.common import Tensor class GaussianProcessNetworkBase(mx.gluon.HybridBlock): """ Defines a Gluon block used for GP training and predictions. """ # The two subclasses GaussianProcessTrainingNetwork and # GaussianProcessPredictionNetwork define how to # compute the loss and how to generate predictions, respectively. @validated() def __init__( self, prediction_length: int, context_length: int, cardinality: int, kernel_output: KernelOutputDict, params_scaling: bool, ctx: mx.Context, float_type: DType, max_iter_jitter: int, jitter_method: str, **kwargs, ) -> None: """ Parameters ---------- prediction_length Prediction length. context_length Training length. cardinality Number of time series. kernel_output KernelOutput instance to determine which kernel subclass to be instantiated. params_scaling Determines whether or not to scale the model parameters. ctx Determines whether to compute on the cpu or gpu. float_type Determines whether to use single or double precision. max_iter_jitter Maximum number of iterations for jitter to iteratively make the matrix positive definite. jitter_method Iteratively jitter method or use eigenvalue decomposition depending on problem size. **kwargs Arbitrary keyword arguments. """ super().__init__(**kwargs) self.prediction_length = prediction_length self.context_length = context_length self.cardinality = cardinality self.kernel_output = kernel_output self.params_scaling = params_scaling self.float_type = float_type self.ctx = ctx self.max_iter_jitter = max_iter_jitter self.jitter_method = jitter_method with self.name_scope(): self.proj_kernel_args = kernel_output.get_args_proj( self.float_type ) self.num_hyperparams = kernel_output.get_num_args() self.embedding = mx.gluon.nn.Embedding( # Noise sigma is additional parameter so add 1 to output dim input_dim=self.cardinality, output_dim=self.num_hyperparams + 1, dtype=self.float_type, ) # noinspection PyMethodOverriding,PyPep8Naming def get_gp_params( self, F, past_target: Tensor, past_time_feat: Tensor, feat_static_cat: Tensor, ) -> Tuple: """ This function returns the GP hyper-parameters for the model. Parameters ---------- F A module that can either refer to the Symbol API or the NDArray API in MXNet. past_target Training time series values of shape (batch_size, context_length). past_time_feat Training features of shape (batch_size, context_length, num_features). feat_static_cat Time series indices of shape (batch_size, 1). Returns ------- Tuple Tuple of kernel hyper-parameters of length num_hyperparams. Each is a Tensor of shape (batch_size, 1, 1). Model noise sigma. Tensor of shape (batch_size, 1, 1). """ output = self.embedding( feat_static_cat.squeeze() ) # Shape (batch_size, num_hyperparams + 1) kernel_args = self.proj_kernel_args(output) sigma = softplus( F, output.slice_axis( # sigma is the last hyper-parameter axis=1, begin=self.num_hyperparams, end=self.num_hyperparams + 1, ), ) if self.params_scaling: scalings = self.kernel_output.gp_params_scaling( F, past_target, past_time_feat ) sigma = F.broadcast_mul(sigma, scalings[self.num_hyperparams]) kernel_args = ( F.broadcast_mul(kernel_arg, scaling) for kernel_arg, scaling in zip( kernel_args, scalings[0 : self.num_hyperparams] ) ) min_value = 1e-5 max_value = 1e8 kernel_args = ( kernel_arg.clip(min_value, max_value).expand_dims(axis=2) for kernel_arg in kernel_args ) sigma = sigma.clip(min_value, max_value).expand_dims(axis=2) return kernel_args, sigma class GaussianProcessTrainingNetwork(GaussianProcessNetworkBase): # noinspection PyMethodOverriding,PyPep8Naming @validated() def __init__(self, *args, **kwargs) -> None: """ Parameters ---------- *args Variable length argument list. **kwargs Arbitrary keyword arguments. """ super().__init__(*args, **kwargs) # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, past_target: Tensor, past_time_feat: Tensor, feat_static_cat: Tensor, ) -> Tensor: """ Parameters ---------- F A module that can either refer to the Symbol API or the NDArray API in MXNet. past_target Training time series values of shape (batch_size, context_length). past_time_feat Training features of shape (batch_size, context_length, num_features). feat_static_cat Time series indices of shape (batch_size, 1). Returns ------- Tensor GP loss of shape (batch_size, 1) """ kernel_args, sigma = self.get_gp_params( F, past_target, past_time_feat, feat_static_cat ) kernel = self.kernel_output.kernel(kernel_args) gp = GaussianProcess( sigma=sigma, kernel=kernel, context_length=self.context_length, ctx=self.ctx, float_type=self.float_type, max_iter_jitter=self.max_iter_jitter, jitter_method=self.jitter_method, ) return gp.log_prob(past_time_feat, past_target) class GaussianProcessPredictionNetwork(GaussianProcessNetworkBase): @validated() def __init__( self, num_parallel_samples: int, sample_noise: bool, *args, **kwargs ) -> None: r""" Parameters ---------- num_parallel_samples Number of samples to be drawn. sample_noise Boolean to determine whether to add :math:`\sigma^2I` to the predictive covariance matrix. *args Variable length argument list. **kwargs Arbitrary keyword arguments. """ super().__init__(*args, **kwargs) self.num_parallel_samples = num_parallel_samples self.sample_noise = sample_noise # noinspection PyMethodOverriding,PyPep8Naming def hybrid_forward( self, F, past_target: Tensor, past_time_feat: Tensor, future_time_feat: Tensor, feat_static_cat: Tensor, ) -> Tensor: """ Parameters ---------- F A module that can either refer to the Symbol API or the NDArray API in MXNet. past_target Training time series values of shape (batch_size, context_length). past_time_feat Training features of shape (batch_size, context_length, num_features). future_time_feat Test features of shape (batch_size, prediction_length, num_features). feat_static_cat Time series indices of shape (batch_size, 1). Returns ------- Tensor GP samples of shape (batch_size, num_samples, prediction_length). """ kernel_args, sigma = self.get_gp_params( F, past_target, past_time_feat, feat_static_cat ) gp = GaussianProcess( sigma=sigma, kernel=self.kernel_output.kernel(kernel_args), context_length=self.context_length, prediction_length=self.prediction_length, num_samples=self.num_parallel_samples, ctx=self.ctx, float_type=self.float_type, max_iter_jitter=self.max_iter_jitter, jitter_method=self.jitter_method, sample_noise=self.sample_noise, ) samples, _, _ = gp.exact_inference( past_time_feat, past_target, future_time_feat ) # Shape (batch_size, prediction_length, num_samples) return samples.swapaxes(1, 2)
9,726
33.01049
102
py
rankpredictor
rankpredictor-master/sub/gluonts/model/gp_forecaster/_estimator.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import List, Optional # Third-party imports import numpy as np from mxnet.gluon import HybridBlock # First-party imports from gluonts.core.component import DType, validated from gluonts.dataset.field_names import FieldName from gluonts.kernels import KernelOutput, RBFKernelOutput from gluonts.model.estimator import GluonEstimator from gluonts.model.predictor import Predictor, RepresentableBlockPredictor from gluonts.support.util import copy_parameters from gluonts.time_feature import TimeFeature, time_features_from_frequency_str from gluonts.trainer import Trainer from gluonts.transform import ( AddTimeFeatures, AsNumpyArray, CanonicalInstanceSplitter, Chain, SetFieldIfNotPresent, TestSplitSampler, Transformation, ) # Relative imports from ._network import ( GaussianProcessPredictionNetwork, GaussianProcessTrainingNetwork, ) class GaussianProcessEstimator(GluonEstimator): r""" GaussianProcessEstimator shows how to build a local time series model using Gaussian Processes (GP). Each time series has a GP with its own hyper-parameters. For the radial basis function (RBF) Kernel, the learnable hyper-parameters are the amplitude and lengthscale. The periodic kernel has those hyper-parameters with an additional learnable frequency parameter. The RBFKernel is the default, but either kernel can be used by inputting the desired KernelOutput object. The noise sigma in the model is another learnable hyper-parameter for both kernels. These parameters are fit using an Embedding of the integer time series indices (each time series has its set of hyper-parameter that is static in time). The observations are the time series values. In this model, the time features are hour of the day and day of the week. Parameters ---------- freq Time series frequency. prediction_length Prediction length. cardinality Number of time series. trainer Trainer instance to be used for model training (default: Trainer()). context_length Training length (default: None, in which case context_length = prediction_length). kernel_output KernelOutput instance to determine which kernel subclass to be instantiated (default: RBFKernelOutput()). params_scaling Determines whether or not to scale the model parameters (default: True). float_type Determines whether to use single or double precision (default: np.float64). max_iter_jitter Maximum number of iterations for jitter to iteratively make the matrix positive definite (default: 10). jitter_method Iteratively jitter method or use eigenvalue decomposition depending on problem size (default: "iter"). sample_noise Boolean to determine whether to add :math:`\sigma^2I` to the predictive covariance matrix (default: True). time_features Time features to use as inputs of the model (default: None, in which case these are automatically determined based on the frequency). num_parallel_samples Number of evaluation samples per time series to increase parallelism during inference. This is a model optimization that does not affect the accuracy (default: 100). """ @validated() def __init__( self, freq: str, prediction_length: int, cardinality: int, trainer: Trainer = Trainer(), context_length: Optional[int] = None, kernel_output: KernelOutput = RBFKernelOutput(), params_scaling: bool = True, dtype: DType = np.float64, max_iter_jitter: int = 10, jitter_method: str = "iter", sample_noise: bool = True, time_features: Optional[List[TimeFeature]] = None, num_parallel_samples: int = 100, ) -> None: self.float_type = dtype super().__init__(trainer=trainer, dtype=self.float_type) assert ( prediction_length > 0 ), "The value of `prediction_length` should be > 0" assert cardinality > 0, "The value of `cardinality` should be > 0" assert ( context_length is None or context_length > 0 ), "The value of `context_length` should be > 0" assert ( num_parallel_samples > 0 ), "The value of `num_parallel_samples` should be > 0" self.freq = freq self.prediction_length = prediction_length self.context_length = ( context_length if context_length is not None else prediction_length ) self.cardinality = cardinality self.kernel_output = kernel_output self.params_scaling = params_scaling self.max_iter_jitter = max_iter_jitter self.jitter_method = jitter_method self.sample_noise = sample_noise self.time_features = ( time_features if time_features is not None else time_features_from_frequency_str(self.freq) ) self.num_parallel_samples = num_parallel_samples def create_transformation(self) -> Transformation: return Chain( [ AsNumpyArray(field=FieldName.TARGET, expected_ndim=1), AddTimeFeatures( start_field=FieldName.START, target_field=FieldName.TARGET, output_field=FieldName.FEAT_TIME, time_features=self.time_features, pred_length=self.prediction_length, ), SetFieldIfNotPresent( field=FieldName.FEAT_STATIC_CAT, value=[0.0] ), AsNumpyArray(field=FieldName.FEAT_STATIC_CAT, expected_ndim=1), CanonicalInstanceSplitter( target_field=FieldName.TARGET, is_pad_field=FieldName.IS_PAD, start_field=FieldName.START, forecast_start_field=FieldName.FORECAST_START, instance_sampler=TestSplitSampler(), time_series_fields=[FieldName.FEAT_TIME], instance_length=self.context_length, use_prediction_features=True, prediction_length=self.prediction_length, ), ] ) def create_training_network(self) -> HybridBlock: return GaussianProcessTrainingNetwork( prediction_length=self.prediction_length, context_length=self.context_length, cardinality=self.cardinality, kernel_output=self.kernel_output, params_scaling=self.params_scaling, ctx=self.trainer.ctx, float_type=self.float_type, max_iter_jitter=self.max_iter_jitter, jitter_method=self.jitter_method, ) def create_predictor( self, transformation: Transformation, trained_network: HybridBlock ) -> Predictor: prediction_network = GaussianProcessPredictionNetwork( prediction_length=self.prediction_length, context_length=self.context_length, cardinality=self.cardinality, num_parallel_samples=self.num_parallel_samples, params=trained_network.collect_params(), kernel_output=self.kernel_output, params_scaling=self.params_scaling, ctx=self.trainer.ctx, float_type=self.float_type, max_iter_jitter=self.max_iter_jitter, jitter_method=self.jitter_method, sample_noise=self.sample_noise, ) copy_parameters( net_source=trained_network, net_dest=prediction_network ) return RepresentableBlockPredictor( input_transform=transformation, prediction_net=prediction_network, batch_size=self.trainer.batch_size, freq=self.freq, prediction_length=self.prediction_length, ctx=self.trainer.ctx, dtype=self.float_type, )
8,691
38.689498
114
py
rankpredictor
rankpredictor-master/sub/gluonts/kernels/_kernel_output.py
# Copyright 2018 Amazon.com, Inc. or its affiliates. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"). # You may not use this file except in compliance with the License. # A copy of the License is located at # # http://www.apache.org/licenses/LICENSE-2.0 # # or in the "license" file accompanying this file. This file 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. # Standard library imports from typing import Dict, Tuple import numpy as np from mxnet import gluon # First-party imports from gluonts.core.component import DType, validated from gluonts.distribution.distribution_output import ArgProj from gluonts.model.common import Tensor # Relative imports from . import Kernel class KernelOutput: """ Class to connect a network to a kernel. """ def get_args_proj(self, float_type: DType) -> gluon.HybridBlock: raise NotImplementedError() def kernel(self, args) -> Kernel: raise NotImplementedError() # noinspection PyMethodOverriding,PyPep8Naming @staticmethod def compute_std(F, data: Tensor, axis: int) -> Tensor: """ This function computes the standard deviation of the data along a given axis. Parameters ---------- F : ModuleType A module that can either refer to the Symbol API or the NDArray API in MXNet. data : Tensor Data to be used to compute the standard deviation. axis : int Axis along which to compute the standard deviation. Returns ------- Tensor The standard deviation of the given data. """ return F.sqrt( F.mean( F.broadcast_minus( data, F.mean(data, axis=axis).expand_dims(axis=axis) ) ** 2, axis=axis, ) ) class KernelOutputDict(KernelOutput): args_dim: Dict[str, int] kernel_cls: type @validated() def __init__(self) -> None: pass def get_num_args(self) -> int: return len(self.args_dim) def get_args_proj(self, float_type: DType = np.float32) -> ArgProj: """ This method calls the ArgProj block in distribution_output to project from a dense layer to kernel arguments. Parameters ---------- float_type : DType Determines whether to use single or double precision. Returns ------- ArgProj """ return ArgProj( args_dim=self.args_dim, domain_map=gluon.nn.HybridLambda(self.domain_map), dtype=float_type, ) # noinspection PyMethodOverriding,PyPep8Naming def gp_params_scaling( self, F, past_target: Tensor, past_time_feat: Tensor ) -> Tuple[Tensor, Tensor, Tensor]: raise NotImplementedError() # noinspection PyMethodOverriding,PyPep8Naming def domain_map(self, F, *args: Tensor): raise NotImplementedError() def kernel(self, kernel_args) -> Kernel: """ Parameters ---------- kernel_args Variable length argument list. Returns ------- Kernel Instantiated specified Kernel subclass object. """ return self.kernel_cls(*kernel_args)
3,541
27.111111
79
py
malware-uncertainty
malware-uncertainty-master/tools/utils.py
from __future__ import absolute_import from __future__ import division from __future__ import print_function import tensorflow as tf import numpy as np import tensorflow_probability as tfp import os import sys import shutil import hashlib import warnings import functools from collections import namedtuple, defaultdict class ParamWrapper(object): def __init__(self, params): if not isinstance(params, dict): params = vars(params) self.params = params def __getattr__(self, name): val = self.params.get(name) if val is None: MSG = "Setting params ({}) is deprecated" warnings.warn(MSG.format(name)) return val def retrive_files_set(base_dir, dir_ext, file_ext): """ get file paths given the directory :param base_dir: basic directory :param dir_ext: directory append at the rear of base_dir :param file_ext: file extension :return: set of file paths. Avoid the repetition """ def get_file_name(root_dir, file_ext): for dir_path, dir_names, file_names in os.walk(root_dir, topdown=True): for file_name in file_names: _ext = file_ext if os.path.splitext(file_name)[1] == _ext: yield os.path.join(dir_path, file_name) elif '.' not in file_ext: _ext = '.' + _ext if os.path.splitext(file_name)[1] == _ext: yield os.path.join(dir_path, file_name) else: pass else: pass if file_ext is not None: file_exts = file_ext.split("|") else: file_exts = [''] file_path_list = list() for ext in file_exts: file_path_list.extend(get_file_name(os.path.join(base_dir, dir_ext), ext)) # remove duplicate elements from collections import OrderedDict return list(OrderedDict.fromkeys(file_path_list)) def get_file_name(path): return os.path.splitext(os.path.basename(path))[0] def get_file_nameext(path): return os.path.basename(path) def dump_pickle(data, path): try: import pickle as pkl except Exception as e: import cPickle as pkl if not os.path.exists(os.path.dirname(path)): mkdir(os.path.dirname(path)) with open(path, 'wb') as wr: pkl.dump(data, wr) return True def read_pickle(path): try: import pickle as pkl except Exception as e: import cPickle as pkl if os.path.isfile(path): with open(path, 'rb') as fr: return pkl.load(fr) else: raise IOError("The {0} is not been found.".format(path)) def dump_joblib(data, path): if not os.path.exists(os.path.dirname(path)): mkdir(os.path.dirname(path)) try: import joblib with open(path, 'wb') as wr: joblib.dump(data, wr) return except IOError: raise IOError("Dump data failed.") def read_joblib(path): import joblib if os.path.isfile(path): with open(path, 'rb') as fr: return joblib.load(fr) else: raise IOError("The {0} is not a file.".format(path)) def read_txt(path, mode='r'): if os.path.isfile(path): with open(path, mode) as f_r: lines = f_r.read().strip().splitlines() return lines else: raise ValueError("{} does not seen like a file path.\n".format(path)) def dump_txt(data_str, path, mode='w'): if not isinstance(data_str, str): raise TypeError with open(path, mode) as f_w: f_w.write(data_str) def readdata_np(data_path): try: with open(data_path, 'rb') as f_r: data = np.load(f_r) return data except IOError as e: raise IOError("Unable to open {0}: {1}.\n".format(data_path, str(e))) def dumpdata_np(data, data_path): if not isinstance(data, np.ndarray): warnings.warn("The array is not the numpy.ndarray type.") data_dir = os.path.dirname(data_path) try: if not os.path.exists(data_dir): os.makedirs(data_dir) with open(data_path, 'wb') as f_s: np.save(f_s, data) except OSError as e: sys.stderr.write(e) def safe_load_json(json_path): try: import yaml with open(json_path, 'r') as rh: return yaml.safe_load(rh) except IOError as ex: raise IOError(str(ex) + ": Unable to load json file.") def load_json(json_path): try: import json with open(json_path, 'r') as rh: return json.load(rh) except IOError as ex: raise IOError(str(ex) + ": Unable to load json file.") def dump_json(obj_dict, file_path): try: import json if not os.path.exists(os.path.dirname(file_path)): mkdir(os.path.dirname(file_path)) with open(file_path, 'w') as fh: json.dump(obj_dict, fh) except IOError as ex: raise IOError(str(ex) + ": Fail to dump dict using json toolbox") def mkdir(target): try: if os.path.isfile(target): target = os.path.dirname(target) if not os.path.exists(target): os.makedirs(target) return 0 except IOError as e: raise Exception("Fail to create directory! Error:" + str(e)) def copy_files(src_file_list, dst_dir): if not isinstance(src_file_list, list): raise TypeError if os.path.isdir(dst_dir): raise ValueError for src in src_file_list: if not os.path.isfile(src): continue shutil.copy(src, dst_dir) def get_sha256(file_path): assert os.path.isfile(file_path), 'permit only file path' fh = open(file_path, 'rb') sha256 = hashlib.sha256() while True: data = fh.read(8192) if not data: break sha256.update(data) fh.close() return sha256.hexdigest() def merge_namedtuples(tp1, tp2): from collections import namedtuple _TP12 = namedtuple('tp12', tp1._fields + tp2._fields) return _TP12(*(tp1 + tp2)) def expformat(f, pos, prec=0, exp_digits=1, sign='off'): """Scientific-format a number with a given number of digits in the exponent. Optionally remove the sign in the exponent""" s = "%.*e" % (prec, f) mantissa, exp = s.split('e') if sign == 'on': # add 1 to digits as 1 is taken by sign +/- return "%se%+0*d" % (mantissa, exp_digits+1, int(exp)) else : return "%se%0*d" % (mantissa, exp_digits, int(exp)) def bootstrap(data, fun, n_resamples=1000, alpha=0.05, seed=0): """Compute confidence interval for values of function fun Parameters ========== data: list of arguments to fun """ assert isinstance(data, list) n_samples = len(data[0]) np.random.seed(seed) idx = np.random.randint(0, n_samples, (n_resamples, n_samples)) def select(sample): return [d[sample] for d in data] def evaluate(sample): result = select(sample) values = [] for elems in zip(*result): values.append(fun(*elems)) return np.stack(values, axis=0) values = evaluate(idx) idx = idx[np.argsort(values, axis=0, kind='mergesort')] values = np.sort(values, axis=0, kind='mergesort') stat = namedtuple('stat', ['value', 'index']) low = stat(value=values[int((alpha / 2.0) * n_resamples)], index=idx[int((alpha / 2.0) * n_resamples)]) high = stat(value=values[int((1 - alpha / 2.0) * n_resamples)], index=idx[int((1 - alpha / 2.0) * n_resamples)]) mean = stat(value=np.mean(values, axis=0), index=None) return low, high, mean ######################################################################################## ############################# functions for tf models ################################## ######################################################################################## ensemble_method_scope = ['vanilla', 'mc_dropout', 'deep_ensemble', 'weighted_ensemble', 'bayesian'] class DenseDropout(tf.keras.layers.Layer): def __init__(self, units, dropout_rate, activation=None, use_dropout=True, **kwargs): """ Initialize a dense-dropout layer :param units: number of neurons :param dropout_rate: a float value between 0 and 1. A portion of activations will be dropped randomly :param activation: activation function param use_dropout: performing dropout in both training and testing phases :param kwargs: other arguments for tf.keras.layers.Dense """ super(DenseDropout, self).__init__() self.units = units self.activation = activation self.dropout_rate = dropout_rate self.use_dropout = use_dropout self.kwargs = kwargs self.dense_layer = tf.keras.layers.Dense(units, activation=self.activation, **self.kwargs) self.dropout_layer = tf.keras.layers.Dropout(rate=dropout_rate) def call(self, inputs, training=True): return self.dropout_layer(self.dense_layer(inputs), training=self.use_dropout) class Conv2DDropout(tf.keras.layers.Layer): def __init__(self, filters, kernel_size, dropout_rate, activation=None, use_dropout=True, **kwargs): """ Initialize a convolution-dropout layer :param filters: Positive integer, number of ouput channels :param kernel_size: An integer or tuple/list of 2 integers, specifying the height and width of 2D convolution window :param dropout_rate: a float value between 0 and 1. A portion of activations will be dropped randomly :param activation: activation function :param use_dropout: performing dropout in both training and testing phases :param kwargs: other arguments for tf.keras.layers.Conv2D """ super(Conv2DDropout, self).__init__() self.filters = filters self.kernel_size = kernel_size self.dropout_rate = dropout_rate self.use_dropout = use_dropout self.conv2d_layer = tf.keras.layers.Conv2D(filters, kernel_size, activation=activation, **kwargs) self.dropout_layer = tf.keras.layers.Dropout(rate=dropout_rate) def call(self, inputs, training=True): return self.dropout_layer(self.conv2d_layer(inputs), training=self.use_dropout) class LSTMDropout(tf.keras.layers.Layer): def __init__(self, units, dropout_rate, use_dropout=True, go_backwards=True, return_sequences=True, **kwargs): """ Initialize a LSTM-dropout layer :param dropout_rate: a float value between 0 and 1. A portion of activations will be dropped randomly :param units: Positive Integer, number of neurons :param use_dropout: performing dropout in both training and testing phases :param kwargs: other arguments for tf.keras.layers.LSTM """ super(LSTMDropout, self).__init__() self.units = units self.dropout_rate = dropout_rate self.use_dropout = use_dropout self.go_backwards = go_backwards self.return_sequences = return_sequences self.lstm = tf.keras.layers.LSTM(units, dropout=self.dropout_rate, return_sequences=self.return_sequences, **kwargs) def call(self, inputs, training=True): return self.lstm(inputs, training=self.use_dropout) @property def return_state(self): return self.lstm.return_state def get_config(self): config = super(LSTMDropout, self).get_config() config['dropout_rate'] = self.dropout_rate config['units'] = self.units config['use_dropout'] = self.use_dropout config['go_backwards'] = self.go_backwards return config class DropoutDense(tf.keras.layers.Layer): def __init__(self, units, dropout_rate, activation=None, use_dropout=True, **kwargs): """ Initialize a dense-dropout layer :param units: number of neurons :param dropout_rate: a float value between 0 and 1. A portion of activations will be dropped randomly :param activation: activation function param use_dropout: performing dropout in both training and testing phases :param kwargs: other arguments for tf.keras.layers.Dense """ super(DropoutDense, self).__init__() self.units = units self.activation = activation self.dropout_rate = dropout_rate self.use_dropout = use_dropout self.kwargs = kwargs self.dense_layer = tf.keras.layers.Dense(units, activation=self.activation, **self.kwargs) self.dropout_layer = tf.keras.layers.Dropout(rate=dropout_rate) def call(self, inputs, training=True): return self.dense_layer(self.dropout_layer(inputs, training=self.use_dropout)) def dense_dropout(dropout_rate=0.4): return functools.partial(DenseDropout, dropout_rate=dropout_rate) def conv2d_dropout(dropout_rate=0.4): return functools.partial(Conv2DDropout, dropout_rate=dropout_rate) def lstm_dropout(dropout_rate=0.4): return functools.partial(LSTMDropout, dropout_rate=dropout_rate) def dropout_dense(dropout_rate=0.4): return functools.partial(DropoutDense, dropout_rate=dropout_rate) def scaled_reparameterization_layer(tfp_varitional_layer_obj, scale_factor=1. / 10000): def scaled_kl_fn(q, p, _): return tfp.distributions.kl_divergence(q, p) * scale_factor return functools.partial(tfp_varitional_layer_obj, kernel_divergence_fn=scaled_kl_fn, bias_divergence_fn=scaled_kl_fn) def customized_reparameterization_dense_layer(scale_factor=1. / 10000): # code from: https://github.com/tensorflow/probability/issues/409 # and https://github.com/tensorflow/probability/blob/v0.11.0/tensorflow_probability/python/layers/util.py#L202-L224 tfd = tfp.distributions def _posterior_mean_field(kernel_size, bias_size=0, dtype=None): n = kernel_size + bias_size c = np.log(np.expm1(1.)) return tf.keras.Sequential([ tfp.layers.VariableLayer(2 * n, dtype=dtype), tfp.layers.DistributionLambda(lambda t: tfd.Independent( tfd.Normal(loc=t[..., :n], scale=1e-5 + tf.nn.softplus(c + t[..., n:])), reinterpreted_batch_ndims=1)), ]) def _non_trainable_prior_fn(kernel_size, bias_size=0, dtype=None): def _distribution_fn(_): return tfd.Independent(tfd.Normal(loc=tf.zeros(kernel_size + bias_size, dtype=dtype), scale=1.), reinterpreted_batch_ndims=1) return _distribution_fn def _trainable_prior_fn(kernel_size, bias_size=0, dtype=None): return tf.keras.Sequential([ tfp.layers.VariableLayer(kernel_size + bias_size, dtype=dtype), tfp.layers.DistributionLambda( lambda mu: tfd.Independent(tfd.Normal(loc=mu, scale=1), reinterpreted_batch_ndims=1)), ]) return functools.partial( tfp.layers.DenseVariational, make_posterior_fn=_posterior_mean_field, make_prior_fn=_non_trainable_prior_fn, kl_weight=scale_factor) def produce_layer(ensemble_type=None, **kwargs): assert ensemble_type in ensemble_method_scope, 'only support ensemble method {}.'.format( ','.join(ensemble_method_scope) ) if ensemble_type == 'vanilla' or ensemble_type == 'deep_ensemble' or ensemble_type == 'weighted_ensemble': Dense = tf.keras.layers.Dense Conv2D = tf.keras.layers.Conv2D LSTM = tf.keras.layers.LSTM last_Dense = tf.keras.layers.Dense elif ensemble_type == 'mc_dropout': Dense = dense_dropout(kwargs['dropout_rate']) Conv2D = conv2d_dropout(kwargs['dropout_rate']) LSTM = lstm_dropout(kwargs['dropout_rate']) last_Dense = dropout_dense(kwargs['dropout_rate']) elif ensemble_type == 'bayesian': Dense = scaled_reparameterization_layer(tfp.layers.DenseReparameterization, kwargs['kl_scaler']) # customized_reparameterization_dense_layer(kwargs['kl_scaler']) # Conv2D = scaled_reparameterization_layer(tfp.layers.Convolution2DReparameterization, kwargs['kl_scaler']) LSTM = tf.keras.layers.LSTM last_Dense = scaled_reparameterization_layer(tfp.layers.DenseReparameterization, kwargs['kl_scaler']) # customized_reparameterization_dense_layer(kwargs['kl_scaler']) # else: raise ValueError('only support ensemble method {}.'.format(','.join(ensemble_method_scope))) return Dense, Conv2D, LSTM, last_Dense ##### neural network initialization ########### def get_fans(shape): fan_in = shape[0] if len(shape) == 2 else np.prod(shape[:-1]) fan_out = shape[1] if len(shape) == 2 else shape[-1] return fan_in, fan_out def glorot_uniform(shape): if len(shape) > 1: fan_in, fan_out = get_fans(shape) scale = np.sqrt(6. / (fan_in + fan_out)) return np.random.uniform(low=-scale, high=scale, size=shape) else: return np.zeros(shape, dtype=np.float32)
17,663
32.904031
176
py
malware-uncertainty
malware-uncertainty-master/core/feature/feature_extraction.py
import os import multiprocessing import collections import warnings import tempfile import numpy as np import pandas as pd from sklearn.preprocessing import MinMaxScaler from sklearn.cluster import KMeans from tools import utils from tools import progressbar_wrapper from core.ensemble.dataset_lib import build_dataset_from_numerical_data, \ build_dataset_via_generator, \ build_dataset_from_img_generator from config import logging, ErrorHandler logger = logging.getLogger('core.feature.feature_extraction') logger.addHandler(ErrorHandler) class FeatureExtraction(object): """Produce features for ML algorithms""" def __init__(self, naive_data_save_dir, intermediate_save_dir, file_ext=None, update=False, proc_number=2): """ initialization :param naive_data_save_dir: a directory for saving intermediates :param intermediate_save_dir: a directory for saving meta information :param file_ext: file extent :param update: boolean indicator for recomputing the naive features :param proc_number: process number """ self.naive_data_save_dir = naive_data_save_dir utils.mkdir(self.naive_data_save_dir) self.meta_data_save_dir = intermediate_save_dir utils.mkdir(self.meta_data_save_dir) self.file_ext = file_ext self.update = update self.proc_number = int(proc_number) def feature_extraction(self, sample_dir, use_order_features=False): """ extract the android features from Android packages and save the extractions into designed directory :param sample_dir: malicious / benign samples for the subsequent process of feature extraction :param use_order_features: following the order of the provided sample paths """ raise NotImplementedError def feature_preprocess(self, feature_path_list, gt_labels): """ pre-processing the naive data to accommodate the input format of ML algorithms :param feature_path_list: feature paths produced by the method of feature_extraction :param gt_labels: corresponding ground truth labels """ raise NotImplementedError def feature2ipt(self, feature_path_list, labels=None, is_training_set=False): """ Mapping features to the input space :param feature_path_list, a list of paths point to the features :param labels, ground truth labels :param is_training_set, boolean type """ raise NotImplementedError @staticmethod def _check(sample_dir): """ check a valid directory and produce a list of file paths """ if isinstance(sample_dir, str): if not os.path.exists(sample_dir): MSG = "No such directory or file {} exists!".format(sample_dir) raise ValueError(MSG) elif os.path.isfile(sample_dir): sample_path_list = [sample_dir] elif os.path.isdir(sample_dir): sample_path_list = list(utils.retrive_files_set(sample_dir, "", ".apk|")) assert len(sample_path_list) > 0, 'No files' else: raise ValueError(" No such path {}".format(sample_dir)) elif isinstance(sample_dir, list): sample_path_list = [path for path in sample_dir if os.path.isfile(path)] else: MSG = "A directory or a list of paths are allowed!" raise ValueError(MSG) return sample_path_list class DrebinFeature(FeatureExtraction): def __init__(self, naive_data_save_dir, intermediate_save_dir, file_ext='.drebin', update=False, proc_number=2): super(DrebinFeature, self).__init__(naive_data_save_dir, intermediate_save_dir, file_ext, update, proc_number) def feature_extraction(self, sample_dir, use_order_features=False): """ drebin features :return: 2D list, [[a list of features from an apk],...,[a list of features from an apk]] """ from core.feature.drebin import AxplorerMapping, get_drebin_feature sample_path_list = self._check(sample_dir) pool = multiprocessing.Pool(self.proc_number) pbar = progressbar_wrapper.ProgressBar() process_results = [] tasks = [] pmap = AxplorerMapping() for i, apk_path in enumerate(sample_path_list): sha256 = os.path.splitext(os.path.basename(apk_path))[0] # utils.get_sha256(apk_path) save_path = os.path.join(self.naive_data_save_dir, sha256 + self.file_ext) if os.path.exists(save_path) and (not self.update): continue tasks.append(apk_path) process_results = pool.apply_async(get_drebin_feature, args=(apk_path, pmap, save_path), callback=pbar.CallbackForProgressBar) pool.close() if process_results: pbar.DisplayProgressBar(process_results, len(tasks), type='hour') pool.join() feature_path_list = [] for i, apk_path in enumerate(sample_path_list): sha256_code = os.path.splitext(os.path.basename(apk_path))[0] # utils.get_sha256(apk_path) save_path = os.path.join(self.naive_data_save_dir, sha256_code + self.file_ext) if os.path.exists(save_path): feature_path_list.append(save_path) else: warnings.warn("Fail to perform feature extraction for '{}'".format(apk_path)) return feature_path_list def load_features(self, feature_path_list): """ load features :param feature_path_list: feature paths produced by the method of feature_extraction :return: a list of features """ from core.feature.drebin import wrapper_load_features feature_list = [] n_proc = 1 if multiprocessing.cpu_count() // 2 <= 1 else multiprocessing.cpu_count() // 2 pool = multiprocessing.Pool(n_proc) for res in pool.imap(wrapper_load_features, feature_path_list): if not isinstance(res, Exception): feature_list.append(res) else: print(str(res)) return feature_list def feature_preprocess(self, feature_path_list, gt_labels): """ pre-processing the naive data to accommodate the input format of ML algorithms :param feature_path_list: feature paths produced by the method of feature_extraction :param gt_labels: corresponding ground truth labels """ vocab_path = os.path.join(self.meta_data_save_dir, 'drebin.vocab') if self.update or (not os.path.exists(vocab_path)): assert len(feature_path_list) == len(gt_labels) features = self.load_features(feature_path_list) tmp_vocab = self.get_vocabulary(features) # we select 10,000 features selected_vocab = self.feature_selection(features, gt_labels, tmp_vocab, dim=10000) vocab_path = os.path.join(self.meta_data_save_dir, 'drebin.vocab') utils.dump_pickle(selected_vocab, vocab_path) return def feature2ipt(self, feature_path_list, labels=None, is_training_set=False): """ Mapping features to the input space :param feature_path_list: the feature paths produced by the method of feature_extraction :param labels: the ground truth labels correspond to features :param is_training_set, not used here :return tf.data; input dimension of an item of data :rtype tf.data.Dataset object; integer """ # load vocab_path = os.path.join(self.meta_data_save_dir, 'drebin.vocab') if not os.path.exists(vocab_path): raise ValueError("A vocabulary is needed.") vocab = utils.read_pickle(vocab_path) features = self.load_features(feature_path_list) dataX_np = self.get_feature_representation(features, vocab) input_dim = dataX_np.shape[1] if labels is not None: return build_dataset_from_numerical_data((dataX_np, labels)), input_dim else: return build_dataset_from_numerical_data(dataX_np), input_dim def feature_selection(self, train_features, train_y, vocab, dim): """ feature selection :param train_features: 2D feature :type train_features: numpy object :param train_y: ground truth labels :param vocab: a list of words (i.e., features) :param dim: the number of remained words :return: chose vocab """ is_malware = (train_y == 1) mal_features = np.array(train_features, dtype=object)[is_malware] ben_features = np.array(train_features, dtype=object)[~is_malware] if (len(mal_features) <= 0) or (len(ben_features) <= 0): return vocab mal_representations = self.get_feature_representation(mal_features, vocab) mal_frequency = np.sum(mal_representations, axis=0) / float(len(mal_features)) ben_representations = self.get_feature_representation(ben_features, vocab) ben_frequency = np.sum(ben_representations, axis=0) / float(len(ben_features)) # eliminate the words showing zero occurrence in apk files is_null_feature = np.all(mal_representations == 0, axis=0) & np.all(ben_representations, axis=0) mal_representations, ben_representations = None, None vocab_filtered = list(np.array(vocab)[~is_null_feature]) if len(vocab_filtered) <= dim: return vocab_filtered else: feature_frq_diff = np.abs(mal_frequency[~is_null_feature] - ben_frequency[~is_null_feature]) position_flag = np.argsort(feature_frq_diff)[::-1][:dim] vocab_selected = [] for p in position_flag: vocab_selected.append(vocab_filtered[p]) return vocab_selected def load_vocabulary(self): vocab_path = os.path.join(self.meta_data_save_dir, 'drebin.vocab') if not os.path.exists(vocab_path): raise ValueError("A vocabulary is needed.") vocab = utils.read_pickle(vocab_path) return vocab @staticmethod def get_vocabulary(feature_list, n=300000): """ obtain the vocabulary based on the feature :param feature_list: 2D list of naive feature :param n: the number of top frequency items :return: feature vocabulary """ c = collections.Counter() for features in feature_list: for feature in features: c[feature] = c[feature] + 1 vocab, count = zip(*c.most_common(n)) return list(vocab) @staticmethod def get_feature_representation(feature_list, vocab): """ mapping feature to numerical representation :param feature_list: 2D feature list with shape [number of files, number of feature] :param vocab: a list of words :return: 2D representation :rtype numpy.ndarray """ N = len(feature_list) M = len(vocab) assert N > 0 and M > 0 representations = np.zeros((N, M), dtype=np.float32) dictionary = dict(zip(vocab, range(len(vocab)))) for i, features in enumerate(feature_list): if len(features) > 0: filled_positions = [idx for idx in list(map(dictionary.get, features)) if idx is not None] if len(filled_positions) != 0: representations[i, filled_positions] = 1. else: warnings.warn("Produce zero feature vector.") return representations class OpcodeSeq(FeatureExtraction): """ get opcode sequences """ def __init__(self, naive_data_save_dir, intermediate_save_dir=None, file_ext='.opcode', update=False, proc_number=2): """ initialization :param naive_data_save_dir: a directory for saving intermediates :param file_ext: file extent :param update: boolean indicator for recomputing the naive features :param proc_number: process number """ super(OpcodeSeq, self).__init__(naive_data_save_dir, intermediate_save_dir, file_ext, update, proc_number) def feature_extraction(self, sample_dir, use_order_features=False): from core.feature.opcodeseq import feature_extr_wrapper sample_path_list = self._check(sample_dir) pool = multiprocessing.Pool(self.proc_number) pbar = progressbar_wrapper.ProgressBar() process_results = [] tasks = [] for i, apk_path in enumerate(sample_path_list): sha256_code = os.path.splitext(os.path.basename(apk_path))[0] # utils.get_sha256(apk_path) save_path = os.path.join(self.naive_data_save_dir, sha256_code + self.file_ext) if os.path.exists(save_path) and not self.update: continue tasks.append(apk_path) process_results = pool.apply_async(feature_extr_wrapper, args=(apk_path, save_path), callback=pbar.CallbackForProgressBar) pool.close() if process_results: pbar.DisplayProgressBar(process_results, len(tasks), type='hour') pool.join() feature_paths = [] for i, apk_path in enumerate(sample_path_list): sha256_code = os.path.splitext(os.path.basename(apk_path))[0] # utils.get_sha256(apk_path) save_path = os.path.join(self.naive_data_save_dir, sha256_code + self.file_ext) if os.path.exists(save_path): feature_paths.append(save_path) else: warnings.warn("Fail to perform feature extraction for '{}'".format(apk_path)) return feature_paths def feature_preprocess(self, feature_path_list, gt_labels): """ pre-processing the naive data to accommodate the input format of ML algorithms :param feature_path_list: a list of paths directing to saved features :param gt_labels: corresponding ground truth labels """ return def feature2ipt(self, feature_path_list, labels=None, is_training_set=False): """ Mapping features to the input space """ from core.feature.opcodeseq import read_opcode, read_opcode_wrapper from core.ensemble.model_hp import text_cnn_hparam def padding_opcodes(features_of_an_apk, padding_char=0): padding_seq = [] padding_chars = [padding_char] * text_cnn_hparam.kernel_size for i, seq in enumerate(features_of_an_apk): padding_seq.extend(seq) padding_seq.extend(padding_chars) if len(padding_seq) > text_cnn_hparam.max_sequence_length: break return np.array(padding_seq[:text_cnn_hparam.max_sequence_length]) def generator(): if labels is not None: for path, label in zip(feature_path_list, labels): data_padded = padding_opcodes(read_opcode(path)) yield data_padded[:text_cnn_hparam.max_sequence_length], label else: for path in feature_path_list: data_padded = padding_opcodes(read_opcode(path)) yield data_padded[:text_cnn_hparam.max_sequence_length] with tempfile.NamedTemporaryFile() as f: return build_dataset_via_generator(generator, labels, f.name), None class MultiModality(FeatureExtraction): def __init__(self, naive_data_save_dir, intermediate_save_dir, use_feature_selection=True, feature_dimension=10000, cluster_centers=100, similar_threshold=0.5, file_ext='.multimod', update=False, proc_number=2 ): """ initialization :param naive_data_save_dir: a directory for saving intermediates :param intermediate_save_dir: a directory for saving meta information :param use_feature_selection: select features with top frequencies :param feature_dimension: the number of selected features, default 10,000 :param cluster_centers: the number of cluster centers, default 100 :param file_ext: file extent :param update: boolean indicator for recomputing the naive features :param proc_number: process number """ super(MultiModality, self).__init__(naive_data_save_dir, intermediate_save_dir, file_ext, update, proc_number ) self.use_feature_selection = use_feature_selection self.feature_dimension = feature_dimension self.cluster_centers = cluster_centers self.similar_threshold = similar_threshold def feature_extraction(self, sample_dir, use_order_features=False): """ extract the android features from Android packages and save the extractions into designed directory """ from core.feature.multimodality import API_LIST, get_multimod_feature sample_path_list = self._check(sample_dir) pool = multiprocessing.Pool(self.proc_number) pbar = progressbar_wrapper.ProgressBar() process_results = [] tasks = [] for i, apk_path in enumerate(sample_path_list): sha256_code = os.path.splitext(os.path.basename(apk_path))[0] # utils.get_sha256(apk_path) save_path = os.path.join(self.naive_data_save_dir, sha256_code + self.file_ext) if os.path.exists(save_path) and (not self.update): continue tasks.append(apk_path) process_results = pool.apply_async(get_multimod_feature, args=(apk_path, API_LIST, save_path), callback=pbar.CallbackForProgressBar) pool.close() if process_results: pbar.DisplayProgressBar(process_results, len(tasks), type='hour') pool.join() feature_paths = [] for i, apk_path in enumerate(sample_path_list): sha256_code = os.path.splitext(os.path.basename(apk_path))[0] # utils.get_sha256(apk_path) save_path = os.path.join(self.naive_data_save_dir, sha256_code + self.file_ext) if os.path.exists(save_path): feature_paths.append(save_path) else: warnings.warn("Fail to perform feature extraction for '{}'".format(apk_path)) return feature_paths def feature_preprocess(self, feature_path_list, gt_labels): """ pre-processing the naive data to accommodate the input format of ML algorithms :param feature_path_list: a list of paths directing to save features. For each apk, features produced by the method of feature_extraction, 2D list [[feature type 1,...,feature type 5],...,] :param gt_labels: corresponding ground truth labels """ assert len(feature_path_list) == len(gt_labels), 'inconsistent dataset' vocab_path = os.path.join(self.meta_data_save_dir, 'multimodality.vocab') if os.path.exists(vocab_path) and not self.update: vocab_list = self.load_meta_info(vocab_path) else: vocab_list = self.get_vocab(feature_path_list, gt_labels, self.use_feature_selection, self.feature_dimension) # saving self.save_meta_info(vocab_list, vocab_path) dataX_list = self.feature_mapping(feature_path_list, vocab_list) # further processing scaler_path = os.path.join(self.meta_data_save_dir, 'multimodality.scaler') scaled_dataX_list = self.data_scaling(dataX_list, scaler_path) # clustering for last three types of features cluster_center_path = os.path.join(self.meta_data_save_dir, 'multimodality.center') if not os.path.exists(cluster_center_path) or self.update: cluster_centers = [] for i, dataX in enumerate(scaled_dataX_list[2:]): # produce the last three similarity-based feature center_vec = self.k_means_clustering(dataX, self.cluster_centers) cluster_centers.append(center_vec) # saving self.save_meta_info(cluster_centers, cluster_center_path) return def feature2ipt(self, feature_path_list, labels=None, is_training_set=False): """ Mapping features to the input space """ # assert self._check_features(features) vocab_path = os.path.join(self.meta_data_save_dir, 'multimodality.vocab') vocab_list = self.load_meta_info(vocab_path) scaler_path = os.path.join(self.meta_data_save_dir, 'multimodality.scaler') scalers = self.load_meta_info(scaler_path) cluster_center_path = os.path.join(self.meta_data_save_dir, 'multimodality.center') centers = self.load_meta_info(cluster_center_path) dataX_list = self.feature_mapping(feature_path_list, vocab_list) for i, dataX in enumerate(dataX_list): dataX_list[i] = scalers[i].transform(dataX) for i, center in enumerate(centers): dataX_list[2 + i] = self._get_similarity(dataX_list[2 + i], center, self.similar_threshold) input_dim = [] for x in dataX_list: input_dim.append(x.shape[1]) # build dataset if labels is not None: data_tf = build_dataset_from_numerical_data(tuple(dataX_list)) y = build_dataset_from_numerical_data(labels) from tensorflow import data return data.Dataset.zip((data_tf, y)), input_dim else: return dataX_list, input_dim def save_meta_info(self, data, path): if self.update or (not os.path.exists(path)): utils.dump_joblib(data, path) return @staticmethod def load_meta_info(path): if os.path.exists(path): return utils.read_joblib(path) else: raise ValueError("No such data.") @staticmethod def get_vocab(feature_path_list, gt_labels=None, use_feature_selection=False, dim=10000): """ build vocabulary for five kinds of feature, including permission/component/environment, string, method api, method opcodes, shared library, each of which are presented in the 'collections.defaultdict' format :param feature_path_list: a list of paths redirecting to save features :param gt_labels: ground truth labels (optional) :param use_feature_selection: conducting feature extraction or not (False means no, and True means yes) :param dim: the number of selected feature (optional) :return: list of vocabularies corresponding to five kinds of feature """ from core.feature.multimodality import wrapper_load_features feature_list = [] n_proc = 1 if multiprocessing.cpu_count() // 2 <= 1 else multiprocessing.cpu_count() // 2 pool = multiprocessing.Pool(n_proc) for res in pool.imap(wrapper_load_features, feature_path_list): if not isinstance(res, Exception): feature_list.append(res) else: print(str(res)) assert isinstance(feature_list, list) and len(feature_list) > 0, 'Type: {} and length: {}'.format( type(feature_list), len(feature_list)) number_of_types = len(feature_list[0]) number_of_samples = len(feature_list) vocabulary_list = [] for t in range(number_of_types): c = collections.Counter() for j in range(number_of_samples): feature_dict = feature_list[j][t] for k, v in feature_dict.items(): c[k] += v if not use_feature_selection: if len(c) > 0: vocab, count = zip(*c.items()) else: vocab = [] else: if len(c) > 0: vocab, count = zip(*c.most_common(dim)) # filter out words with low frequency else: vocab = [] vocabulary_list.append(list(vocab)) return vocabulary_list @staticmethod def feature_mapping(feature_path_list, vocab_list): """ mapping feature to numerical representation :param feature_path_list: a list of paths redirecting to saved features :param vocab_list: several lists of words :return: 2D representation :rtype numpy.ndarray """ from core.feature.multimodality import wrapper_load_features feature_list = [] n_proc = 1 if multiprocessing.cpu_count() // 2 <= 1 else multiprocessing.cpu_count() // 2 pool = multiprocessing.Pool(n_proc) for res in pool.imap(wrapper_load_features, feature_path_list): if not isinstance(res, Exception): feature_list.append(res) else: print(str(res)) pool.close() pool.join() assert len(feature_list[0]) == len(vocab_list) number_of_feature_types = len(vocab_list) representation_list = [] for t in range(number_of_feature_types): # feature_dict = np.array(feature_list)[:, t] number_of_samples = len(feature_list) vocab = vocab_list[t] M = len(list(vocab)) representation = np.zeros((number_of_samples, M), dtype=np.float32) dictionary = dict(zip(vocab, range(M))) for j in range(number_of_samples): feature_dict = feature_list[j][t] if len(feature_dict) > 0: filled_positions = [idx for idx in list(map(dictionary.get, list(feature_dict.keys()))) if idx is not None] filled_values = [feature_dict.get(key) for key in list(feature_dict.keys()) if dictionary.get(key) is not None] if len(filled_positions) != 0: representation[j, filled_positions] = filled_values[:] else: warnings.warn("Produce zero feature vector.") representation_list.append(representation) return representation_list def data_scaling(self, data_x_list, scalar_saving_path=None): """ minmax scaling for numerical feature representations :param data_x_list: a list of un-normalized feature representation :param scalar_saving_path: :return: scaled feature representation :rtype : list of 2d numpy.ndarray """ if os.path.exists(scalar_saving_path) and not self.update: scalers = self.load_meta_info(scalar_saving_path) else: scalers = [] for i, dataX in enumerate(data_x_list): scaler = MinMaxScaler() scaler.fit(dataX) scalers.append(scaler) self.save_meta_info(scalers, scalar_saving_path) for i, dataX in enumerate(data_x_list): data_x_list[i] = scalers[i].transform(dataX) return data_x_list @staticmethod def k_means_clustering(data_x, number_of_cluster_centers=100): N = data_x.shape[0] n_clusters = number_of_cluster_centers if number_of_cluster_centers < N else N // 2 kmeans = KMeans(n_clusters=n_clusters, random_state=0).fit(data_x) return kmeans.cluster_centers_ @staticmethod def _get_similarity(data_x, anchor, threshold=0.5): """ get similarity matrix :return: similarity-based feature representation """ # The following method of calculating the similarity matrix might be different to the proposal # in the paper (i.e., Algorithm 2), which is confusing to us. similar_mat = 1. / np.column_stack( [np.max(np.square(data_x - center) ** 0.5 + 1, axis=-1) for center in anchor]) return np.greater(similar_mat, threshold).astype(np.float32) @staticmethod def _check_features(features): """ check the completeness :param features: a list of features, each item presented in the 'collections.defaultdict' format :return: True or False """ return (isinstance(features, list)) and (len(features) > 0) and ( isinstance(features[0][0], dict)) class DexToImage(FeatureExtraction): """ Convert the dex files to a RGB image """ def __init__(self, naive_data_save_dir, intermediate_save_dir, file_ext='.jpg', update=False, proc_number=2 ): super(DexToImage, self).__init__(naive_data_save_dir, intermediate_save_dir, file_ext, update, proc_number) def feature_extraction(self, sample_dir, use_order_features=False): from core.feature.dex2img import dex2img sample_path_list = self._check(sample_dir) pool = multiprocessing.Pool(self.proc_number) pbar = progressbar_wrapper.ProgressBar() process_results = [] tasks = [] for i, apk_path in enumerate(sample_path_list): sha256_code = os.path.splitext(os.path.basename(apk_path))[0] # utils.get_sha256(apk_path) save_path = os.path.join(self.naive_data_save_dir, sha256_code + self.file_ext) if os.path.exists(save_path) and (not self.update): continue tasks.append(apk_path) process_results = pool.apply_async(dex2img, args=(apk_path, save_path), callback=pbar.CallbackForProgressBar) pool.close() if process_results: pbar.DisplayProgressBar(process_results, len(tasks), type='hour') pool.join() result_paths = [] for i, apk_path in enumerate(sample_path_list): sha256_code = os.path.splitext(os.path.basename(apk_path))[0] # utils.get_sha256(apk_path) res_path = os.path.join(self.naive_data_save_dir, sha256_code + self.file_ext) if os.path.exists(res_path): result_paths.append(res_path) else: warnings.warn("Fail to perform feature extraction for '{}'".format(apk_path)) return result_paths def feature_preprocess(self, feature_path_list, gt_labels): """ pre-processing the naive data to accommodate the input format of ML algorithms :param feature_path_list: a list of paths directing to image files :param gt_labels: corresponding ground truth labels """ return def feature2ipt(self, feature_path_list, labels=None, is_training_set=False, image_size=[500, 500]): """ Mapping features to the input space """ image_names = [os.path.basename(path) for path in feature_path_list] from tensorflow.keras.preprocessing.image import ImageDataGenerator from core.ensemble.model_hp import train_hparam if len(image_names) % train_hparam.batch_size == 0: batches = len(image_names) // train_hparam.batch_size else: batches = len(image_names) // train_hparam.batch_size + 1 assert batches >= 0, 'No data. Exit' img_generator_obj = ImageDataGenerator(rescale=1. / 255) if is_training_set and labels is not None: labels_string = [(lambda x: 'malware' if x else 'benware')(label) for label in labels] w = float(len(labels) - np.sum(labels)) / np.sum(labels) w = 1. if w <= 1. else w weights = [(lambda x: w if x else 1.)(label) for label in labels] img_pd = pd.DataFrame({'image': image_names, 'labels': labels_string, 'sample_weights': weights}) img_generator = img_generator_obj.flow_from_dataframe(img_pd, directory=self.naive_data_save_dir, x_col='image', y_col='labels', weight_col='sample_weights', classes=['benware', 'malware'], target_size=image_size, class_mode='binary', shuffle=True, batch_size=train_hparam.batch_size ) elif not is_training_set and labels is not None: labels_string = [(lambda x: 'malware' if x else 'benware')(label) for label in labels] img_pd = pd.DataFrame({'image': image_names, 'labels': labels_string}) img_generator = img_generator_obj.flow_from_dataframe(img_pd, directory=self.naive_data_save_dir, x_col='image', y_col='labels', classes=['benware', 'malware'], target_size=image_size, class_mode='binary', shuffle=False, batch_size=train_hparam.batch_size ) else: img_pd = pd.DataFrame({'image': image_names}) img_generator = img_generator_obj.flow_from_dataframe(img_pd, directory=self.naive_data_save_dir, x_col='image', target_size=image_size, shuffle=False, class_mode=None, batch_size=train_hparam.batch_size # lead to ) def generator(): for _ in range(batches): yield next(img_generator) return (build_dataset_from_img_generator(generator, input_dim=[*image_size, 3], y=labels, is_training=is_training_set), [*image_size, 3]) class APISequence(FeatureExtraction): """Obtain api sequences based on the function call graph""" def __init__(self, naive_data_save_dir, intermediate_save_dir, use_feature_selection=True, ratio=0.25, file_ext='.seq', update=False, proc_number=2 ): """ initialization :param naive_data_save_dir: a directory for saving intermediates :param intermediate_save_dir: a directory for saving meta information :param use_feature_selection: use feature selection to filtering out entities with high frequencies :param ratio: resides the range of [0, 1] and denotes a portion of features will be neglected :param file_ext: file extent :param update: boolean indicator for recomputing the naive features :param proc_number: process number """ super(APISequence, self).__init__(naive_data_save_dir, intermediate_save_dir, file_ext, update, proc_number) self.use_feature_selection = use_feature_selection self.ratio = ratio from core.ensemble.model_hp import droidetec_hparam self.maximum_vocab_size = droidetec_hparam.vocab_size def feature_extraction(self, sample_dir, use_order_features=False): """ save the android features and return the saved paths """ from core.feature.apiseq import get_api_sequence sample_path_list = self._check(sample_dir) pool = multiprocessing.Pool(self.proc_number) pbar = progressbar_wrapper.ProgressBar() process_results = [] tasks = [] for i, apk_path in enumerate(sample_path_list): sha256_code = os.path.splitext(os.path.basename(apk_path))[0] # utils.get_sha256(apk_path) save_path = os.path.join(self.naive_data_save_dir, sha256_code + self.file_ext) if os.path.exists(save_path) and (not self.update): continue tasks.append(apk_path) process_results = pool.apply_async(get_api_sequence, args=(apk_path, save_path), callback=pbar.CallbackForProgressBar) pool.close() if process_results: pbar.DisplayProgressBar(process_results, len(tasks), type='hour') pool.join() feature_paths = [] for i, apk_path in enumerate(sample_path_list): sha256_code = os.path.splitext(os.path.basename(apk_path))[0] # utils.get_sha256(apk_path) save_path = os.path.join(self.naive_data_save_dir, sha256_code + self.file_ext) if os.path.exists(save_path): feature_paths.append(save_path) else: warnings.warn("Fail to perform feature extraction for '{}'".format(apk_path)) return feature_paths def feature_preprocess(self, feature_path_list, gt_labels): """ pre-processing the naive data to accommodate the input format of ML algorithms """ dict_saving_path = os.path.join(self.meta_data_save_dir, 'apiseq.dict') if not os.path.exists(dict_saving_path) or self.update: vocab = self.get_vocab(feature_path_list, gt_labels) dictionary = dict(zip(vocab, range(len(vocab)))) # saving utils.dump_joblib(dictionary, dict_saving_path) return def feature2ipt(self, feature_path_list, labels=None, is_training_set=False): """ Mapping features to the input space """ dict_saving_path = os.path.join(self.meta_data_save_dir, 'apiseq.dict') if os.path.exists(dict_saving_path): dictionary = utils.read_joblib(dict_saving_path) else: raise ValueError("No meta data") from core.ensemble.model_hp import droidetec_hparam def generator(): if labels is not None: discrete_features = self.feature_mapping(feature_path_list, dictionary) assert len(discrete_features) == len(labels), 'inconsistent data vs. corresponding label' for data, label in zip(*(discrete_features, labels)): yield data[:droidetec_hparam.max_sequence_length], label else: from core.feature.apiseq import wrapper_mapping for feature_path in feature_path_list: data = wrapper_mapping([feature_path, dictionary]) if isinstance(data, Exception): raise ValueError("Cannot load the specified feature:{}".format( feature_path )) yield data[:droidetec_hparam.max_sequence_length] with tempfile.NamedTemporaryFile() as f: return build_dataset_via_generator(generator, labels, f.name), None def get_vocab(self, feature_path_list, gt_labels): """ create vocabulary based on a list of feature :param feature_path_list: a list of feature paths :param gt_labels: ground truth labels :return: vocabulary :rtype: list """ if self.use_feature_selection: assert 0. < self.ratio <= 1., 'the ratio should be (0,1]' from core.feature.apiseq import wrapper_load_feature c_mal = collections.Counter() c_ben = collections.Counter() # has side-effect, consuming lots of local disk n_proc = 1 if multiprocessing.cpu_count() // 2 <= 1 else multiprocessing.cpu_count() // 2 pool = multiprocessing.Pool(n_proc) for res, label in zip(pool.imap(wrapper_load_feature, feature_path_list), gt_labels): if isinstance(res, list): if label: c_mal.update(res) else: c_ben.update(res) elif isinstance(res, Exception): print(str(res)) else: raise ValueError pool.close() pool.join() if not self.use_feature_selection: c_mal.update(c_ben) # c_all = c_mal else: api_num_mal = len(c_mal) api_hf_mal = dict(c_mal.most_common(int(api_num_mal * self.ratio))).keys() api_num_ben = len(c_ben) api_hf_ben = dict(c_ben.most_common(int(api_num_ben * self.ratio))).keys() common_apis = [e for e in api_hf_mal if e in api_hf_ben] for api in common_apis: c_mal[api] = 0 c_ben[api] = 0 c_mal.update(c_ben) logger.info('Filtering out {} apis.'.format(len(common_apis))) c_all = dict(c_mal.most_common(self.maximum_vocab_size - 1)) # saving a slot for null features c_all['sos'] = 1 vocab, count = zip(*c_all.items()) return list(vocab) def feature_mapping(self, feature_path_list, dictionary): """ mapping feature to numerical representation :param feature_path_list: a list of feature paths :param dictionary: vocabulary -> index :return: 2D representation :rtype numpy.ndarray """ numerical_features = [] from core.feature.apiseq import wrapper_mapping from core.ensemble.model_hp import droidetec_hparam n_proc = 1 if multiprocessing.cpu_count() // 2 <= 1 else multiprocessing.cpu_count() // 2 pool = multiprocessing.Pool(n_proc) pargs = [(path, dictionary) for path in feature_path_list] for res, path in zip(pool.imap(wrapper_mapping, pargs), feature_path_list): if not isinstance(res, Exception): numerical_features.append(res[:droidetec_hparam.max_sequence_length]) else: warnings.warn(str(res) + ': ' + path) pool.close() pool.join() return numerical_features
44,809
43.322453
115
py
malware-uncertainty
malware-uncertainty-master/core/ensemble/deep_ensemble.py
from os import path import time import numpy as np import tensorflow as tf from core.ensemble.vanilla import Vanilla from core.ensemble.model_hp import train_hparam from core.ensemble.dataset_lib import build_dataset_from_numerical_data from tools import utils from config import logging, ErrorHandler logger = logging.getLogger('ensemble.deep_ensemble') logger.addHandler(ErrorHandler) class DeepEnsemble(Vanilla): def __init__(self, architecture_type='dnn', base_model=None, n_members=5, model_directory=None, name='DEEPENSEMBLE' ): super(DeepEnsemble, self).__init__(architecture_type, base_model, n_members, model_directory, name) self.hparam = train_hparam self.ensemble_type = 'deep_ensemble' class WeightedDeepEnsemble(Vanilla): def __init__(self, architecture_type='dnn', base_model=None, n_members=5, model_directory=None, name='WEIGTHEDDEEPENSEMBLE' ): super(WeightedDeepEnsemble, self).__init__(architecture_type, base_model, n_members, model_directory, name) self.hparam = train_hparam self.ensemble_type = 'deep_ensemble' self.weight_modular = None def get_weight_modular(self): class Simplex(tf.keras.constraints.Constraint): def __call__(self, w): return tf.math.softmax(w - tf.math.reduce_max(w), axis=0) inputs = tf.keras.Input(shape=(self.n_members,)) outs = tf.keras.layers.Dense(1, use_bias=False, activation=None, kernel_constraint=Simplex(), name='simplex')( inputs) return tf.keras.Model(inputs=inputs, outputs=outs) def predict(self, x, use_prob=False): """ conduct prediction """ self.base_model = None self.weight_modular = None self.weights_list = [] self._optimizers_dict = [] self.load_ensemble_weights() output_list = [] start_time = time.time() for base_model in self.model_generator(): if isinstance(x, tf.data.Dataset): output_list.append(base_model.predict(x, verbose=1)) elif isinstance(x, (np.ndarray, list)): output_list.append(base_model.predict(x, batch_size=self.hparam.batch_size, verbose=1)) else: raise ValueError total_time = time.time() - start_time logger.info('Inference costs {} seconds.'.format(total_time)) assert self.weight_modular is not None output = self.weight_modular(np.hstack(output_list)).numpy() if not use_prob: return np.stack(output_list, axis=1), self.weight_modular.get_layer('simplex').get_weights() else: return output def fit(self, train_set, validation_set=None, input_dim=None, **kwargs): """ fit the ensemble by producing a lists of model weights :param train_set: tf.data.Dataset, the type shall accommodate to the input format of Tensorflow models :param validation_set: validation data, optional :param input_dim: integer or list, input dimension except for the batch size """ # training preparation if self.base_model is None: self.build_model(input_dim=input_dim) if self.weight_modular is None: self.weight_modular = self.get_weight_modular() self.base_model.compile( optimizer=tf.keras.optimizers.Adam(learning_rate=self.hparam.learning_rate, clipvalue=self.hparam.clipvalue), loss=tf.keras.losses.BinaryCrossentropy(), metrics=[tf.keras.metrics.BinaryAccuracy()], ) self.weight_modular.compile( optimizer=tf.keras.optimizers.Adam(learning_rate=self.hparam.learning_rate, clipvalue=self.hparam.clipvalue), loss=tf.keras.losses.BinaryCrossentropy(), metrics=[tf.keras.metrics.BinaryAccuracy()], ) # training logger.info("hyper-parameters:") logger.info(dict(self.hparam._asdict())) logger.info("...training start!") best_val_accuracy = 0. total_time = 0. for epoch in range(self.hparam.n_epochs): for member_idx in range(self.n_members): if member_idx < len(self.weights_list): # loading former weights self.base_model.set_weights(self.weights_list[member_idx]) self.base_model.optimizer.set_weights(self._optimizers_dict[member_idx]) elif member_idx == 0: pass # do nothing else: self.reinitialize_base_model() msg = 'Epoch {}/{}, member {}/{}, and {} member(s) in list'.format(epoch + 1, self.hparam.n_epochs, member_idx + 1, self.n_members, len(self.weights_list)) print(msg) start_time = time.time() self.base_model.fit(train_set, epochs=epoch + 1, initial_epoch=epoch, validation_data=validation_set ) self.update_weights(member_idx, self.base_model.get_weights(), self.base_model.optimizer.get_weights()) end_time = time.time() total_time += end_time - start_time # training weight modular msg = "train the weight modular at epoch {}/{}" print(msg.format(epoch, self.hparam.n_epochs)) start_time = time.time() history = self.fit_weight_modular(train_set, validation_set, epoch) end_time = time.time() total_time += end_time - start_time # saving logger.info('Training ensemble costs {} in total (including validation).'.format(total_time)) train_acc = history.history['binary_accuracy'][0] val_acc = history.history['val_binary_accuracy'][0] msg = 'Epoch {}/{}: training accuracy {:.5f}, validation accuracy {:.5f}.'.format( epoch + 1, self.hparam.n_epochs, train_acc, val_acc ) logger.info(msg) if (epoch + 1) % self.hparam.interval == 0: if val_acc > best_val_accuracy: self.save_ensemble_weights() best_val_accuracy = val_acc msg = '\t The best validation accuracy is {:.5f}, obtained at epoch {}/{}'.format( best_val_accuracy, epoch + 1, self.hparam.n_epochs ) logger.info(msg) return def fit_weight_modular(self, train_set, validation_set, epoch): """ fit weight modular :param train_set: training set :param validation_set: validation set :param epoch: integer, training epoch :return: None """ # obtain data def get_data(x_y_set): tsf_x = [] tsf_y = [] for _x, _y in x_y_set: _x_list = [] for base_model in self.model_generator(): _x_pred = base_model(_x) _x_list.append(_x_pred) tsf_x.append(np.hstack(_x_list)) tsf_y.append(_y) return np.vstack(tsf_x), np.concatenate(tsf_y) transform_train_set = build_dataset_from_numerical_data(get_data(train_set)) transform_val_set = build_dataset_from_numerical_data(get_data(validation_set)) history = self.weight_modular.fit(transform_train_set, epochs=epoch + 1, initial_epoch=epoch, validation_data=transform_val_set ) return history def save_ensemble_weights(self): if not path.exists(self.save_dir): utils.mkdir(self.save_dir) # save model configuration try: config = self.base_model.to_json() utils.dump_json(config, path.join(self.save_dir, self.architecture_type + '.json')) # lightweight method for saving model configurature except Exception as e: pass finally: if not path.exists(path.join(self.save_dir, self.architecture_type)): utils.mkdir(path.join(self.save_dir, self.architecture_type)) self.base_model.save(path.join(self.save_dir, self.architecture_type)) print("Save the model configuration to directory {}".format(self.save_dir)) # save model weights utils.dump_joblib(self.weights_list, path.join(self.save_dir, self.architecture_type + '.model')) utils.dump_joblib(self._optimizers_dict, path.join(self.save_dir, self.architecture_type + '.model.metadata')) print("Save the model weights to directory {}".format(self.save_dir)) # save weight modular self.weight_modular.save(path.join(self.save_dir, self.architecture_type + '_weight_modular')) print("Save the weight modular weights to directory {}".format(self.save_dir)) return def load_ensemble_weights(self): if path.exists(path.join(self.save_dir, self.architecture_type + '.json')): config = utils.load_json(path.join(self.save_dir, self.architecture_type + '.json')) self.base_model = tf.keras.models.model_from_json(config) elif path.exists(path.join(self.save_dir, self.architecture_type)): self.base_model = tf.keras.models.load_model(path.join(self.save_dir, self.architecture_type)) else: logger.error("File not found: ".format(path.join(self.save_dir, self.architecture_type + '.json'))) raise FileNotFoundError print("Load model config from {}.".format(self.save_dir)) if path.exists(path.join(self.save_dir, self.architecture_type + '.model')): self.weights_list = utils.read_joblib(path.join(self.save_dir, self.architecture_type + '.model')) else: logger.error("File not found: ".format(path.join(self.save_dir, self.architecture_type + '.model'))) raise FileNotFoundError print("Load model weights from {}.".format(self.save_dir)) if path.exists(path.join(self.save_dir, self.architecture_type + '.model.metadata')): self._optimizers_dict = utils.read_joblib( path.join(self.save_dir, self.architecture_type + '.model.metadata')) else: self._optimizers_dict = [None] * len(self.weights_list) if path.exists(path.join(self.save_dir, self.architecture_type + '_weight_modular')): self.weight_modular = tf.keras.models.load_model( path.join(self.save_dir, self.architecture_type + '_weight_modular')) return
11,862
44.98062
133
py
malware-uncertainty
malware-uncertainty-master/core/ensemble/anchor_ensemble.py
import os.path as path import time import tensorflow as tf import numpy as np from core.ensemble.vanilla import Vanilla from core.ensemble.model_hp import train_hparam, anchor_hparam from core.ensemble.model_lib import model_builder from tools import utils from config import logging logger = logging.getLogger('ensemble.vanilla') class AnchorEnsemble(Vanilla): def __init__(self, architecture_type='dnn', base_model=None, n_members=2, model_directory=None, name='ANCHOR'): """ initialization :param architecture_type: the type of base model :param base_model: an object of base model :param n_members: number of base models :param model_directory: a folder for saving ensemble weights """ super(AnchorEnsemble, self).__init__(architecture_type, base_model, n_members, model_directory) self.hparam = utils.merge_namedtuples(train_hparam, anchor_hparam) self.ensemble_type = 'anchor' self.name = name.lower() self.save_dir = path.join(self.model_directory, self.name) def build_model(self, input_dim=None): """ Build an ensemble model -- only the homogeneous structure is considered :param input_dim: integer or list, input dimension shall be set in some cases under eager mode """ callable_graph = model_builder(self.architecture_type) @callable_graph(input_dim) def _builder(): seed = np.random.choice(self.hparam.random_seed) return utils.produce_layer(self.ensemble_type, scale=self.hparam.scale, batch_size=self.hparam.batch_size, seed=seed) self.base_model = _builder() def model_generator(self): try: for m in range(self.n_members): self.base_model = None self.load_ensemble_weights(m) yield self.base_model except Exception as e: raise Exception("Cannot load model:{}.".format(str(e))) def fit(self, train_set, validation_set=None, input_dim=None, **kwargs): """ fit the ensemble by producing a lists of model weights :param train_set: tf.data.Dataset, the type shall accommodate to the input format of Tensorflow models :param validation_set: validation data, optional :param input_dim: integer or list, input dimension except for the batch size """ # training logger.info("hyper-parameters:") logger.info(dict(self.hparam._asdict())) logger.info("...training start!") np.random.seed(self.hparam.random_seed) train_set = train_set.shuffle(buffer_size=100, reshuffle_each_iteration=True) for member_idx in range(self.n_members): self.base_model = None self.build_model(input_dim=input_dim) self.base_model.compile( optimizer=tf.keras.optimizers.Adam(learning_rate=self.hparam.learning_rate), loss=tf.keras.losses.BinaryCrossentropy(), metrics=[tf.keras.metrics.BinaryAccuracy()], ) for epoch in range(self.hparam.n_epochs): total_time = 0. msg = 'Epoch {}/{}, and member {}/{}'.format(epoch + 1, self.hparam.n_epochs, member_idx + 1, self.n_members) print(msg) start_time = time.time() self.base_model.fit(train_set, epochs=epoch + 1, initial_epoch=epoch, validation_data=validation_set ) end_time = time.time() total_time += end_time - start_time # saving logger.info('Training ensemble costs {} seconds at this epoch'.format(total_time)) if (epoch + 1) % self.hparam.interval == 0: self.save_ensemble_weights(member_idx) def save_ensemble_weights(self, member_idx=0): if not path.exists(path.join(self.save_dir, self.architecture_type + '_{}'.format(member_idx))): utils.mkdir(path.join(self.save_dir, self.architecture_type + '_{}'.format(member_idx))) # save model configuration self.base_model.save(path.join(self.save_dir, self.architecture_type + '_{}'.format(member_idx))) print("Save the model to directory {}".format(self.save_dir)) def load_ensemble_weights(self, member_idx=0): if path.exists(path.join(self.save_dir, self.architecture_type + '_{}'.format(member_idx))): self.base_model = tf.keras.models.load_model( path.join(self.save_dir, self.architecture_type + '_{}'.format(member_idx))) def get_n_members(self): return self.n_members def update_weights(self, member_idx, model_weights, optimizer_weights=None): raise NotImplementedError
5,238
42.658333
110
py
malware-uncertainty
malware-uncertainty-master/core/ensemble/vanilla.py
import os.path as path import time import tensorflow as tf import numpy as np from core.ensemble.ensemble import Ensemble from core.ensemble.model_hp import train_hparam, finetuning_hparam from core.ensemble.model_lib import model_builder from tools import utils from config import logging, ErrorHandler logger = logging.getLogger('ensemble.vanilla') logger.addHandler(ErrorHandler) class Vanilla(Ensemble): """ vanilla model, i.e., the so-called ensemble just has a single model """ def __init__(self, architecture_type='dnn', base_model=None, n_members=1, model_directory=None, name='VANILLA'): """ initialization :param architecture_type: the type of base model :param base_model: an object of base model :param n_members: number of base models :param model_directory: a folder for saving ensemble weights """ super(Vanilla, self).__init__(architecture_type, base_model, n_members, model_directory) self.hparam = train_hparam self.ensemble_type = 'vanilla' self.name = name.lower() self.save_dir = path.join(self.model_directory, self.name) def build_model(self, input_dim=None): """ Build an ensemble model -- only the homogeneous structure is considered :param input_dim: integer or list, input dimension shall be set in some cases under eager mode """ callable_graph = model_builder(self.architecture_type) @callable_graph(input_dim) def _builder(): return utils.produce_layer(self.ensemble_type) self.base_model = _builder() return def predict(self, x, use_prob=False): """ conduct prediction """ self.base_model = None self.weights_list = [] self._optimizers_dict = [] self.load_ensemble_weights() output_list = [] start_time = time.time() for base_model in self.model_generator(): if isinstance(x, tf.data.Dataset): output_list.append(base_model.predict(x, verbose=1)) elif isinstance(x, (np.ndarray, list)): output_list.append(base_model.predict(x, verbose=1, batch_size=self.hparam.batch_size)) else: raise ValueError total_time = time.time() - start_time logger.info('Inference costs {} seconds.'.format(total_time)) if not use_prob: return np.stack(output_list, axis=1) else: return np.mean(np.stack(output_list, axis=1), axis=1) def evaluate(self, x, gt_labels, threshold=0.5, name='test'): """ get some statistical values :param x: tf.data.Dataset object :param gt_labels: ground truth labels :param threshold: float value between 0 and 1, to decide the predicted label :return: None """ x_prob = self.predict(x, use_prob=True) x_pred = (x_prob >= threshold).astype(np.int32) # metrics from sklearn.metrics import f1_score, accuracy_score, confusion_matrix, balanced_accuracy_score accuracy = accuracy_score(gt_labels, x_pred) b_accuracy = balanced_accuracy_score(gt_labels, x_pred) MSG = "The accuracy on the {} dataset is {:.5f}%" logger.info(MSG.format(name, accuracy * 100)) MSG = "The balanced accuracy on the {} dataset is {:.5f}%" logger.info(MSG.format(name, b_accuracy * 100)) is_single_class = False if np.all(gt_labels == 1.) or np.all(gt_labels == 0.): is_single_class = True if not is_single_class: tn, fp, fn, tp = confusion_matrix(gt_labels, x_pred).ravel() fpr = fp / float(tn + fp) fnr = fn / float(tp + fn) f1 = f1_score(gt_labels, x_pred, average='binary') print("Other evaluation metrics we may need:") MSG = "False Negative Rate (FNR) is {:.5f}%, False Positive Rate (FPR) is {:.5f}%, F1 score is {:.5f}%" logger.info(MSG.format(fnr * 100, fpr * 100, f1 * 100)) return def model_generator(self): try: if len(self.weights_list) <= 0: self.load_ensemble_weights() except Exception as e: raise Exception("Cannot load model weights:{}.".format(str(e))) for i, weights in enumerate(self.weights_list): self.base_model.set_weights(weights=weights) # if i in self._optimizers_dict and self.base_model.optimizer is not None: # self.base_model.optimizer.set_weights(self._optimizers_dict[i]) yield self.base_model def fit(self, train_set, validation_set=None, input_dim=None, **kwargs): """ fit the ensemble by producing a lists of model weights :param train_set: tf.data.Dataset, the type shall accommodate to the input format of Tensorflow models :param validation_set: validation data, optional :param input_dim: integer or list, input dimension except for the batch size """ # training preparation if self.base_model is None: self.build_model(input_dim=input_dim) self.base_model.compile( optimizer=tf.keras.optimizers.Adam(learning_rate=self.hparam.learning_rate, clipvalue=self.hparam.clipvalue), loss=tf.keras.losses.BinaryCrossentropy(), metrics=[tf.keras.metrics.BinaryAccuracy()], ) # training logger.info("hyper-parameters:") logger.info(dict(self.hparam._asdict())) logger.info("The number of trainable variables: {}".format(len(self.base_model.trainable_variables))) logger.info("...training start!") best_val_accuracy = 0. total_time = 0. for epoch in range(self.hparam.n_epochs): train_acc = 0. val_acc = 0. for member_idx in range(self.n_members): if member_idx < len(self.weights_list): # loading former weights self.base_model.set_weights(self.weights_list[member_idx]) self.base_model.optimizer.set_weights(self._optimizers_dict[member_idx]) elif member_idx == 0: pass # do nothing else: self.reinitialize_base_model() msg = 'Epoch {}/{}, member {}/{}, and {} member(s) in list'.format(epoch + 1, self.hparam.n_epochs, member_idx + 1, self.n_members, len(self.weights_list)) print(msg) start_time = time.time() history = self.base_model.fit(train_set, epochs=epoch + 1, initial_epoch=epoch, validation_data=validation_set ) train_acc += history.history['binary_accuracy'][0] val_acc += history.history['val_binary_accuracy'][0] self.update_weights(member_idx, self.base_model.get_weights(), self.base_model.optimizer.get_weights()) end_time = time.time() total_time += end_time - start_time # saving logger.info('Training ensemble costs {} seconds in total (including validation).'.format(total_time)) train_acc = train_acc / self.n_members val_acc = val_acc / self.n_members msg = 'Epoch {}/{}: training accuracy {:.5f}, validation accuracy {:.5f}.'.format( epoch + 1, self.hparam.n_epochs, train_acc, val_acc ) logger.info(msg) if (epoch + 1) % self.hparam.interval == 0: if val_acc > best_val_accuracy: self.save_ensemble_weights() best_val_accuracy = val_acc msg = '\t The best validation accuracy is {:.5f}, obtained at epoch {}/{}'.format( best_val_accuracy, epoch + 1, self.hparam.n_epochs ) logger.info(msg) return def fit_finetuning(self, train_set, validation_set=None, input_dim=None, **kwargs): """ just for experiments of r2d2 model :param train_set: tf.data.Dataset, the type shall accommodate to the input format of Tensorflow models :param validation_set: validation data, optional :param input_dim: integer or list, input dimension except for the batch size """ # training preparation if self.base_model is None: self.build_model(input_dim=input_dim) # training logger.info("hyper-parameters:") logger.info(dict(finetuning_hparam._asdict())) logger.info("...training start!") def train(n_epochs, learning_rate, ft=False): logger.info("The number of trainable variables: {}".format(len(self.base_model.trainable_variables))) optimizer = tf.keras.optimizers.Adam(learning_rate=learning_rate, clipvalue=self.hparam.clipvalue) if ft: optimizer = tf.keras.optimizers.RMSprop(learning_rate=learning_rate, clipvalue=self.hparam.clipvalue) self.base_model.compile( optimizer=optimizer, loss=tf.keras.losses.BinaryCrossentropy(), metrics=[tf.keras.metrics.BinaryAccuracy()], ) if ft: self.load_ensemble_weights() best_val_accuracy = 0. total_time = 0. for epoch in range(n_epochs): train_acc = 0. val_acc = 0. for member_idx in range(self.n_members): if member_idx < len(self.weights_list): # loading former weights self.base_model.set_weights(self.weights_list[member_idx]) if epoch > 0: self.base_model.optimizer.set_weights(self._optimizers_dict[member_idx]) elif member_idx == 0: pass # do nothing else: self.reinitialize_base_model() msg = 'Epoch {}/{}, member {}/{}, and {} member(s) in list'.format(epoch + 1, n_epochs, member_idx + 1, self.n_members, len(self.weights_list)) print(msg) start_time = time.time() history = self.base_model.fit(train_set, epochs=epoch + 1, initial_epoch=epoch, validation_data=validation_set ) train_acc += history.history['binary_accuracy'][0] val_acc += history.history['val_binary_accuracy'][0] self.update_weights(member_idx, self.base_model.get_weights(), self.base_model.optimizer.get_weights()) end_time = time.time() total_time += end_time - start_time # saving logger.info('Training ensemble costs {} in total (including validation).'.format(total_time)) train_acc = train_acc / self.n_members val_acc = val_acc / self.n_members msg = 'Epoch {}/{}: training accuracy {:.5f}, validation accuracy {:.5f}.'.format( epoch + 1, self.hparam.n_epochs, train_acc, val_acc ) logger.info(msg) if (epoch + 1) % self.hparam.interval == 0: if val_acc > best_val_accuracy: self.save_ensemble_weights() best_val_accuracy = val_acc msg = '\t The best validation accuracy is {:.5f}, obtained at epoch {}/{}'.format( best_val_accuracy, epoch + 1, self.hparam.n_epochs ) logger.info(msg) train(finetuning_hparam.n_epochs, finetuning_hparam.learning_rate) logger.info("...training finished!") # finetuning: for layer in self.base_model.layers[-finetuning_hparam.unfreezed_layers:]: layer.trainable = True logger.info("...fine-tuning start!") train(finetuning_hparam.n_epochs_ft, finetuning_hparam.learning_rate_ft, True) logger.info("...fine-tuning finished!") return def update_weights(self, member_idx, model_weights, optimizer_weights=None): if member_idx < len(self.weights_list): self.weights_list[member_idx] = model_weights self._optimizers_dict[member_idx] = optimizer_weights else: # append the weights at the rear of list assert len(self.weights_list) == len(self._optimizers_dict) self.weights_list.append(model_weights) set_idx = len(self.weights_list) - 1 self._optimizers_dict[set_idx] = optimizer_weights return def save_ensemble_weights(self): # if not path.exists(self.save_dir): # utils.mkdir(self.save_dir) # # save model configuration # try: # config = self.base_model.to_json() # utils.dump_json(config, path.join(self.save_dir, # self.architecture_type + '.json')) # lightweight method for saving model configurature # except Exception as e: # pass # finally: if not path.exists(path.join(self.save_dir, self.architecture_type)): utils.mkdir(path.join(self.save_dir, self.architecture_type)) self.base_model.save(path.join(self.save_dir, self.architecture_type)) print("Save the model configuration to directory {}".format(self.save_dir)) # save model weights utils.dump_joblib(self.weights_list, path.join(self.save_dir, self.architecture_type + '.model')) utils.dump_joblib(self._optimizers_dict, path.join(self.save_dir, self.architecture_type + '.model.metadata')) print("Save the model weights to directory {}".format(self.save_dir)) return def load_ensemble_weights(self): # if path.exists(path.join(self.save_dir, self.architecture_type + '.json')): # config = utils.load_json(path.join(self.save_dir, self.architecture_type + '.json')) # self.base_model = tf.keras.models.model_from_json(config) if path.exists(path.join(self.save_dir, self.architecture_type)): self.base_model = tf.keras.models.load_model(path.join(self.save_dir, self.architecture_type)) else: logger.error("File not found: ".format(path.join(self.save_dir, self.architecture_type + '.json'))) raise FileNotFoundError print("Load model config from {}.".format(self.save_dir)) if path.exists(path.join(self.save_dir, self.architecture_type + '.model')): self.weights_list = utils.read_joblib(path.join(self.save_dir, self.architecture_type + '.model')) else: logger.error("File not found: ".format(path.join(self.save_dir, self.architecture_type + '.model'))) raise FileNotFoundError print("Load model weights from {}.".format(self.save_dir)) if path.exists(path.join(self.save_dir, self.architecture_type + '.model.metadata')): self._optimizers_dict = utils.read_joblib( path.join(self.save_dir, self.architecture_type + '.model.metadata')) else: self._optimizers_dict = [None] * len(self.weights_list) return def get_n_members(self): return len(self.weights_list) def reinitialize_base_model(self): new_weights = [] for w in self.base_model.weights: if w.trainable: if '/kernel' in w.name: # default name new_w = utils.glorot_uniform(w.numpy().shape) elif '/recurrent_kernel' in w.name: initilizer = tf.keras.initializers.Orthogonal() new_w = initilizer(w.numpy().shape).numpy() elif '/bias' in w.name: new_w = utils.glorot_uniform(w.numpy().shape) else: new_w = utils.glorot_uniform(w.numpy().shape) else: new_w = w.numpy() new_weights.append(new_w) self.base_model.set_weights(new_weights) return def gradient_loss_wrt_input(self, x, y=None): if self.base_model is None: raise ValueError("A learned model is expected. Please try load_ensemble_weights() first") # we set y[...]=1 by default y = np.ones(shape=x.shape[0], dtype=np.int64) binary_ce = tf.losses.binary_crossentropy grad = 0. for model_fn in self.model_generator(): with tf.GradientTape() as g: g.watch(x) loss = binary_ce(y, model_fn(x)) grad += g.gradient(loss, x) return grad
17,864
46.387268
135
py
malware-uncertainty
malware-uncertainty-master/core/ensemble/bayesian_ensemble.py
import time import tensorflow as tf from core.ensemble.vanilla import model_builder from core.ensemble.mc_dropout import MCDropout from core.ensemble.model_hp import train_hparam, bayesian_ensemble_hparam from config import logging, ErrorHandler from tools import utils logger = logging.getLogger('ensemble.bayesian_ensemble') logger.addHandler(ErrorHandler) class BayesianEnsemble(MCDropout): def __init__(self, architecture_type='dnn', base_model=None, n_members=1, model_directory=None, name='BAYESIAN_ENSEMBLE' ): super(BayesianEnsemble, self).__init__(architecture_type, base_model, n_members, model_directory, name) self.hparam = utils.merge_namedtuples(train_hparam, bayesian_ensemble_hparam) self.ensemble_type = 'bayesian' def build_model(self, input_dim=None, scaler=1. / 10000): """ Build an ensemble model -- only the homogeneous structure is considered :param input_dim: integer or list, input dimension shall be set in some cases under eager mode :param scaler: float value in the rage of [0, 1], weighted kl divergence """ callable_graph = model_builder(self.architecture_type) @callable_graph(input_dim) def _builder(): return utils.produce_layer(self.ensemble_type, kl_scaler=scaler) self.base_model = _builder() return def fit(self, train_set, validation_set=None, input_dim=None, **kwargs): """ fit the ensemble by producing a lists of model weights :param train_set: tf.data.Dataset, the type shall accommodate to the input format of Tensorflow models :param validation_set: validation data, optional :param input_dim: integer or list, input dimension except for the batch size """ # training preparation if self.base_model is None: # scaler = 1. / (len(list(train_set)) * self.hparam.batch_size) # time-consuming scaler = 1. / 50000. self.build_model(input_dim=input_dim, scaler=scaler) self.base_model.compile( optimizer=tf.keras.optimizers.Adam(learning_rate=self.hparam.learning_rate, clipvalue=self.hparam.clipvalue), loss=tf.keras.losses.BinaryCrossentropy(), metrics=[tf.keras.metrics.BinaryAccuracy()], experimental_run_tf_function=False ) # training logger.info("hyper-parameters:") logger.info(dict(self.hparam._asdict())) logger.info("...training start!") best_val_accuracy = 0. total_time = 0. for epoch in range(self.hparam.n_epochs): train_acc = 0. val_acc = 0. for member_idx in range(self.n_members): if member_idx < len(self.weights_list): # loading former weights self.base_model.set_weights(self.weights_list[member_idx]) self.base_model.optimizer.set_weights(self._optimizers_dict[member_idx]) elif member_idx == 0: pass # do nothing else: self.reinitialize_base_model() msg = 'Epoch {}/{}, member {}/{}, and {} member(s) in list'.format(epoch + 1, self.hparam.n_epochs, member_idx + 1, self.n_members, len(self.weights_list)) print(msg) start_time = time.time() history = self.base_model.fit(train_set, epochs=epoch + 1, initial_epoch=epoch, validation_data=validation_set ) train_acc += history.history['binary_accuracy'][0] val_acc += history.history['val_binary_accuracy'][0] self.update_weights(member_idx, self.base_model.get_weights(), self.base_model.optimizer.get_weights()) end_time = time.time() total_time += end_time - start_time # saving logger.info('Training ensemble costs {} in total (including validation).'.format(total_time)) train_acc = train_acc / self.n_members val_acc = val_acc / self.n_members msg = 'Epoch {}/{}: training accuracy {:.5f}, validation accuracy {:.5f}.'.format( epoch + 1, self.hparam.n_epochs, train_acc, val_acc ) logger.info(msg) if (epoch + 1) % self.hparam.interval == 0: if val_acc >= best_val_accuracy: self.save_ensemble_weights() best_val_accuracy = val_acc msg = '\t The best validation accuracy is {:.5f}, obtained at epoch {}/{}'.format( best_val_accuracy, epoch + 1, self.hparam.n_epochs ) logger.info(msg) return
5,667
44.709677
110
py
malware-uncertainty
malware-uncertainty-master/core/ensemble/model_lib.py
""" This script is for building model graph""" import tensorflow as tf from tools import utils from config import logging, ErrorHandler logger = logging.getLogger('core.ensemble.model_lib') logger.addHandler(ErrorHandler) def model_builder(architecture_type='dnn'): assert architecture_type in model_name_type_dict, 'models are {}'.format(','.join(model_name_type_dict.keys())) return model_name_type_dict[architecture_type] def _change_scaler_to_list(scaler): if not isinstance(scaler, (list, tuple)): return [scaler] else: return scaler def _dnn_graph(input_dim=None, use_mc_dropout=False): """ The deep neural network based malware detector. The implement is based on the paper, entitled ``Adversarial Examples for Malware Detection'', which can be found here: http://patrickmcdaniel.org/pubs/esorics17.pdf We slightly change the model architecture by reducing the number of neurons at the last layer to one. """ input_dim = _change_scaler_to_list(input_dim) from core.ensemble.model_hp import dnn_hparam logger.info(dict(dnn_hparam._asdict())) def wrapper(func): def graph(): Dense, _1, _2, _3 = func() model = tf.keras.Sequential() model.add(tf.keras.layers.InputLayer(input_shape=(input_dim[0],))) for units in dnn_hparam.hidden_units: model.add(Dense(units, activation=dnn_hparam.activation)) if use_mc_dropout: model.add(tf.keras.layers.Dense(dnn_hparam.output_dim, activation=tf.nn.sigmoid)) else: model.add(tf.keras.layers.Dropout(dnn_hparam.dropout_rate)) model.add(Dense(dnn_hparam.output_dim, activation=tf.nn.sigmoid)) return model return graph return wrapper def _text_cnn_graph(input_dim=None, use_mc_dropout=False): """ deep android malware detection The implement is based on the paper, entitled ``Deep Android Malware Detection'', which can be found here: https://dl.acm.org/doi/10.1145/3029806.3029823 """ input_dim = _change_scaler_to_list(input_dim) # dynamical input shape is permitted from core.ensemble.model_hp import text_cnn_hparam logger.info(dict(text_cnn_hparam._asdict())) def wrapper(func): def graph(): Dense, Conv2D, _1, _2 = func() class TextCNN(tf.keras.models.Model): def __init__(self): super(TextCNN, self).__init__() self.embedding = tf.keras.layers.Embedding(text_cnn_hparam.vocab_size, text_cnn_hparam.n_embedding_dim) self.spatial_dropout = tf.keras.layers.SpatialDropout2D(rate=text_cnn_hparam.dropout_rate) self.conv = Conv2D(text_cnn_hparam.n_conv_filters, text_cnn_hparam.kernel_size, activation=text_cnn_hparam.activation) self.conv_dropout = tf.keras.layers.Dropout(rate=text_cnn_hparam.dropout_rate) self.pooling = tf.keras.layers.GlobalMaxPool2D() # produce a fixed length vector self.denses = [Dense(neurons, activation='relu') for neurons in text_cnn_hparam.hidden_units] self.dropout = tf.keras.layers.Dropout(text_cnn_hparam.dropout_rate) if use_mc_dropout: self.d_out = tf.keras.layers.Dense(text_cnn_hparam.output_dim, activation=tf.nn.sigmoid) else: self.d_out = Dense(text_cnn_hparam.output_dim, activation=tf.nn.sigmoid) def call(self, x, training=False): embed_code = self.embedding(x) # batch_size, seq_length, embedding_dim, 1. Note: seq_length >= conv_kernel_size embed_code = tf.expand_dims(embed_code, axis=-1) if text_cnn_hparam.use_spatial_dropout: embed_code = self.spatial_dropout(embed_code, training=training) conv_x = self.conv(embed_code) if text_cnn_hparam.use_conv_dropout: conv_x = self.conv_dropout(conv_x) flatten_x = self.pooling(conv_x) for i, dense in enumerate(self.denses): flatten_x = dense(flatten_x) if not use_mc_dropout: flatten_x = self.dropout(flatten_x, training=training) return self.d_out(flatten_x) return TextCNN() return graph return wrapper def _multimodalitynn(input_dim=None, use_mc_dropout=False): """ A Multimodal Deep Learning Method for Android Malware Detection Using Various Features The implement is based on our understanding of the paper, entitled ``A Multimodal Deep Learning Method for Android Malware Detection Using Various Features'': @ARTICLE{8443370, author={T. {Kim} and B. {Kang} and M. {Rho} and S. {Sezer} and E. G. {Im}}, journal={IEEE Transactions on Information Forensics and Security}, title={A Multimodal Deep Learning Method for Android Malware Detection Using Various Features}, year={2019}, volume={14}, number={3}, pages={773-788},} """ input_dim = _change_scaler_to_list(input_dim) assert isinstance(input_dim, (list, tuple)), 'a list of input dimensions are mandatory.' from core.ensemble.model_hp import multimodalitynn_hparam assert len(input_dim) == multimodalitynn_hparam.n_modalities, 'Expected input number {}, but got {}'.format( multimodalitynn_hparam.n_modalities, len(input_dim)) logger.info(dict(multimodalitynn_hparam._asdict())) def wrapper(func): def graph(): input_layers = [] Dense, _1, _2, _3 = func() for idx, header in enumerate(range(multimodalitynn_hparam.n_modalities)): input_layers.append( tf.keras.Input(input_dim[idx], name='HEADER_{}'.format(idx + 1)) ) x_initial_out = [] for x in input_layers: for units in multimodalitynn_hparam.initial_hidden_units: x = Dense(units, activation=multimodalitynn_hparam.activation)(x) x_initial_out.append(x) x_out = tf.keras.layers.concatenate(x_initial_out) for units in multimodalitynn_hparam.hidden_units: x_out = Dense(units, activation=multimodalitynn_hparam.activation)(x_out) if use_mc_dropout: out = tf.keras.layers.Dense(multimodalitynn_hparam.output_dim, activation=tf.nn.sigmoid)(x_out) else: out = tf.keras.layers.Dropout(rate=multimodalitynn_hparam.dropout_rate)(x_out) out = Dense(multimodalitynn_hparam.output_dim, activation=tf.nn.sigmoid)(out) return tf.keras.Model(inputs=input_layers, outputs=out) return graph return wrapper def _r2d2(input_dim=None, use_mc_dropout=False): """ R2-D2: ColoR-inspired Convolutional NeuRal Network (CNN)-based AndroiD Malware Detections The implement is based on our understanding of the paper, entitled ``R2-D2: ColoR-inspired Convolutional NeuRal Network (CNN)-based AndroiD Malware Detections'': @INPROCEEDINGS{8622324, author={T. H. {Huang} and H. {Kao}}, booktitle={2018 IEEE International Conference on Big Data (Big Data)}, title={R2-D2: ColoR-inspired Convolutional NeuRal Network (CNN)-based AndroiD Malware Detections}, year={2018}, volume={}, number={}, pages={2633-2642},} """ input_dim = _change_scaler_to_list(input_dim) from core.ensemble.model_hp import r2d2_hparam logger.info(dict(r2d2_hparam._asdict())) def wrapper(func): def graph(): Dense, _1, _2, last_Dense = func() base_model = tf.keras.applications.MobileNetV2(input_shape=input_dim, include_top=False, weights='imagenet') base_model.trainable = False for layer in base_model.layers[-r2d2_hparam.unfreezed_layers:]: layer.trainable = True x_new = base_model.layers[-1].output x_new = tf.keras.layers.GlobalAveragePooling2D()(x_new) if use_mc_dropout: # x_new = tf.nn.dropout(x_new, rate=mc_droput_rate) # out = tf.keras.layers.Dense(r2d2_hparam.output_dim, activation=tf.nn.sigmoid)(x_new) out = last_Dense(r2d2_hparam.output_dim, activation=tf.nn.sigmoid)(x_new) else: x_new = tf.keras.layers.Dropout(r2d2_hparam.dropout_rate)(x_new) out = Dense(r2d2_hparam.output_dim, activation=tf.nn.sigmoid)(x_new) return tf.keras.Model(inputs=base_model.input, outputs=out) return graph return wrapper def _droidectc_graph(input_dim=None, use_mc_dropout=False): """ DROIDETEC: Android Malware Detection and Malicious Code Localization through Deep Learning The implement is based on our understanding of the paper, entitled ``DROIDETEC: Android Malware Detection and Malicious Code Localization through Deep Learning'': @article{ma2020droidetec, title={Droidetec: Android malware detection and malicious code localization through deep learning}, author={Ma, Zhuo and Ge, Haoran and Wang, Zhuzhu and Liu, Yang and Liu, Ximeng}, journal={arXiv preprint arXiv:2002.03594}, year={2020} } """ input_dim = _change_scaler_to_list(input_dim) # dynamic input shape is permitted from core.ensemble.model_hp import droidetec_hparam logger.info(dict(droidetec_hparam._asdict())) def wrapper(func): def graph(): Dense, _1, LSTM, last_Dense = func() class BiLSTMAttention(tf.keras.models.Model): def __init__(self): super(BiLSTMAttention, self).__init__() self.embedding = tf.keras.layers.Embedding(droidetec_hparam.vocab_size, droidetec_hparam.n_embedding_dim) self.bi_lstm = tf.keras.layers.Bidirectional(LSTM(droidetec_hparam.lstm_units, return_sequences=True), merge_mode='sum' ) self.dense_layer = tf.keras.layers.Dense(droidetec_hparam.lstm_units, use_bias=False) # for units in droidetec_hparam.hidden_units: # self.dense_layers.append(Dense(droidetec_hparam.hidden_units, use_bias=False)) if use_mc_dropout: self.output_layer = last_Dense(droidetec_hparam.output_dim, activation=tf.nn.sigmoid) else: self.output_layer = Dense(droidetec_hparam.output_dim, activation=tf.nn.sigmoid) def call(self, x, training=False): embed_x = self.embedding(x) # if use_mc_dropout: # stateful_x = self.bi_lstm(embed_x, training=True) # else: stateful_x = self.bi_lstm(embed_x) alpha_wights = tf.nn.softmax(self.dense_layer(tf.nn.tanh(stateful_x)), axis=1) attn_x = tf.reduce_sum(alpha_wights * stateful_x, axis=1) # if use_mc_dropout: # attn_x = tf.nn.dropout(attn_x, rate=mc_dropout_rate) return self.output_layer(attn_x) return BiLSTMAttention() return graph return wrapper model_name_type_dict = { 'dnn': _dnn_graph, 'text_cnn': _text_cnn_graph, 'multimodalitynn': _multimodalitynn, 'r2d2': _r2d2, 'droidectc': _droidectc_graph } def build_models(input_x, architecture_type, ensemble_type='vanilla', input_dim=None, use_mc_dropout=False): builder = model_builder(architecture_type) @builder(input_dim, use_mc_dropout) def graph(): return utils.produce_layer(ensemble_type, dropout_rate=0.4) model = graph() return model(input_x)
12,509
42.741259
115
py
diaparser
diaparser-master/setup.py
# -*- coding: utf-8 -*- from setuptools import find_packages, setup setup( name='diaparser', version='1.1.3', author='Yu Zhang, Giuseppe Attardi', author_email='yzhang.cs@outlook.com, attardi@di.unipi.it', description='Direct Attentive Dependency Parser', long_description=open('README.md', 'r').read(), long_description_content_type='text/markdown', url='https://github.com/Unipisa/diaparser', packages=find_packages(), classifiers=[ 'Programming Language :: Python :: 3', 'License :: OSI Approved :: MIT License', 'Operating System :: OS Independent', 'Topic :: Scientific/Engineering :: Artificial Intelligence', 'Topic :: Text Processing :: Linguistic' ], setup_requires=[ 'setuptools>=18.0', ], # stanza 1.3 has incompatible changes (attribute feat_dropout instead of dropout) install_requires=['torch>=2.0', 'transformers', 'nltk', 'stanza', 'numpy'], entry_points={ 'console_scripts': [ 'diaparser=diaparser.cmds.biaffine_dependency:main', ] }, python_requires='>=3.6', zip_safe=False )
1,148
31.828571
85
py
diaparser
diaparser-master/diaparser/modules/bert.py
# -*- coding: utf-8 -*- import torch import torch.nn as nn from torch.nn.utils.rnn import pad_sequence from transformers import AutoModel, AutoConfig from .scalar_mix import ScalarMix from .dropout import TokenDropout # from torch.cuda import memory_allocated class BertEmbedding(nn.Module): r""" A module that directly utilizes the pretrained models in `transformers`_ to produce BERT representations. While mainly tailored to provide input preparation and post-processing for the BERT model, it is also compatible with other pretrained language models like XLNet, RoBERTa and ELECTRA, etc. Args: model (str): Path or name of the pretrained models registered in `transformers`_, e.g., ``'bert-base-cased'``. n_layers (int): The number of layers from the model to use. If 0, uses all layers. n_out (int): The requested size of the embeddings. If 0, uses the size of the pretrained embedding model. stride (int): A sequence longer than the limited max length will be splitted into several small pieces with a window size of ``stride``. Default: 5. pad_index (int): The index of the padding token in the BERT vocabulary. Default: 0. dropout (float): The dropout ratio of BERT layers. Default: 0. This value will be passed into the :class:`ScalarMix` layer. requires_grad (bool): If ``True``, the model parameters will be updated together with the downstream task. Default: ``False``. .. _transformers: https://github.com/huggingface/transformers """ def __init__(self, model, n_layers, n_out, stride=5, pad_index=0, dropout=0, requires_grad=False, mask_token_id=0, token_dropout=0.0, mix_dropout=0.0, use_hidden_states=True, use_attentions=False, attention_head=0, attention_layer=8): r""" :param model (str): path or name of the pretrained model. :param n_layers (int): number of layers from the model to use. If 0, use all layers. :param n_out (int): the requested size of the embeddings. If 0, use the size of the pretrained embedding model :param requires_grad (bool): whether to fine tune the embeddings. :param mask_token_id (int): the value of the [MASK] token to use for dropped tokens. :param dropout (float): drop layers with this probability when comuting their weighted average with ScalarMix. :param use_hidden_states (bool): use the output hidden states from bert if True, or else the outputs. :param use_attentions (bool): wheth4er to use attention weights. :param attention_head (int): which attention head to use. :param attention_layer (int): which attention layer to use. """ super().__init__() config = AutoConfig.from_pretrained(model, output_hidden_states=True, output_attentions=use_attentions) self.bert = AutoModel.from_pretrained(model, config=config) self.bert.requires_grad_(requires_grad) self.model = model self.n_layers = n_layers or self.bert.config.num_hidden_layers self.hidden_size = self.bert.config.hidden_size self.n_out = n_out or self.hidden_size self.stride = stride self.pad_index = self.bert.config.pad_token_id self.mix_dropout = mix_dropout self.token_dropout = token_dropout self.requires_grad = requires_grad self.max_len = self.bert.config.max_position_embeddings self.use_hidden_states = use_hidden_states self.mask_token_id = mask_token_id self.use_attentions = use_attentions self.head = attention_head self.attention_layer = attention_layer self.token_dropout = TokenDropout(token_dropout, mask_token_id) if token_dropout else None self.scalar_mix = ScalarMix(self.n_layers, mix_dropout) if self.hidden_size != self.n_out: self.projection = nn.Linear(self.hidden_size, self.n_out, False) def __repr__(self): s = f"{self.model}, n_layers={self.n_layers}, n_out={self.n_out}" s += f", pad_index={self.pad_index}" s += f", max_len={self.max_len}" if self.mix_dropout > 0: s += f", mix_dropout={self.mix_dropout}" if self.use_attentions: s += f", use_attentions={self.use_attentions}" if self.hidden_size != self.n_out: s += f", projection=({self.hidden_size} x {self.n_out})" s += f", mask_token_id={self.mask_token_id}" if self.requires_grad: s += f", requires_grad={self.requires_grad}" return f"{self.__class__.__name__}({s})" def forward(self, subwords): r""" Args: subwords (~torch.Tensor): ``[batch_size, seq_len, fix_len]``. Returns: ~torch.Tensor: BERT embeddings of shape ``[batch_size, seq_len, n_out]``. """ batch_size, seq_len, fix_len = subwords.shape if self.token_dropout: subwords = self.token_dropout(subwords) mask = subwords.ne(self.pad_index) lens = mask.sum((1, 2)) if not self.requires_grad: self.bert.eval() # CHECK_ME (supar does not do) # [batch_size, n_subwords] subwords = pad_sequence(subwords[mask].split(lens.tolist()), True) bert_mask = pad_sequence(mask[mask].split(lens.tolist()), True) # Outputs from the transformer: # - last_hidden_state: [batch, seq_len, hidden_size] # - pooler_output: [batch, hidden_size], # - hidden_states (optional): [[batch_size, seq_length, hidden_size]] * (1 + layers) # - attentions (optional): [[batch_size, num_heads, seq_length, seq_length]] * layers # print('<BERT, GPU MiB:', memory_allocated() // (1024*1024)) # DEBUG outputs = self.bert(subwords[:, :self.max_len], attention_mask=bert_mask[:, :self.max_len].float()) # print('BERT>, GPU MiB:', memory_allocated() // (1024*1024)) # DEBUG if self.use_hidden_states: bert_idx = -2 if self.use_attentions else -1 bert = outputs[bert_idx] # [n_layers, batch_size, n_subwords, hidden_size] bert = bert[-self.n_layers:] # [batch_size, n_subwords, hidden_size] bert = self.scalar_mix(bert) for i in range(self.stride, (subwords.shape[1]-self.max_len+self.stride-1)//self.stride*self.stride+1, self.stride): part = self.bert(subwords[:, i:i+self.max_len], attention_mask=bert_mask[:, i:i+self.max_len].float())[bert_idx] bert = torch.cat((bert, self.scalar_mix(part[-self.n_layers:])[:, self.max_len-self.stride:]), 1) else: bert = outputs[0] # [batch_size, n_subwords] bert_lens = mask.sum(-1) bert_lens = bert_lens.masked_fill_(bert_lens.eq(0), 1) # [batch_size, seq_len, fix_len, hidden_size] embed = bert.new_zeros(*mask.shape, self.hidden_size) embed = embed.masked_scatter_(mask.unsqueeze(-1), bert[bert_mask]) # [batch_size, seq_len, hidden_size] embed = embed.sum(2) / bert_lens.unsqueeze(-1) # sum wordpieces seq_attn = None if self.use_attentions: # (a list of layers) = [ [batch, num_heads, sent_len, sent_len] ] attns = outputs[-1] # [batch, n_subwords, n_subwords] attn = attns[self.attention_layer][:,self.head,:,:] # layer 9 represents syntax # squeeze out multiword tokens mask2 = ~mask mask2[:,:,0] = True # keep first column sub_masks = pad_sequence(mask2[mask].split(lens.tolist()), True) seq_mask = torch.einsum('bi,bj->bij', sub_masks, sub_masks) # outer product seq_lens = seq_mask.sum((1,2)) # [batch_size, seq_len, seq_len] sub_attn = attn[seq_mask].split(seq_lens.tolist()) # fill a tensor [batch_size, seq_len, seq_len] seq_attn = attn.new_zeros(batch_size, seq_len, seq_len) for i, attn_i in enumerate(sub_attn): size = sub_masks[i].sum(0) attn_i = attn_i.view(size, size) seq_attn[i,:size,:size] = attn_i if hasattr(self, 'projection'): embed = self.projection(embed) return embed, seq_attn
8,608
46.563536
128
py
diaparser
diaparser-master/diaparser/modules/lstm.py
# -*- coding: utf-8 -*- import torch import torch.nn as nn from .dropout import SharedDropout from torch.nn.modules.rnn import apply_permutation from torch.nn.utils.rnn import PackedSequence class LSTM(nn.Module): r""" LSTM is an variant of the vanilla bidirectional LSTM adopted by Biaffine Parser with the only difference of the dropout strategy. It drops nodes in the LSTM layers (input and recurrent connections) and applies the same dropout mask at every recurrent timesteps. APIs are roughly the same as :class:`~torch.nn.LSTM` except that we only allows :class:`~torch.nn.utils.rnn.PackedSequence` as input. References: - Timothy Dozat and Christopher D. Manning. 2017. `Deep Biaffine Attention for Neural Dependency Parsing`_. Args: input_size (int): The number of expected features in the input. hidden_size (int): The number of features in the hidden state `h`. num_layers (int): The number of recurrent layers. Default: 1. bidirectional (bool): If ``True``, becomes a bidirectional LSTM. Default: ``False`` dropout (float): If non-zero, introduces a :class:`SharedDropout` layer on the outputs of each LSTM layer except the last layer. Default: 0. .. _Deep Biaffine Attention for Neural Dependency Parsing: https://openreview.net/forum?id=Hk95PK9le """ def __init__(self, input_size, hidden_size, num_layers=1, bidirectional=False, dropout=0): super().__init__() self.input_size = input_size self.hidden_size = hidden_size self.num_layers = num_layers self.bidirectional = bidirectional self.dropout = dropout self.f_cells = nn.ModuleList() if bidirectional: self.b_cells = nn.ModuleList() for _ in range(self.num_layers): self.f_cells.append(nn.LSTMCell(input_size=input_size, hidden_size=hidden_size)) if bidirectional: self.b_cells.append(nn.LSTMCell(input_size=input_size, hidden_size=hidden_size)) input_size = hidden_size * (1 + self.bidirectional) self.reset_parameters() def __repr__(self): s = f"{self.input_size}, {self.hidden_size}" if self.num_layers > 1: s += f", num_layers={self.num_layers}" if self.bidirectional: s += f", bidirectional={self.bidirectional}" if self.dropout > 0: s += f", dropout={self.dropout}" return f"{self.__class__.__name__}({s})" def reset_parameters(self): for param in self.parameters(): # apply orthogonal_ to weight if len(param.shape) > 1: nn.init.orthogonal_(param) # apply zeros_ to bias else: nn.init.zeros_(param) def permute_hidden(self, hx, permutation): if permutation is None: return hx h = apply_permutation(hx[0], permutation) c = apply_permutation(hx[1], permutation) return h, c def layer_forward(self, x, hx, cell, batch_sizes, reverse=False): hx_0 = hx_i = hx hx_n, output = [], [] steps = reversed(range(len(x))) if reverse else range(len(x)) if self.training: hid_mask = SharedDropout.get_mask(hx_0[0], self.dropout) for t in steps: last_batch_size, batch_size = len(hx_i[0]), batch_sizes[t] if last_batch_size < batch_size: hx_i = [torch.cat((h, ih[last_batch_size:batch_size])) for h, ih in zip(hx_i, hx_0)] else: hx_n.append([h[batch_size:] for h in hx_i]) hx_i = [h[:batch_size] for h in hx_i] hx_i = [h for h in cell(x[t], hx_i)] output.append(hx_i[0]) if self.training: hx_i[0] = hx_i[0] * hid_mask[:batch_size] if reverse: hx_n = hx_i output.reverse() else: hx_n.append(hx_i) hx_n = [torch.cat(h) for h in zip(*reversed(hx_n))] output = torch.cat(output) return output, hx_n def forward(self, sequence, hx=None): r""" Args: sequence (~torch.nn.utils.rnn.PackedSequence): A packed variable length sequence. hx (~torch.Tensor, ~torch.Tensor): A tuple composed of two tensors `h` and `c`. `h` of shape ``[num_layers*num_directions, batch_size, hidden_size]`` holds the initial hidden state for each element in the batch. `c` of shape ``[num_layers*num_directions, batch_size, hidden_size]`` holds the initial cell state for each element in the batch. If `hx` is not provided, both `h` and `c` default to zero. Default: ``None``. Returns: ~torch.nn.utils.rnn.PackedSequence, (~torch.Tensor, ~torch.Tensor): The first is a packed variable length sequence. The second is a tuple of tensors `h` and `c`. `h` of shape ``[num_layers*num_directions, batch_size, hidden_size]`` holds the hidden state for `t=seq_len`. Like output, the layers can be separated using ``h.view(num_layers, 2, batch_size, hidden_size)`` and similarly for c. `c` of shape ``[num_layers*num_directions, batch_size, hidden_size]`` holds the cell state for `t=seq_len`. """ x, batch_sizes = sequence.data, sequence.batch_sizes.tolist() batch_size = batch_sizes[0] h_n, c_n = [], [] if hx is None: ih = x.new_zeros(self.num_layers * 2, batch_size, self.hidden_size) h, c = ih, ih else: h, c = self.permute_hidden(hx, sequence.sorted_indices) h = h.view(self.num_layers, 2, batch_size, self.hidden_size) c = c.view(self.num_layers, 2, batch_size, self.hidden_size) for i in range(self.num_layers): x = torch.split(x, batch_sizes) if self.training: mask = SharedDropout.get_mask(x[0], self.dropout) x = [i * mask[:len(i)] for i in x] x_i, (h_i, c_i) = self.layer_forward(x=x, hx=(h[i, 0], c[i, 0]), cell=self.f_cells[i], batch_sizes=batch_sizes) if self.bidirectional: x_b, (h_b, c_b) = self.layer_forward(x=x, hx=(h[i, 1], c[i, 1]), cell=self.b_cells[i], batch_sizes=batch_sizes, reverse=True) x_i = torch.cat((x_i, x_b), -1) h_i = torch.stack((h_i, h_b)) c_i = torch.stack((c_i, c_b)) x = x_i h_n.append(h_i) c_n.append(h_i) x = PackedSequence(x, sequence.batch_sizes, sequence.sorted_indices, sequence.unsorted_indices) hx = torch.cat(h_n, 0), torch.cat(c_n, 0) hx = self.permute_hidden(hx, sequence.unsorted_indices) return x, hx
7,586
39.790323
125
py
diaparser
diaparser-master/diaparser/modules/mlp.py
# -*- coding: utf-8 -*- import torch.nn as nn from ..modules.dropout import SharedDropout class MLP(nn.Module): r""" Applies a linear transformation together with :class:`~torch.nn.LeakyReLU` activation to the incoming tensor: :math:`y = \mathrm{LeakyReLU}(x A^T + b)` Args: n_in (~torch.Tensor): The size of each input feature. n_out (~torch.Tensor): The size of each output feature. dropout (float): If non-zero, introduce a :class:`SharedDropout` layer on the output with this dropout ratio. Default: 0. """ def __init__(self, n_in, n_out, dropout=0): super().__init__() self.n_in = n_in self.n_out = n_out self.linear = nn.Linear(n_in, n_out) self.activation = nn.LeakyReLU(negative_slope=0.1) self.dropout = SharedDropout(p=dropout) self.reset_parameters() def __repr__(self): s = f"n_in={self.n_in}, n_out={self.n_out}" if self.dropout.p > 0: s += f", dropout={self.dropout.p}" return f"{self.__class__.__name__}({s})" def reset_parameters(self): nn.init.orthogonal_(self.linear.weight) nn.init.zeros_(self.linear.bias) def forward(self, x): r""" Args: x (~torch.Tensor): The size of each input feature is `n_in`. Returns: A tensor with the size of each output feature `n_out`. """ x = self.linear(x) x = self.activation(x) x = self.dropout(x) return x
1,581
26.275862
116
py
diaparser
diaparser-master/diaparser/modules/affine.py
# -*- coding: utf-8 -*- import torch import torch.nn as nn class Biaffine(nn.Module): def __init__(self, n_in, n_out=1, bias_x=True, bias_y=True): super(Biaffine, self).__init__() self.n_in = n_in self.n_out = n_out self.bias_x = bias_x self.bias_y = bias_y self.weight = nn.Parameter(torch.Tensor(n_out, n_in + bias_x, n_in + bias_y)) self.reset_parameters() def extra_repr(self): s = f"n_in={self.n_in}, n_out={self.n_out}" if self.bias_x: s += f", bias_x={self.bias_x}" if self.bias_y: s += f", bias_y={self.bias_y}" return s def reset_parameters(self): nn.init.zeros_(self.weight) def forward(self, x, y): if self.bias_x: x = torch.cat((x, torch.ones_like(x[..., :1])), -1) if self.bias_y: y = torch.cat((y, torch.ones_like(y[..., :1])), -1) # [batch_size, n_out, seq_len, seq_len] s = torch.einsum('bxi,oij,byj->boxy', x, self.weight, y) # remove dim 1 if n_out == 1 s = s.squeeze(1) return s
1,220
26.75
64
py
diaparser
diaparser-master/diaparser/modules/char_lstm.py
# -*- coding: utf-8 -*- import torch import torch.nn as nn from torch.nn.utils.rnn import pack_padded_sequence class CharLSTM(nn.Module): r""" CharLSTM aims to generate character-level embeddings for tokens. It summerizes the information of characters in each token to an embedding using a LSTM layer. Args: n_char (int): The number of characters. n_embed (int): The size of each embedding vector as input to LSTM. n_out (int): The size of each output vector. pad_index (int): The index of the padding token in the vocabulary. Default: 0. """ def __init__(self, n_chars, n_word_embed, n_out, pad_index=0): super().__init__() self.n_chars = n_chars self.n_word_embed = n_word_embed self.n_out = n_out self.pad_index = pad_index # the embedding layer self.embed = nn.Embedding(num_embeddings=n_chars, embedding_dim=n_word_embed) # the lstm layer self.lstm = nn.LSTM(input_size=n_word_embed, hidden_size=n_out//2, batch_first=True, bidirectional=True) def __repr__(self): return f"{self.__class__.__name__}({self.n_chars}, {self.n_embed}, n_out={self.n_out}, pad_index={self.pad_index})" def forward(self, x): r""" Args: x (~torch.Tensor): ``[batch_size, seq_len, fix_len]``. Characters of all tokens. Each token holds no more than `fix_len` characters, and the excess is cut off directly. Returns: ~torch.Tensor: The embeddings of shape ``[batch_size, seq_len, n_out]`` derived from the characters. """ # [batch_size, seq_len, fix_len] mask = x.ne(self.pad_index) # [batch_size, seq_len] lens = mask.sum(-1) char_mask = lens.gt(0) # [n, fix_len, n_word_embed] x = self.embed(x[char_mask]) x = pack_padded_sequence(x, lens[char_mask], True, False) x, (h, _) = self.lstm(x) # [n, fix_len, n_out] h = torch.cat(torch.unbind(h), dim=-1) # [batch_size, seq_len, n_out] embed = h.new_zeros(*lens.shape, self.n_out) embed = embed.masked_scatter_(char_mask.unsqueeze(-1), h) return embed
2,424
32.680556
123
py
diaparser
diaparser-master/diaparser/modules/dropout.py
# -*- coding: utf-8 -*- import torch import torch.nn as nn class SharedDropout(nn.Module): r""" SharedDropout differs from the vanilla dropout strategy in that the dropout mask is shared across one dimension. Args: p (float): The probability of an element to be zeroed. Default: 0.5. batch_first (bool): If ``True``, the input and output tensors are provided as ``[batch_size, seq_len, *]``. Default: ``True``. Examples: >>> x = torch.ones(1, 3, 5) >>> nn.Dropout()(x) tensor([[[0., 2., 2., 0., 0.], [2., 2., 0., 2., 2.], [2., 2., 2., 2., 0.]]]) >>> SharedDropout()(x) tensor([[[2., 0., 2., 0., 2.], [2., 0., 2., 0., 2.], [2., 0., 2., 0., 2.]]]) """ def __init__(self, p=0.5, batch_first=True): super().__init__() self.p = p self.batch_first = batch_first def __repr__(self): s = f"p={self.p}" if self.batch_first: s += f", batch_first={self.batch_first}" return f"{self.__class__.__name__}({s})" def forward(self, x): r""" Args: x (~torch.Tensor): A tensor of any shape. Returns: The returned tensor is of the same shape as `x`. """ if self.training: if self.batch_first: mask = self.get_mask(x[:, 0], self.p).unsqueeze(1) else: mask = self.get_mask(x[0], self.p) x *= mask return x @staticmethod def get_mask(x, p): return x.new_empty(x.shape).bernoulli_(1 - p) / (1 - p) class IndependentDropout(nn.Module): r""" For :math:`N` tensors, they use different dropout masks respectively. When :math:`N-M` of them are dropped, the remaining :math:`M` ones are scaled by a factor of :math:`N/M` to compensate, and when all of them are dropped together, zeros are returned. Args: p (float): The probability of an element to be zeroed. Default: 0.5. Examples: >>> x, y = torch.ones(1, 3, 5), torch.ones(1, 3, 5) >>> x, y = IndependentDropout()(x, y) >>> x tensor([[[1., 1., 1., 1., 1.], [0., 0., 0., 0., 0.], [2., 2., 2., 2., 2.]]]) >>> y tensor([[[1., 1., 1., 1., 1.], [2., 2., 2., 2., 2.], [0., 0., 0., 0., 0.]]]) """ def __init__(self, p=0.5): super().__init__() self.p = p def __repr__(self): return f"{self.__class__.__name__}(p={self.p})" def forward(self, *items): r""" Args: items (list[~torch.Tensor]): A list of tensors that have the same shape except the last dimension. Returns: The returned tensors are of the same shape as `items`. """ if self.training: masks = [x.new_empty(x.shape[:2]).bernoulli_(1 - self.p) for x in items] total = sum(masks) scale = len(items) / total.max(torch.ones_like(total)) masks = [mask * scale for mask in masks] items = [item * mask.unsqueeze(dim=-1) for item, mask in zip(items, masks)] return items class TokenDropout(nn.Module): def __init__(self, p=0.5, value=0): super(TokenDropout, self).__init__() self.p = p self.value = value def extra_repr(self): return f"p={self.p}, value={self.value}" def forward(self, x): if self.training: mask = torch.rand_like(x, dtype=torch.float) < self.p x.masked_fill_(mask, self.value) return x
3,804
26.977941
123
py
diaparser
diaparser-master/diaparser/modules/matrix_tree_theorem.py
# -*- coding: utf-8 -*- import torch import torch.autograd as autograd import torch.nn as nn class MatrixTreeTheorem(nn.Module): def __init__(self, *args, **kwargs): super(MatrixTreeTheorem, self).__init__(*args, **kwargs) @torch.enable_grad() def forward(self, scores, mask, target=None): scores = scores.double() mask = mask.index_fill(1, mask.new_tensor(0).long(), 1) A = scores.requires_grad_().exp() A = A * mask.unsqueeze(1) * mask.unsqueeze(-1) batch_size, seq_len, _ = A.shape # D is the weighted degree matrix D = torch.zeros_like(A) D.diagonal(0, 1, 2).copy_(A.sum(-1)) # Laplacian matrix L = nn.init.eye_(torch.empty_like(A[0])).repeat(batch_size, 1, 1) L[mask] = (D - A)[mask] # calculate the partition (a.k.a normalization) term logZ = L[:, 1:, 1:].slogdet()[1].sum() mask = mask.index_fill(1, mask.new_tensor(0).long(), 0) # calculate the marginal probablities probs, = autograd.grad(logZ, scores, retain_graph=scores.requires_grad) probs = probs.float() if target is None: return probs score = scores.gather(-1, target.unsqueeze(-1)).squeeze(-1)[mask].sum() loss = (logZ - score).float() return loss, probs
1,329
33.102564
79
py
diaparser
diaparser-master/diaparser/modules/scalar_mix.py
# -*- coding: utf-8 -*- import torch import torch.nn as nn class ScalarMix(nn.Module): r""" Computes a parameterised scalar mixture of :math:`N` tensors, :math:`mixture = \gamma * \sum_{k}(s_k * tensor_k)` where :math:`s = \mathrm{softmax}(w)`, with :math:`w` and :math:`\gamma` scalar parameters. Args: n_layers (int): The number of layers to be mixed, i.e., :math:`N`. dropout (float): The dropout ratio of the layer weights. If dropout > 0, then for each scalar weight, adjust its softmax weight mass to 0 with the dropout probability (i.e., setting the unnormalized weight to -inf). This effectively redistributes the dropped probability mass to all other weights. Default: 0. """ def __init__(self, n_layers: int, dropout: float = 0.0): super().__init__() self.n_layers = n_layers self.weights = nn.Parameter(torch.zeros(n_layers)) self.gamma = nn.Parameter(torch.tensor([1.0])) self.dropout = nn.Dropout(dropout) def __repr__(self): s = f"n_layers={self.n_layers}" if self.dropout.p > 0: s += f", dropout={self.dropout.p}" return f"{self.__class__.__name__}({s})" def forward(self, tensors): r""" Args: tensors (list[~torch.Tensor]): :math:`N` tensors to be mixed. Returns: The mixture of :math:`N` tensors. """ normed_weights = self.dropout(self.weights.softmax(-1)) weighted_sum = sum(w * h for w, h in zip(normed_weights, tensors)) return self.gamma * weighted_sum
1,679
30.698113
117
py
diaparser
diaparser-master/diaparser/models/dependency.py
# -*- coding: utf-8 -*- import torch import torch.nn as nn from ..modules import MLP, BertEmbedding, Biaffine, LSTM, CharLSTM from ..modules.dropout import IndependentDropout, SharedDropout from ..utils.config import Config from ..utils.alg import eisner, mst from ..utils.transform import CoNLL from torch.nn.utils.rnn import pack_padded_sequence, pad_packed_sequence from typing import Tuple class BiaffineDependencyModel(nn.Module): r""" The implementation of Biaffine Dependency Parser. References: - Timothy Dozat and Christopher D. Manning. 2017. `Deep Biaffine Attention for Neural Dependency Parsing`_. Args: n_words (int): The size of the word vocabulary. n_feats (int): The size of the feat vocabulary. n_rels (int): The number of labels in the treebank. feat (str): Specifies which type of additional feature to use: ``'char'`` | ``'bert'`` | ``'tag'``. ``'char'``: Character-level representations extracted by CharLSTM. ``'bert'``: BERT representations, other pretrained langugae models like XLNet are also feasible. ``'tag'``: POS tag embeddings. Default: ``'char'``. n_word_embed (int): The size of word embeddings. Default: 100. n_feat_embed (int): The size of feature representations. Default: 100. n_char_embed (int): The size of character embeddings serving as inputs of CharLSTM, required if ``feat='char'``. Default: 50. bert (str): Specifies which kind of language model to use, e.g., ``'bert-base-cased'`` and ``'xlnet-base-cased'``. This is required if ``feat='bert'``. The full list can be found in `transformers`_. Default: ``None``. n_bert_layers (int): Specifies how many last layers to use. Required if ``feat='bert'``. The final outputs would be the weight sum of the hidden states of these layers. Default: 4. bert_fine_tune (bool): Weather to fine tune the BERT model. Deafult: False. mix_dropout (float): The dropout ratio of BERT layers. Required if ``feat='bert'``. Default: .0. token_dropout (float): The dropout ratio of tokens. Default: .0. embed_dropout (float): The dropout ratio of input embeddings. Default: .33. n_lstm_hidden (int): The size of LSTM hidden states. Default: 400. n_lstm_layers (int): The number of LSTM layers. Default: 3. lstm_dropout (float): The dropout ratio of LSTM. Default: .33. n_mlp_arc (int): Arc MLP size. Default: 500. n_mlp_rel (int): Label MLP size. Default: 100. mlp_dropout (float): The dropout ratio of MLP layers. Default: .33. use_hidden_states (bool): Wethre to use hidden states rather than outputs from BERT. Default: True. use_attentions (bool): Wethre to use attention heads from BERT. Default: False. attention_head (int): Which attention head from BERT to use. Default: 0. attention_layer (int): Which attention layer from BERT to use; use all if 0. Default: 6. feat_pad_index (int): The index of the padding token in the feat vocabulary. Default: 0. pad_index (int): The index of the padding token in the word vocabulary. Default: 0. unk_index (int): The index of the unknown token in the word vocabulary. Default: 1. .. _Deep Biaffine Attention for Neural Dependency Parsing: https://openreview.net/forum?id=Hk95PK9le .. _transformers: https://github.com/huggingface/transformers """ def __init__(self, n_words, n_feats, n_rels, feat='char', n_word_embed=100, n_feat_embed=100, n_char_embed=50, bert=None, n_bert_layers=4, bert_fine_tune=False, mix_dropout=.0, token_dropout=.0, embed_dropout=.33, n_lstm_hidden=400, n_lstm_layers=3, lstm_dropout=.33, n_mlp_arc=500, n_mlp_rel=100, mask_token_id=.0, mlp_dropout=.33, use_hidden_states=True, use_attentions=False, attention_head=0, attention_layer=6, feat_pad_index=0, pad_index=0, unk_index=1, **kwargs): super().__init__() # cant use Config(**locals()) because it includes self self.args = Config().update(locals()) args = self.args if args.n_word_embed: # the embedding layer self.word_embed = nn.Embedding(num_embeddings=args.n_words, embedding_dim=args.n_word_embed) self.unk_index = args.unk_index else: self.word_embed = None if args.feat == 'char': self.feat_embed = CharLSTM(n_chars=args.n_feats, n_word_embed=args.n_char_embed, n_out=args.n_feat_embed, pad_index=args.feat_pad_index) elif args.feat == 'bert': self.feat_embed = BertEmbedding(model=args.bert, n_layers=args.n_bert_layers, n_out=args.n_feat_embed, requires_grad=args.bert_fine_tune, mask_token_id=args.mask_token_id, token_dropout=args.token_dropout, mix_dropout=args.mix_dropout, use_hidden_states=args.use_hidden_states, use_attentions=args.use_attentions, attention_layer=args.attention_layer) # Setting this requires rebuilding models: # args.n_mlp_arc = self.feat_embed.bert.config.max_position_embeddings args.n_feat_embed = self.feat_embed.n_out # taken from the model args.n_bert_layers = self.feat_embed.n_layers # taken from the model elif args.feat == 'tag': self.feat_embed = nn.Embedding(num_embeddings=args.n_feats, embedding_dim=args.n_feat_embed) else: raise RuntimeError("The feat type should be in ['char', 'bert', 'tag'].") self.embed_dropout = IndependentDropout(p=args.embed_dropout) if args.n_lstm_layers: # the lstm layer self.lstm = LSTM(input_size=args.n_word_embed+args.n_feat_embed, hidden_size=args.n_lstm_hidden, num_layers=args.n_lstm_layers, bidirectional=True, dropout=args.lstm_dropout) self.lstm_dropout = SharedDropout(p=args.lstm_dropout) mlp_input_size = args.n_lstm_hidden*2 else: self.lstm = None mlp_input_size = args.n_word_embed + args.n_feat_embed # the MLP layers self.mlp_arc_d = MLP(n_in=mlp_input_size, n_out=args.n_mlp_arc, dropout=args.mlp_dropout) self.mlp_arc_h = MLP(n_in=mlp_input_size, n_out=args.n_mlp_arc, dropout=args.mlp_dropout) self.mlp_rel_d = MLP(n_in=mlp_input_size, n_out=args.n_mlp_rel, dropout=args.mlp_dropout) self.mlp_rel_h = MLP(n_in=mlp_input_size, n_out=args.n_mlp_rel, dropout=args.mlp_dropout) # the Biaffine layers self.arc_attn = Biaffine(n_in=args.n_mlp_arc, bias_x=True, bias_y=False) self.rel_attn = Biaffine(n_in=args.n_mlp_rel, n_out=args.n_rels, bias_x=True, bias_y=True) # transformer attention if args.use_attentions: self.attn_mix = nn.Parameter(torch.randn(1)) self.criterion = nn.CrossEntropyLoss() def extra_repr(self): total_params = sum(p.numel() for p in self.parameters()) trainable_params = sum(p.numel() for p in self.parameters() if p.requires_grad) return f"Total parameters: {total_params}\n" \ f"Trainable parameters: {trainable_params}\n" \ f"Features: {self.args.n_feats}" def load_pretrained(self, embed=None): if embed is not None: self.pretrained = nn.Embedding.from_pretrained(embed) nn.init.zeros_(self.word_embed.weight) return self def forward(self, words: torch.Tensor, feats: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]: r""" Args: words (~torch.LongTensor): ``[batch_size, seq_len]``. Word indices. feats (~torch.LongTensor): Feat indices. If feat is ``'char'`` or ``'bert'``, the size of feats should be ``[batch_size, seq_len, fix_len]``. if ``'tag'``, the size is ``[batch_size, seq_len]``. Returns: ~torch.Tensor, ~torch.Tensor: The first tensor of shape ``[batch_size, seq_len, seq_len]`` holds scores of all possible arcs. The second of shape ``[batch_size, seq_len, seq_len, n_labels]`` holds scores of all possible labels on each arc. """ # words, feats are the first two items in the batch from DataLoader.__iter__() whole_words = feats[:, :, 0] # drop subpiece dimension batch_size, seq_len = whole_words.shape # get the mask and lengths of given batch mask = whole_words.ne(self.feat_embed.pad_index) lens = mask.sum(dim=1).cpu() # BUG fix: https://github.com/pytorch/pytorch/issues/43227 # feat_embed: [batch_size, seq_len, n_feat_embed] # attn: [batch_size, seq_len, seq_len] feat_embed, attn = self.feat_embed(feats) if self.word_embed: ext_words = words # set the indices larger than num_embeddings to unk_index if hasattr(self, 'pretrained'): ext_mask = words.ge(self.word_embed.num_embeddings) ext_words = words.masked_fill(ext_mask, self.unk_index) # get outputs from embedding layers word_embed = self.word_embed(ext_words) if hasattr(self, 'pretrained'): word_embed += self.pretrained(words) word_embed, feat_embed = self.embed_dropout(word_embed, feat_embed) # concatenate the word and feat representations embed = torch.cat((word_embed, feat_embed), dim=-1) else: embed = self.embed_dropout(feat_embed)[0] if self.lstm: x = pack_padded_sequence(embed, lens, True, False) x, _ = self.lstm(x) x, _ = pad_packed_sequence(x, True, total_length=seq_len) x = self.lstm_dropout(x) else: x = embed # apply MLPs to the BiLSTM output states arc_d = self.mlp_arc_d(x) arc_h = self.mlp_arc_h(x) rel_d = self.mlp_rel_d(x) rel_h = self.mlp_rel_h(x) # [batch_size, seq_len, seq_len] s_arc = self.arc_attn(arc_d, arc_h) # [batch_size, seq_len, seq_len, n_rels] s_rel = self.rel_attn(rel_d, rel_h).permute(0, 2, 3, 1) # mix bert attentions if attn is not None: s_arc += self.attn_mix * attn # set the scores that exceed the length of each sentence to -inf s_arc.masked_fill_(~mask.unsqueeze(1), float('-inf')) # Lower the diagonal, because the head of a word can't be itself. s_arc += torch.diag(s_arc.new(seq_len).fill_(float('-inf'))) return s_arc, s_rel def loss(self, s_arc: torch.Tensor, s_rel: torch.Tensor, arcs: torch.Tensor, rels: torch.Tensor, mask: torch.Tensor, partial: bool = False) -> torch.Tensor: r""" Computes the arc and tag loss for a sequence given gold heads and tags. Args: s_arc (~torch.Tensor): ``[batch_size, seq_len, seq_len]``. Scores of all possible arcs. s_rel (~torch.Tensor): ``[batch_size, seq_len, seq_len, n_labels]``. Scores of all possible labels on each arc. arcs (~torch.LongTensor): ``[batch_size, seq_len]``. The tensor of gold-standard arcs. rels (~torch.LongTensor): ``[batch_size, seq_len]``. The tensor of gold-standard labels. mask (~torch.BoolTensor): ``[batch_size, seq_len]``. The mask for covering the unpadded tokens. partial (bool): ``True`` denotes the trees are partially annotated. Default: ``False``. Returns: ~torch.Tensor: The training loss. """ if partial: mask = mask & arcs.ge(0) s_arc, arcs = s_arc[mask], arcs[mask] s_rel, rels = s_rel[mask], rels[mask] # select the predicted relations towards the correct heads s_rel = s_rel[torch.arange(len(arcs)), arcs] arc_loss = self.criterion(s_arc, arcs) rel_loss = self.criterion(s_rel, rels) return arc_loss + rel_loss def decode(self, s_arc: torch.Tensor, s_rel: torch.Tensor, mask: torch.Tensor, tree: bool = False, proj: bool = False) -> Tuple[torch.Tensor, torch.Tensor]: r""" Args: s_arc (~torch.Tensor): ``[batch_size, seq_len, seq_len]``. Scores of all possible arcs. s_rel (~torch.Tensor): ``[batch_size, seq_len, seq_len, n_labels]``. Scores of all possible labels on each arc. mask (~torch.BoolTensor): ``[batch_size, seq_len]``. The mask for covering the unpadded tokens. tree (bool): If ``True``, ensures to output well-formed trees. Default: ``False``. proj (bool): If ``True``, ensures to output projective trees. Default: ``False``. Returns: ~torch.Tensor, ~torch.Tensor: Predicted arcs and labels of shape ``[batch_size, seq_len]``. """ lens = mask.sum(1) # prevent self-loops s_arc.diagonal(0, 1, 2).fill_(float('-inf')) # select the most likely arcs arc_preds = s_arc.argmax(-1) if tree: # ensure the arcs form a tree bad = [not CoNLL.istree(seq[1:i+1], proj) for i, seq in zip(lens.tolist(), arc_preds.tolist())] if any(bad): alg = eisner if proj else mst arc_preds[bad] = alg(s_arc[bad], mask[bad]) # select the most likely rels rel_preds = s_rel.argmax(-1) # choose those corresponding to the predicted arcs rel_preds = rel_preds.gather(-1, arc_preds.unsqueeze(-1)).squeeze(-1) return arc_preds, rel_preds
15,848
42.541209
117
py
diaparser
diaparser-master/diaparser/utils/embedding.py
# -*- coding: utf-8 -*- import torch class Embedding(): def __init__(self, tokens, vectors, unk=None): self.tokens = tokens self.vectors = torch.tensor(vectors) self.pretrained = {w: v for w, v in zip(tokens, vectors)} self.unk = unk def __len__(self): return len(self.tokens) def __contains__(self, token): return token in self.pretrained @property def dim(self): return self.vectors.size(1) @property def unk_index(self): if self.unk is not None: return self.tokens.index(self.unk) else: raise AttributeError @classmethod def load(cls, path, unk=None): with open(path, 'r') as f: lines = [line for line in f] splits = [line.split() for line in lines] tokens, vectors = zip(*[(s[0], list(map(float, s[1:]))) for s in splits]) return cls(tokens, vectors, unk=unk)
982
23.575
65
py
diaparser
diaparser-master/diaparser/utils/field.py
# -*- coding: utf-8 -*- from collections import Counter from ..utils.fn import pad from ..utils.vocab import Vocab, FieldVocab import torch from typing import List class RawField(): r""" Defines a general datatype. A :class:`RawField` object does not assume any property of the datatype and it holds parameters relating to how a datatype should be processed. Args: name (str): The name of the field. fn (function): The function used for preprocessing the examples. Default: ``None``. """ def __init__(self, name, fn=None): self.name = name self.fn = fn def __repr__(self): return f"({self.name}): {self.__class__.__name__}()" def preprocess(self, sequence): return self.fn(sequence) if self.fn is not None else sequence def transform(self, sequences): return [self.preprocess(seq) for seq in sequences] def compose(self, sequences): return sequences class Field(RawField): r""" Defines a datatype together with instructions for converting to :class:`~torch.Tensor`. :class:`Field` models common text processing datatypes that can be represented by tensors. It holds a :class:`Vocab` object that defines the set of possible values for elements of the field and their corresponding numerical representations. The :class:`Field` object also holds other parameters relating to how a datatype should be numericalized, such as a tokenization method. Args: name (str): The name of the field. pad_token (str): The string token used as padding. Default: ``None``. unk_token (str): The string token used to represent OOV words. Default: ``None``. bos_token (str): A token that will be prepended to every example using this field, or ``None`` for no `bos_token`. Default: ``None``. eos_token (str): A token that will be appended to every example using this field, or ``None`` for no `eos_token`. lower (bool): Whether to lowercase the text in this field. Default: ``False``. use_vocab (bool): Whether to use a :class:`Vocab` object. If ``False``, the data in this field should already be numerical. Default: ``True``. tokenize (function): The function used to tokenize strings using this field into sequential examples. Default: ``None``. fn (function): The function used for preprocessing the examples. Default: ``None``. """ def __init__(self, name, pad=None, unk=None, bos=None, eos=None, lower=False, use_vocab=True, tokenize=None, fn=None, mask_token_id=0): self.name = name self.pad = pad self.unk = unk self.bos = bos self.eos = eos self.lower = lower self.use_vocab = use_vocab self.tokenize = tokenize self.fn = fn self.mask_token_id = mask_token_id # Attardi self.specials = [token for token in [pad, unk, bos, eos] if token is not None] def __repr__(self): s, params = f"({self.name}): {self.__class__.__name__}(", [] if self.pad is not None: params.append(f"pad={self.pad}") if self.unk is not None: params.append(f"unk={self.unk}") if self.bos is not None: params.append(f"bos={self.bos}") if self.eos is not None: params.append(f"eos={self.eos}") if self.lower: params.append(f"lower={self.lower}") if not self.use_vocab: params.append(f"use_vocab={self.use_vocab}") s += ", ".join(params) s += ")" return s @property def pad_index(self): if self.pad is None: return 0 if hasattr(self, 'vocab'): return self.vocab[self.pad] return self.specials.index(self.pad) @property def unk_index(self): if self.unk is None: return 0 if hasattr(self, 'vocab'): return self.vocab[self.unk] return self.specials.index(self.unk) @property def bos_index(self): if hasattr(self, 'vocab'): return self.vocab[self.bos] if self.bos else 0 return self.specials.index(self.bos) if self.bos else 0 @property def eos_index(self): if hasattr(self, 'vocab'): return self.vocab[self.eos] if self.eos else 0 return self.specials.index(self.eos) if self.eos else 0 @property def device(self): return 'cuda' if torch.cuda.is_available() else 'cpu' def preprocess(self, sequence): r""" Loads a single example using this field, tokenizing if necessary. The sequence will be first passed to ``fn`` if available. If ``tokenize`` is not None, the input will be tokenized. Then the input will be lowercased optionally. Args: sequence (list): The sequence to be preprocessed. Returns: A list of preprocessed sequence. """ if self.fn is not None: sequence = self.fn(sequence) if self.tokenize is not None: sequence = self.tokenize(sequence) if self.lower: sequence = [str.lower(token) for token in sequence] return sequence def build(self, dataset, min_freq=1, embed=None): r""" Constructs a :class:`Vocab` object for this field from the dataset. If the vocabulary has already existed, this function will have no effect. Args: dataset (Dataset): A :class:`Dataset` object. One of the attributes should be named after the name of this field. min_freq (int): The minimum frequency needed to include a token in the vocabulary. Default: 1. embed (Embedding): An Embedding object, words in which will be extended to the vocabulary. Default: ``None``. """ if hasattr(self, 'vocab'): return sequences = getattr(dataset, self.name) counter = Counter(token for seq in sequences for token in self.preprocess(seq)) self.vocab = Vocab(counter, min_freq, self.specials, self.unk_index) if not embed: self.embed = None else: tokens = self.preprocess(embed.tokens) # if the `unk` token was present in the pretrained, # then replace it with a self-defined one if embed.unk: tokens[embed.unk_index] = self.unk self.vocab.extend(tokens) self.embed = torch.zeros(len(self.vocab), embed.dim) self.embed[self.vocab[tokens]] = embed.vectors self.embed /= torch.std(self.embed) def transform(self, sequences: List[List[str]]) -> List[torch.Tensor]: r""" Turns a list of sequences that use this field into tensors. Each sequence is first preprocessed and then numericalized if needed. Args: sequences (list[list[str]]): A list of sequences. Returns: A list of tensors transformed from the input sequences. """ sequences = [self.preprocess(seq) for seq in sequences] if self.use_vocab: sequences = [self.vocab[seq] for seq in sequences] if self.bos: sequences = [[self.bos_index] + seq for seq in sequences] if self.eos: sequences = [seq + [self.eos_index] for seq in sequences] sequences = [torch.tensor(seq) for seq in sequences] return sequences def compose(self, sequences): r""" Composes a batch of sequences into a padded tensor. Args: sequences (list[~torch.Tensor]): A list of tensors. Returns: A padded tensor converted to proper device. """ return pad(sequences, self.pad_index).to(self.device) class SubwordField(Field): r""" A field that conducts tokenization and numericalization over each token rather the sequence. This is customized for models requiring character/subword-level inputs, e.g., CharLSTM and BERT. Args: fix_len (int): A fixed length that all subword pieces will be padded to. This is used for truncating the subword pieces that exceed the length. To save the memory, the final length will be the smaller value between the max length of subword pieces in a batch and `fix_len`. Examples: >>> from transformers import AutoTokenizer >>> tokenizer = AutoTokenizer.from_pretrained('bert-base-cased') >>> field = SubwordField('bert', pad=tokenizer.pad_token, unk=tokenizer.unk_token, bos=tokenizer.cls_token, eos=tokenizer.sep_token, fix_len=20, tokenize=tokenizer.tokenize) >>> field.vocab = tokenizer.get_vocab() # no need to re-build the vocab >>> field.transform([['This', 'field', 'performs', 'token-level', 'tokenization']])[0] tensor([[ 101, 0, 0], [ 1188, 0, 0], [ 1768, 0, 0], [10383, 0, 0], [22559, 118, 1634], [22559, 2734, 0], [ 102, 0, 0]]) """ def __init__(self, *args, fix_len=0, **kwargs): self.fix_len = fix_len super().__init__(*args, **kwargs) def build(self, dataset, min_freq=1, embed=None): sequences = getattr(dataset, self.name) counter = Counter(piece for seq in sequences for token in seq for piece in self.preprocess(token)) self.vocab = Vocab(counter, min_freq, self.specials, self.unk_index) if not embed: self.embed = None else: tokens = self.preprocess(embed.tokens) # if the `unk` token has existed in the pretrained, # then replace it with a self-defined one if embed.unk: tokens[embed.unk_index] = self.unk self.vocab.extend(tokens) self.embed = torch.zeros(len(self.vocab), embed.dim) self.embed[self.vocab[tokens]] = embed.vectors def transform(self, sequences): sequences = [[self.preprocess(token) for token in seq] for seq in sequences] if self.fix_len <= 0: self.fix_len = max(len(token) for seq in sequences for token in seq) if self.use_vocab: sequences = [[[self.vocab[i] for i in token] for token in seq] for seq in sequences] if self.bos: sequences = [[[self.bos_index]] + seq for seq in sequences] if self.eos: sequences = [seq + [[self.eos_index]] for seq in sequences] lens = [min(self.fix_len, max(len(ids) for ids in seq)) for seq in sequences] sequences = [pad([torch.tensor(ids[:i]) for ids in seq], self.pad_index, i) for i, seq in zip(lens, sequences)] return sequences class BertField(SubwordField): r""" A field that is dealt by a transformer. Args: name (str): name of the field. tokenizer (AutoTokenizer): the tokenizer for the transformer. fix_len (int): A fixed length that all subword pieces will be padded to. This is used for truncating the subword pieces that exceed the length. To save the memory, the final length will be the smaller value between the max length of subword pieces in a batch and `fix_len`. Examples: >>> tokenizer = BertField.tokenizer('bert-base-cased') >>> field = BertField('bert', tokenizer, fix_len=20) >>> field.transform([['This', 'field', 'performs', 'token-level', 'tokenization']])[0] tensor([[ 101, 0, 0], [ 1188, 0, 0], [ 1768, 0, 0], [10383, 0, 0], [22559, 118, 1634], [22559, 2734, 0], [ 102, 0, 0]]) """ def __init__(self, name, tokenizer, **kwargs): if hasattr(tokenizer, 'vocab'): self.vocab = tokenizer.get_vocab() else: self.vocab = FieldVocab(tokenizer.unk_token_id, {tokenizer._convert_id_to_token(i): i for i in range(len(tokenizer))}) super().__init__(name, pad=tokenizer.pad_token, unk=tokenizer.unk_token, bos=tokenizer.bos_token or tokenizer.cls_token, mask_token_id=tokenizer.mask_token_id, tokenize=tokenizer.tokenize, **kwargs) def build(self, dataset): """ Pretrained: nothing to be done. """ return @classmethod def tokenizer(cls, name): """ Create an instance of tokenizer from either path or name. :param name: path or name of tokenizer. """ from transformers import AutoTokenizer tokenizer = AutoTokenizer.from_pretrained(name) tokenizer.bos_token = tokenizer.bos_token or tokenizer.cls_token tokenizer.eos_token = tokenizer.eos_token or tokenizer.sep_token return tokenizer class ChartField(Field): r""" Field dealing with constituency trees. This field receives sequences of binarized trees factorized in pre-order, and returns charts filled with labels on each constituent. Examples: >>> sequence = [(0, 5, 'S'), (0, 4, 'S|<>'), (0, 1, 'NP'), (1, 4, 'VP'), (1, 2, 'VP|<>'), (2, 4, 'S+VP'), (2, 3, 'VP|<>'), (3, 4, 'NP'), (4, 5, 'S|<>')] >>> field.transform([sequence])[0] tensor([[ -1, 37, -1, -1, 107, 79], [ -1, -1, 120, -1, 112, -1], [ -1, -1, -1, 120, 86, -1], [ -1, -1, -1, -1, 37, -1], [ -1, -1, -1, -1, -1, 107], [ -1, -1, -1, -1, -1, -1]]) """ def build(self, dataset, min_freq=1): counter = Counter(label for seq in getattr(dataset, self.name) for i, j, label in self.preprocess(seq)) self.vocab = Vocab(counter, min_freq, self.specials, self.unk_index) def transform(self, sequences): charts = [] for sequence in sequences: sequence = self.preprocess(sequence) seq_len = sequence[0][1] + 1 chart = torch.full((seq_len, seq_len), -1, dtype=torch.long) for i, j, label in sequence: chart[i, j] = self.vocab[label] charts.append(chart) return charts
15,427
35.733333
117
py
diaparser
diaparser-master/diaparser/utils/alg.py
# -*- coding: utf-8 -*- import torch from ..utils.fn import pad, stripe def kmeans(x, k, max_it=32): r""" KMeans algorithm for clustering the sentences by length. Args: x (list[int]): The list of sentence lengths. k (int): The number of clusters. This is an approximate value. The final number of clusters can be less or equal to `k`. max_it (int): Maximum number of iterations. If centroids does not converge after several iterations, the algorithm will be early stopped. Returns: list[float], list[list[int]]: The first list contains average lengths of sentences in each cluster. The second is the list of clusters holding indices of data points. Examples: >>> x = torch.randint(10,20,(10,)).tolist() >>> x [15, 10, 17, 11, 18, 13, 17, 19, 18, 14] >>> centroids, clusters = kmeans(x, 3) >>> centroids [10.5, 14.0, 17.799999237060547] >>> clusters [[1, 3], [0, 5, 9], [2, 4, 6, 7, 8]] """ # the number of clusters must not be greater than the number of datapoints x, k = torch.tensor(x, dtype=torch.float), min(len(x), k) # collect unique datapoints d = x.unique() # initialize k centroids randomly c = d[torch.randperm(len(d))[:k]] # assign each datapoint to the cluster with the closest centroid dists, y = torch.abs_(x.unsqueeze(-1) - c).min(-1) for _ in range(max_it): # if an empty cluster is encountered, # choose the farthest datapoint from the biggest cluster and move that the empty one mask = torch.arange(k).unsqueeze(-1).eq(y) none = torch.where(~mask.any(-1))[0].tolist() while len(none) > 0: for i in none: # the biggest cluster b = torch.where(mask[mask.sum(-1).argmax()])[0] # the datapoint farthest from the centroid of cluster b f = dists[b].argmax() # update the assigned cluster of f y[b[f]] = i # re-calculate the mask mask = torch.arange(k).unsqueeze(-1).eq(y) none = torch.where(~mask.any(-1))[0].tolist() # update the centroids c, old = (x * mask).sum(-1) / mask.sum(-1), c # re-assign all datapoints to clusters dists, y = torch.abs_(x.unsqueeze(-1) - c).min(-1) # stop iteration early if the centroids converge if c.equal(old): break # assign all datapoints to the new-generated clusters # the empty ones are discarded assigned = y.unique().tolist() # get the centroids of the assigned clusters centroids = c[assigned].tolist() # map all values of datapoints to buckets clusters = [torch.where(y.eq(i))[0].tolist() for i in assigned] return centroids, clusters def tarjan(sequence): r""" Tarjan algorithm for finding Strongly Connected Components (SCCs) of a graph. Args: sequence (list): List of head indices. Yields: A list of indices that make up a SCC. All self-loops are ignored. Examples: >>> next(tarjan([2, 5, 0, 3, 1])) # (1 -> 5 -> 2 -> 1) is a cycle [2, 5, 1] """ sequence = [-1] + sequence # record the search order, i.e., the timestep dfn = [-1] * len(sequence) # record the the smallest timestep in a SCC low = [-1] * len(sequence) # push the visited into the stack stack, onstack = [], [False] * len(sequence) def connect(i, timestep): dfn[i] = low[i] = timestep[0] timestep[0] += 1 stack.append(i) onstack[i] = True for j, head in enumerate(sequence): if head != i: continue if dfn[j] == -1: yield from connect(j, timestep) low[i] = min(low[i], low[j]) elif onstack[j]: low[i] = min(low[i], dfn[j]) # a SCC is completed if low[i] == dfn[i]: cycle = [stack.pop()] while cycle[-1] != i: onstack[cycle[-1]] = False cycle.append(stack.pop()) onstack[i] = False # ignore the self-loop if len(cycle) > 1: yield cycle timestep = [0] for i in range(len(sequence)): if dfn[i] == -1: yield from connect(i, timestep) def chuliu_edmonds(s): r""" ChuLiu/Edmonds algorithm for non-projective decoding. Some code is borrowed from `tdozat's implementation`_. Descriptions of notations and formulas can be found in `Non-projective Dependency Parsing using Spanning Tree Algorithms`_. Notes: The algorithm does not guarantee to parse a single-root tree. References: - Ryan McDonald, Fernando Pereira, Kiril Ribarov and Jan Hajic. 2005. `Non-projective Dependency Parsing using Spanning Tree Algorithms`_. Args: s (~torch.Tensor): ``[seq_len, seq_len]``. Scores of all dependent-head pairs. Returns: ~torch.Tensor: A tensor with shape ``[seq_len]`` for the resulting non-projective parse tree. .. _tdozat's implementation: https://github.com/tdozat/Parser-v3 .. _Non-projective Dependency Parsing using Spanning Tree Algorithms: https://www.aclweb.org/anthology/H05-1066/ """ s[0, 1:] = float('-inf') # prevent self-loops s.diagonal()[1:].fill_(float('-inf')) # select heads with highest scores tree = s.argmax(-1) # return the cycle finded by tarjan algorithm lazily cycle = next(tarjan(tree.tolist()[1:]), None) # if the tree has no cycles, then it is a MST if not cycle: return tree # indices of cycle in the original tree cycle = torch.tensor(cycle) # indices of noncycle in the original tree noncycle = torch.ones(len(s)).index_fill_(0, cycle, 0) noncycle = torch.where(noncycle.gt(0))[0] def contract(s): # heads of cycle in original tree cycle_heads = tree[cycle] # scores of cycle in original tree s_cycle = s[cycle, cycle_heads] # calculate the scores of cycle's potential dependents # s(c->x) = max(s(x'->x)), x in noncycle and x' in cycle s_dep = s[noncycle][:, cycle] # find the best cycle head for each noncycle dependent deps = s_dep.argmax(1) # calculate the scores of cycle's potential heads # s(x->c) = max(s(x'->x) - s(a(x')->x') + s(cycle)), x in noncycle and x' in cycle # a(v) is the predecessor of v in cycle # s(cycle) = sum(s(a(v)->v)) s_head = s[cycle][:, noncycle] - s_cycle.view(-1, 1) + s_cycle.sum() # find the best noncycle head for each cycle dependent heads = s_head.argmax(0) contracted = torch.cat((noncycle, torch.tensor([-1]))) # calculate the scores of contracted graph s = s[contracted][:, contracted] # set the contracted graph scores of cycle's potential dependents s[:-1, -1] = s_dep[range(len(deps)), deps] # set the contracted graph scores of cycle's potential heads s[-1, :-1] = s_head[heads, range(len(heads))] return s, heads, deps # keep track of the endpoints of the edges into and out of cycle for reconstruction later s, heads, deps = contract(s) # y is the contracted tree y = chuliu_edmonds(s) # exclude head of cycle from y y, cycle_head = y[:-1], y[-1] # fix the subtree with no heads coming from the cycle # len(y) denotes heads coming from the cycle subtree = y < len(y) # add the nodes to the new tree tree[noncycle[subtree]] = noncycle[y[subtree]] # fix the subtree with heads coming from the cycle subtree = ~subtree # add the nodes to the tree tree[noncycle[subtree]] = cycle[deps[subtree]] # fix the root of the cycle cycle_root = heads[cycle_head] # break the cycle and add the root of the cycle to the tree tree[cycle[cycle_root]] = noncycle[cycle_head] return tree def mst(scores, mask, multiroot=False): r""" MST algorithm for decoding non-pojective trees. This is a wrapper for ChuLiu/Edmonds algorithm. The algorithm first runs ChuLiu/Edmonds to parse a tree and then have a check of multi-roots, If ``multiroot=True`` and there indeed exist multi-roots, the algorithm seeks to find best single-root trees by iterating all possible single-root trees parsed by ChuLiu/Edmonds. Otherwise the resulting trees are directly taken as the final outputs. Args: scores (~torch.Tensor): ``[batch_size, seq_len, seq_len]``. Scores of all dependent-head pairs. mask (~torch.BoolTensor): ``[batch_size, seq_len]``. The mask to avoid parsing over padding tokens. The first column serving as pseudo words for roots should be ``False``. muliroot (bool): Ensures to parse a single-root tree If ``False``. Returns: ~torch.Tensor: A tensor with shape ``[batch_size, seq_len]`` for the resulting non-projective parse trees. Examples: >>> scores = torch.tensor([[[-11.9436, -13.1464, -6.4789, -13.8917], [-60.6957, -60.2866, -48.6457, -63.8125], [-38.1747, -49.9296, -45.2733, -49.5571], [-19.7504, -23.9066, -9.9139, -16.2088]]]) >>> scores[:, 0, 1:] = float('-inf') >>> scores.diagonal(0, 1, 2)[1:].fill_(float('-inf')) >>> mask = torch.tensor([[False, True, True, True]]) >>> mst(scores, mask) tensor([[0, 2, 0, 2]]) """ batch_size, seq_len, _ = scores.shape scores = scores.cpu().unbind() preds = [] for i, length in enumerate(mask.sum(1).tolist()): s = scores[i][:length+1, :length+1] tree = chuliu_edmonds(s) roots = torch.where(tree[1:].eq(0))[0] + 1 if not multiroot and len(roots) > 1: s_root = s[:, 0] s_best = float('-inf') s = s.index_fill(1, torch.tensor(0), float('-inf')) for root in roots: s[:, 0] = float('-inf') s[root, 0] = s_root[root] t = chuliu_edmonds(s) s_tree = s[1:].gather(1, t[1:].unsqueeze(-1)).sum() if s_tree > s_best: s_best, tree = s_tree, t preds.append(tree) return pad(preds, total_length=seq_len).to(mask.device) def eisner(scores, mask): r""" First-order Eisner algorithm for projective decoding. References: - Ryan McDonald, Koby Crammer and Fernando Pereira. 2005. `Online Large-Margin Training of Dependency Parsers`_. Args: scores (~torch.Tensor): ``[batch_size, seq_len, seq_len]``. Scores of all dependent-head pairs. mask (~torch.BoolTensor): ``[batch_size, seq_len]``. The mask to avoid parsing over padding tokens. The first column serving as pseudo words for roots should be ``False``. Returns: ~torch.Tensor: A tensor with shape ``[batch_size, seq_len]`` for the resulting projective parse trees. Examples: >>> scores = torch.tensor([[[-13.5026, -18.3700, -13.0033, -16.6809], [-36.5235, -28.6344, -28.4696, -31.6750], [ -2.9084, -7.4825, -1.4861, -6.8709], [-29.4880, -27.6905, -26.1498, -27.0233]]]) >>> mask = torch.tensor([[False, True, True, True]]) >>> eisner(scores, mask) tensor([[0, 2, 0, 2]]) .. _Online Large-Margin Training of Dependency Parsers: https://www.aclweb.org/anthology/P05-1012/ """ lens = mask.sum(1) batch_size, seq_len, _ = scores.shape scores = scores.permute(2, 1, 0) s_i = torch.full_like(scores, float('-inf')) s_c = torch.full_like(scores, float('-inf')) p_i = scores.new_zeros(seq_len, seq_len, batch_size).long() p_c = scores.new_zeros(seq_len, seq_len, batch_size).long() s_c.diagonal().fill_(0) for w in range(1, seq_len): n = seq_len - w starts = p_i.new_tensor(range(n)).unsqueeze(0) # ilr = C(i->r) + C(j->r+1) ilr = stripe(s_c, n, w) + stripe(s_c, n, w, (w, 1)) # [batch_size, n, w] il = ir = ilr.permute(2, 0, 1) # I(j->i) = max(C(i->r) + C(j->r+1) + s(j->i)), i <= r < j il_span, il_path = il.max(-1) s_i.diagonal(-w).copy_(il_span + scores.diagonal(-w)) p_i.diagonal(-w).copy_(il_path + starts) # I(i->j) = max(C(i->r) + C(j->r+1) + s(i->j)), i <= r < j ir_span, ir_path = ir.max(-1) s_i.diagonal(w).copy_(ir_span + scores.diagonal(w)) p_i.diagonal(w).copy_(ir_path + starts) # C(j->i) = max(C(r->i) + I(j->r)), i <= r < j cl = stripe(s_c, n, w, (0, 0), 0) + stripe(s_i, n, w, (w, 0)) cl_span, cl_path = cl.permute(2, 0, 1).max(-1) s_c.diagonal(-w).copy_(cl_span) p_c.diagonal(-w).copy_(cl_path + starts) # C(i->j) = max(I(i->r) + C(r->j)), i < r <= j cr = stripe(s_i, n, w, (0, 1)) + stripe(s_c, n, w, (1, w), 0) cr_span, cr_path = cr.permute(2, 0, 1).max(-1) s_c.diagonal(w).copy_(cr_span) s_c[0, w][lens.ne(w)] = float('-inf') p_c.diagonal(w).copy_(cr_path + starts + 1) def backtrack(p_i, p_c, heads, i, j, complete): if i == j: return if complete: r = p_c[i, j] backtrack(p_i, p_c, heads, i, r, False) backtrack(p_i, p_c, heads, r, j, True) else: r, heads[j] = p_i[i, j], i i, j = sorted((i, j)) backtrack(p_i, p_c, heads, i, r, True) backtrack(p_i, p_c, heads, j, r + 1, True) preds = [] p_c = p_c.permute(2, 0, 1).cpu() p_i = p_i.permute(2, 0, 1).cpu() for i, length in enumerate(lens.tolist()): heads = p_c.new_zeros(length + 1, dtype=torch.long) backtrack(p_i[i], p_c[i], heads, 0, length, True) preds.append(heads.to(mask.device)) return pad(preds, total_length=seq_len).to(mask.device) def eisner2o(scores, mask): r""" Second-order Eisner algorithm for projective decoding. This is an extension of the first-order one that further incorporates sibling scores into tree scoring. References: - Ryan McDonald and Fernando Pereira. 2006. `Online Learning of Approximate Dependency Parsing Algorithms`_. Args: scores (~torch.Tensor, ~torch.Tensor): A tuple of two tensors representing the first-order and second-order scores repectively. The first (``[batch_size, seq_len, seq_len]``) holds scores of all dependent-head pairs. The second (``[batch_size, seq_len, seq_len, seq_len]``) holds scores of all dependent-head-sibling triples. mask (~torch.BoolTensor): ``[batch_size, seq_len]``. The mask to avoid parsing over padding tokens. The first column serving as pseudo words for roots should be ``False``. Returns: ~torch.Tensor: A tensor with shape ``[batch_size, seq_len]`` for the resulting projective parse trees. Examples: >>> s_arc = torch.tensor([[[ -2.8092, -7.9104, -0.9414, -5.4360], [-10.3494, -7.9298, -3.6929, -7.3985], [ 1.1815, -3.8291, 2.3166, -2.7183], [ -3.9776, -3.9063, -1.6762, -3.1861]]]) >>> s_sib = torch.tensor([[[[ 0.4719, 0.4154, 1.1333, 0.6946], [ 1.1252, 1.3043, 2.1128, 1.4621], [ 0.5974, 0.5635, 1.0115, 0.7550], [ 1.1174, 1.3794, 2.2567, 1.4043]], [[-2.1480, -4.1830, -2.5519, -1.8020], [-1.2496, -1.7859, -0.0665, -0.4938], [-2.6171, -4.0142, -2.9428, -2.2121], [-0.5166, -1.0925, 0.5190, 0.1371]], [[ 0.5827, -1.2499, -0.0648, -0.0497], [ 1.4695, 0.3522, 1.5614, 1.0236], [ 0.4647, -0.7996, -0.3801, 0.0046], [ 1.5611, 0.3875, 1.8285, 1.0766]], [[-1.3053, -2.9423, -1.5779, -1.2142], [-0.1908, -0.9699, 0.3085, 0.1061], [-1.6783, -2.8199, -1.8853, -1.5653], [ 0.3629, -0.3488, 0.9011, 0.5674]]]]) >>> mask = torch.tensor([[False, True, True, True]]) >>> eisner2o((s_arc, s_sib), mask) tensor([[0, 2, 0, 2]]) .. _Online Learning of Approximate Dependency Parsing Algorithms: https://www.aclweb.org/anthology/E06-1011/ """ # the end position of each sentence in a batch lens = mask.sum(1) s_arc, s_sib = scores batch_size, seq_len, _ = s_arc.shape # [seq_len, seq_len, batch_size] s_arc = s_arc.permute(2, 1, 0) # [seq_len, seq_len, seq_len, batch_size] s_sib = s_sib.permute(2, 1, 3, 0) s_i = torch.full_like(s_arc, float('-inf')) s_s = torch.full_like(s_arc, float('-inf')) s_c = torch.full_like(s_arc, float('-inf')) p_i = s_arc.new_zeros(seq_len, seq_len, batch_size).long() p_s = s_arc.new_zeros(seq_len, seq_len, batch_size).long() p_c = s_arc.new_zeros(seq_len, seq_len, batch_size).long() s_c.diagonal().fill_(0) for w in range(1, seq_len): # n denotes the number of spans to iterate, # from span (0, w) to span (n, n+w) given width w n = seq_len - w starts = p_i.new_tensor(range(n)).unsqueeze(0) # I(j->i) = max(I(j->r) + S(j->r, i)), i < r < j | # C(j->j) + C(i->j-1)) # + s(j->i) # [n, w, batch_size] il = stripe(s_i, n, w, (w, 1)) + stripe(s_s, n, w, (1, 0), 0) il += stripe(s_sib[range(w, n+w), range(n)], n, w, (0, 1)) # [n, 1, batch_size] il0 = stripe(s_c, n, 1, (w, w)) + stripe(s_c, n, 1, (0, w - 1)) # il0[0] are set to zeros since the scores of the complete spans starting from 0 are always -inf il[:, -1] = il0.index_fill_(0, lens.new_tensor(0), 0).squeeze(1) il_span, il_path = il.permute(2, 0, 1).max(-1) s_i.diagonal(-w).copy_(il_span + s_arc.diagonal(-w)) p_i.diagonal(-w).copy_(il_path + starts + 1) # I(i->j) = max(I(i->r) + S(i->r, j), i < r < j | # C(i->i) + C(j->i+1)) # + s(i->j) # [n, w, batch_size] ir = stripe(s_i, n, w) + stripe(s_s, n, w, (0, w), 0) ir += stripe(s_sib[range(n), range(w, n+w)], n, w) ir[0] = float('-inf') # [n, 1, batch_size] ir0 = stripe(s_c, n, 1) + stripe(s_c, n, 1, (w, 1)) ir[:, 0] = ir0.squeeze(1) ir_span, ir_path = ir.permute(2, 0, 1).max(-1) s_i.diagonal(w).copy_(ir_span + s_arc.diagonal(w)) p_i.diagonal(w).copy_(ir_path + starts) # [n, w, batch_size] slr = stripe(s_c, n, w) + stripe(s_c, n, w, (w, 1)) slr_span, slr_path = slr.permute(2, 0, 1).max(-1) # S(j, i) = max(C(i->r) + C(j->r+1)), i <= r < j s_s.diagonal(-w).copy_(slr_span) p_s.diagonal(-w).copy_(slr_path + starts) # S(i, j) = max(C(i->r) + C(j->r+1)), i <= r < j s_s.diagonal(w).copy_(slr_span) p_s.diagonal(w).copy_(slr_path + starts) # C(j->i) = max(C(r->i) + I(j->r)), i <= r < j cl = stripe(s_c, n, w, (0, 0), 0) + stripe(s_i, n, w, (w, 0)) cl_span, cl_path = cl.permute(2, 0, 1).max(-1) s_c.diagonal(-w).copy_(cl_span) p_c.diagonal(-w).copy_(cl_path + starts) # C(i->j) = max(I(i->r) + C(r->j)), i < r <= j cr = stripe(s_i, n, w, (0, 1)) + stripe(s_c, n, w, (1, w), 0) cr_span, cr_path = cr.permute(2, 0, 1).max(-1) s_c.diagonal(w).copy_(cr_span) # disable multi words to modify the root s_c[0, w][lens.ne(w)] = float('-inf') p_c.diagonal(w).copy_(cr_path + starts + 1) def backtrack(p_i, p_s, p_c, heads, i, j, flag): if i == j: return if flag == 'c': r = p_c[i, j] backtrack(p_i, p_s, p_c, heads, i, r, 'i') backtrack(p_i, p_s, p_c, heads, r, j, 'c') elif flag == 's': r = p_s[i, j] i, j = sorted((i, j)) backtrack(p_i, p_s, p_c, heads, i, r, 'c') backtrack(p_i, p_s, p_c, heads, j, r + 1, 'c') elif flag == 'i': r, heads[j] = p_i[i, j], i if r == i: r = i + 1 if i < j else i - 1 backtrack(p_i, p_s, p_c, heads, j, r, 'c') else: backtrack(p_i, p_s, p_c, heads, i, r, 'i') backtrack(p_i, p_s, p_c, heads, r, j, 's') preds = [] p_i = p_i.permute(2, 0, 1).cpu() p_s = p_s.permute(2, 0, 1).cpu() p_c = p_c.permute(2, 0, 1).cpu() for i, length in enumerate(lens.tolist()): heads = p_c.new_zeros(length + 1, dtype=torch.long) backtrack(p_i[i], p_s[i], p_c[i], heads, 0, length, 'c') preds.append(heads.to(mask.device)) return pad(preds, total_length=seq_len).to(mask.device) def cky(scores, mask): r""" The implementation of `Cocke-Kasami-Younger`_ (CKY) algorithm to parse constituency trees. References: - Yu Zhang, Houquan Zhou and Zhenghua Li. 2020. `Fast and Accurate Neural CRF Constituency Parsing`_. Args: scores (~torch.Tensor): ``[batch_size, seq_len, seq_len]``. Scores of all candidate constituents. mask (~torch.BoolTensor): ``[batch_size, seq_len, seq_len]``. The mask to avoid parsing over padding tokens. For each square matrix in a batch, the positions except upper triangular part should be masked out. Returns: Sequences of factorized predicted bracketed trees that are traversed in pre-order. Examples: >>> scores = torch.tensor([[[ 2.5659, 1.4253, -2.5272, 3.3011], [ 1.3687, -0.5869, 1.0011, 3.3020], [ 1.2297, 0.4862, 1.1975, 2.5387], [-0.0511, -1.2541, -0.7577, 0.2659]]]) >>> mask = torch.tensor([[[False, True, True, True], [False, False, True, True], [False, False, False, True], [False, False, False, False]]]) >>> cky(scores, mask) [[(0, 3), (0, 1), (1, 3), (1, 2), (2, 3)]] .. _Cocke-Kasami-Younger: https://en.wikipedia.org/wiki/CYK_algorithm .. _Fast and Accurate Neural CRF Constituency Parsing: https://www.ijcai.org/Proceedings/2020/560/ """ lens = mask[:, 0].sum(-1) scores = scores.permute(1, 2, 0) seq_len, seq_len, batch_size = scores.shape s = scores.new_zeros(seq_len, seq_len, batch_size) p = scores.new_zeros(seq_len, seq_len, batch_size).long() for w in range(1, seq_len): n = seq_len - w starts = p.new_tensor(range(n)).unsqueeze(0) if w == 1: s.diagonal(w).copy_(scores.diagonal(w)) continue # [n, w, batch_size] s_span = stripe(s, n, w-1, (0, 1)) + stripe(s, n, w-1, (1, w), 0) # [batch_size, n, w] s_span = s_span.permute(2, 0, 1) # [batch_size, n] s_span, p_span = s_span.max(-1) s.diagonal(w).copy_(s_span + scores.diagonal(w)) p.diagonal(w).copy_(p_span + starts + 1) def backtrack(p, i, j): if j == i + 1: return [(i, j)] split = p[i][j] ltree = backtrack(p, i, split) rtree = backtrack(p, split, j) return [(i, j)] + ltree + rtree p = p.permute(2, 0, 1).tolist() trees = [backtrack(p[i], 0, length) for i, length in enumerate(lens.tolist())] return trees
24,571
39.150327
120
py
diaparser
diaparser-master/diaparser/utils/fn.py
# -*- coding: utf-8 -*- import unicodedata def ispunct(token): return all(unicodedata.category(char).startswith('P') for char in token) def isfullwidth(token): return all(unicodedata.east_asian_width(char) in ['W', 'F', 'A'] for char in token) def islatin(token): return all('LATIN' in unicodedata.name(char) for char in token) def isdigit(token): return all('DIGIT' in unicodedata.name(char) for char in token) def tohalfwidth(token): return unicodedata.normalize('NFKC', token) def isprojective(sequence): sequence = [0] + list(sequence) arcs = [(h, d) for d, h in enumerate(sequence[1:], 1) if h >= 0] for i, (hi, di) in enumerate(arcs): for hj, dj in arcs[i+1:]: (li, ri), (lj, rj) = sorted([hi, di]), sorted([hj, dj]) if (li <= hj <= ri and hi == dj) or (lj <= hi <= rj and hj == di): return False if (li < lj < ri or li < rj < ri) and (li - lj) * (ri - rj) > 0: return False return True def istree(sequence, proj=False, multiroot=False): from ..utils.alg import tarjan if proj and not isprojective(sequence): return False n_roots = sum(head == 0 for head in sequence[1:]) if n_roots == 0: return False if not multiroot and n_roots > 1: return False return next(tarjan(sequence), None) is None def numericalize(sequence): return [int(i) for i in sequence] def stripe(x, n, w, offset=(0, 0), dim=1): r""" Returns a diagonal stripe of the tensor. Args: x (~torch.Tensor): the input tensor with 2 or more dims. n (int): the length of the stripe. w (int): the width of the stripe. offset (tuple): the offset of the first two dims. dim (int): 1 if returns a horizontal stripe; 0 otherwise. Returns: a diagonal stripe of the tensor. Examples: >>> x = torch.arange(25).view(5, 5) >>> x tensor([[ 0, 1, 2, 3, 4], [ 5, 6, 7, 8, 9], [10, 11, 12, 13, 14], [15, 16, 17, 18, 19], [20, 21, 22, 23, 24]]) >>> stripe(x, 2, 3) tensor([[0, 1, 2], [6, 7, 8]]) >>> stripe(x, 2, 3, (1, 1)) tensor([[ 6, 7, 8], [12, 13, 14]]) >>> stripe(x, 2, 3, (1, 1), 0) tensor([[ 6, 11, 16], [12, 17, 22]]) """ x, seq_len = x.contiguous(), x.size(1) stride, numel = list(x.stride()), x[0, 0].numel() stride[0] = (seq_len + 1) * numel stride[1] = (1 if dim == 1 else seq_len) * numel return x.as_strided(size=(n, w, *x.shape[2:]), stride=stride, storage_offset=(offset[0]*seq_len+offset[1])*numel) def pad(tensors, padding_value=0, total_length=None): size = [len(tensors)] + [max(tensor.size(i) for tensor in tensors) for i in range(len(tensors[0].size()))] if total_length is not None: assert total_length >= size[1] size[1] = total_length out_tensor = tensors[0].data.new(*size).fill_(padding_value) for i, tensor in enumerate(tensors): out_tensor[i][[slice(0, i) for i in tensor.size()]] = tensor return out_tensor
3,344
29.135135
78
py
diaparser
diaparser-master/diaparser/utils/data.py
# -*- coding: utf-8 -*- from collections import namedtuple import torch import torch.distributed as dist from ..utils.alg import kmeans class Dataset(torch.utils.data.Dataset): r""" Dataset that is compatible with :class:`torch.utils.data.Dataset`. This serves as a wrapper for manipulating all data fields with the operating behaviours defined in :class:`Transform`. The data fields of all the instantiated sentences can be accessed as an attribute of the dataset. Args: transform (Transform): An instance of :class:`Transform` and its derivations. The instance holds a series of loading and processing behaviours with regard to the specfic data format. data (list[list] or str): A list of list of strings or a filename. This will be passed into :meth:`transform.load`. kwargs (dict): Keyword arguments that will be passed into :meth:`transform.load` together with `data` to control the loading behaviour. Attributes: transform (Transform): An instance of :class:`Transform`. sentences (list[Sentence]): A list of sentences loaded from the data. Each sentence includes fields obeying the data format defined in ``transform``. """ def __init__(self, transform, data, **kwargs): super(Dataset, self).__init__() self.transform = transform self.sentences = transform.load(data, **kwargs) def __repr__(self): s = f"{self.__class__.__name__}(" s += f"n_sentences={len(self.sentences)}" if hasattr(self, 'loader'): s += f", n_batches={len(self.loader)}" if hasattr(self, 'buckets'): s += f", n_buckets={len(self.buckets)}" s += ")" return s def __len__(self): return len(self.sentences) def __getitem__(self, index): if not hasattr(self, 'fields'): raise RuntimeError("The fields are not numericalized. Please build the dataset first.") for d in self.fields.values(): yield d[index] def __getattr__(self, name): if name in self.__dict__: return self.__dict__[name] return [getattr(sentence, name) for sentence in self.sentences] def __setattr__(self, name, value): if 'sentences' in self.__dict__ and name in self.sentences[0]: # restore the order of sequences in the buckets indices = torch.tensor([i for bucket in self.buckets.values() for i in bucket]).argsort() for index, sentence in zip(indices, self.sentences): setattr(sentence, name, value[index]) else: self.__dict__[name] = value def __getstate__(self): # only pickle the Transform object and sentences return {'transform': self.transform, 'sentences': self.sentences} def __setstate__(self, state): self.__dict__.update(state) def collate_fn(self, batch): return {f: d for f, d in zip(self.fields.keys(), zip(*batch))} def build(self, batch_size, n_buckets=1, shuffle=False, distributed=False): # numericalize all fields self.fields = self.transform(self.sentences) # NOTE: the final bucket count is roughly equal to n_buckets self.lengths = [len(i) for i in self.fields[next(iter(self.fields))]] self.buckets = dict(zip(*kmeans(self.lengths, n_buckets))) self.loader = DataLoader(dataset=self, batch_sampler=Sampler(buckets=self.buckets, batch_size=batch_size, shuffle=shuffle, distributed=distributed), collate_fn=self.collate_fn) class DataLoader(torch.utils.data.DataLoader): r""" DataLoader, matching with :class:`Dataset`. """ def __init__(self, *args, **kwargs): super().__init__(*args, **kwargs) def __iter__(self): for batch in super().__iter__(): yield namedtuple('Batch', [f.name for f in batch.keys()])(*[f.compose(d) for f, d in batch.items()]) class Sampler(torch.utils.data.Sampler): r""" Sampler that supports for bucketization and token-level batchification. Args: buckets (dict): A dict that maps each centroid to indices of clustered sentences. The centroid corresponds to the average length of all sentences in the bucket. batch_size (int): Token-level batch size. The resulting batch contains roughly the same number of tokens as ``batch_size``. shuffle (bool): If ``True``, the sampler will shuffle both buckets and samples in each bucket. Default: ``False``. distributed (bool): If ``True``, the sampler will be used in conjunction with :class:`torch.nn.parallel.DistributedDataParallel` that restricts data loading to a subset of the dataset. Default: ``False``. """ def __init__(self, buckets, batch_size, shuffle=False, distributed=False): self.batch_size = batch_size self.shuffle = shuffle self.sizes, self.buckets = zip(*[(size, bucket) for size, bucket in buckets.items()]) # number of chunks in each bucket, clipped by range [1, len(bucket)] self.chunks = [min(len(bucket), max(round(size * len(bucket) / batch_size), 1)) for size, bucket in zip(self.sizes, self.buckets)] self.rank = dist.get_rank() if distributed else 0 self.replicas = dist.get_world_size() if distributed else 1 self.samples = sum(self.chunks) // self.replicas self.epoch = 0 def __iter__(self): g = torch.Generator() g.manual_seed(self.epoch) range_fn = torch.arange # if `shuffle=True`, shuffle both the buckets and samples in each bucket # for distributed training, make sure each process generates the same random sequence at each epoch if self.shuffle: def range_fn(x): return torch.randperm(x, generator=g) total, count = 0, 0 # TODO: more elegant way to deal with uneven data, which we directly discard right now for i in range_fn(len(self.buckets)).tolist(): split_sizes = [(len(self.buckets[i]) - j - 1) // self.chunks[i] + 1 for j in range(self.chunks[i])] # DON'T use `torch.chunk` which may return wrong number of chunks for batch in range_fn(len(self.buckets[i])).split(split_sizes): if count == self.samples: break if total % self.replicas == self.rank: count += 1 yield [self.buckets[i][j] for j in batch.tolist()] total += 1 self.epoch += 1 def __len__(self): return self.samples
7,112
40.354651
120
py
diaparser
diaparser-master/diaparser/utils/parallel.py
# -*- coding: utf-8 -*- import os from random import Random import torch import torch.distributed as dist import torch.nn as nn class DistributedDataParallel(nn.parallel.DistributedDataParallel): def __init__(self, module, **kwargs): super().__init__(module, **kwargs) def __getattr__(self, name): wrapped = super().__getattr__('module') if hasattr(wrapped, name): return getattr(wrapped, name) return super().__getattr__(name) def init_device(device, backend='nccl', host=None, port=None): os.environ['CUDA_VISIBLE_DEVICES'] = device if torch.cuda.device_count() > 1: host = host or os.environ.get('MASTER_ADDR', 'localhost') port = port or os.environ.get('MASTER_PORT', str(Random(0).randint(10000, 20000))) os.environ['MASTER_ADDR'] = host os.environ['MASTER_PORT'] = port dist.init_process_group(backend) torch.cuda.set_device(dist.get_rank()) def is_master(): return not dist.is_available() or not dist.is_initialized() or dist.get_rank() == 0
1,071
28.777778
90
py
diaparser
diaparser-master/diaparser/parsers/parser.py
# -*- coding: utf-8 -*- import os from datetime import datetime, timedelta import torch import torch.distributed as dist from .. import parsers from tokenizer.tokenizer import Tokenizer from ..catalog import select from ..utils import Config, Dataset from ..utils.field import Field, BertField from ..utils.logging import init_logger, logger from ..utils.metric import Metric from ..utils.parallel import DistributedDataParallel as DDP from ..utils.parallel import is_master from torch.optim import Adam from torch.optim.lr_scheduler import ExponentialLR def conll_format(path): """ Check whether a file contains data in CoNLL-U format. """ try: with open(path) as f: for line in f: line = line.strip() if line and not line.startswith('#'): # CoNLL-U format has 10 tsv: return len(line.split('\t')) == 10 return False except: return False class Parser(): MODEL = None def __init__(self, args, model, transform): self.args = args self.model = model self.transform = transform def train(self, train, dev, test, buckets=32, batch_size=5000, lr=2e-3, mu=.9, nu=.9, epsilon=1e-12, clip=5.0, decay=.75, decay_steps=5000, epochs=5000, patience=100, verbose=True, **kwargs): r""" Args: lr (float): learnin rate of adam optimizer. Default: 2e-3. mu (float): beta1 of adam optimizer. Default: .9. nu (float): beta2 of adam optimizer. Default: .9. epsilon (float): epsilon of adam optimizer. Default: 1e-12. buckets (int): number of buckets. Default: 32. epochs (int): number of epochs to train: Default: 5000. patience (int): early stop after these many epochs. Default: 100. """ args = self.args.update(locals()) init_logger(logger, verbose=args.verbose) self.transform.train() if dist.is_initialized(): args.batch_size = args.batch_size // dist.get_world_size() logger.info(f"Load the datasets\n" f"{'train:':6} {train}\n" f"{'dev:':6} {dev}\n") train = Dataset(self.transform, args.train, **args) train.build(args.batch_size, args.buckets, True, dist.is_initialized()) logger.info(f"{'train:':6} {len(train):5} sentences, " f"{len(train.loader):3} batches, " f"{len(train.buckets)} buckets") dev = Dataset(self.transform, args.dev) dev.build(args.batch_size, args.buckets) logger.info(f"{'dev:':6} {len(dev):5} sentences, " f"{len(dev.loader):3} batches, " f"{len(train.buckets)} buckets") if args.test: test = Dataset(self.transform, args.test) test.build(args.batch_size, args.buckets) logger.info(f"{'test:':6} {len(test):5} sentences, " f"{len(test.loader):3} batches, " f"{len(train.buckets)} buckets") else: test = None logger.info(f"Model\n{self.model}\n") if dist.is_initialized(): self.model = DDP(self.model, device_ids=[dist.get_rank()], find_unused_parameters=True) self.optimizer = Adam(self.model.parameters(), args.lr, (args.mu, args.nu), args.epsilon) self.scheduler = ExponentialLR(self.optimizer, args.decay**(1/args.decay_steps)) elapsed = timedelta() best_e, best_metric = 1, Metric() for epoch in range(1, args.epochs + 1): start = datetime.now() logger.info(f"Epoch {epoch} / {args.epochs}:") self._train(train.loader) loss, dev_metric = self._evaluate(dev.loader) logger.info(f"{'dev:':6} - loss: {loss:.4f} - {dev_metric}") if test: loss, test_metric = self._evaluate(test.loader) logger.info(f"{'test:':6} - loss: {loss:.4f} - {test_metric}") t = datetime.now() - start # save the model if it is the best so far if dev_metric > best_metric: best_e, best_metric = epoch, dev_metric if is_master(): self.save(args.path) logger.info(f"{t}s elapsed (saved)\n") else: logger.info(f"{t}s elapsed\n") elapsed += t if epoch - best_e >= args.patience: break logger.info(f"Epoch {best_e} saved") logger.info(f"{'dev:':6} - {best_metric}") if test: loss, metric = self.load(args.path)._evaluate(test.loader) logger.info(f"{'test:':6} - {metric}") logger.info(f"{elapsed}s elapsed, {elapsed / epoch}s/epoch") def evaluate(self, data, buckets=8, batch_size=5000, **kwargs): args = self.args.update(locals()) init_logger(logger, verbose=args.verbose) self.transform.train() logger.info("Loading the data") dataset = Dataset(self.transform, data) dataset.build(args.batch_size, args.buckets) logger.info(f"\n{dataset}") logger.info("Evaluating the dataset") start = datetime.now() loss, metric = self._evaluate(dataset.loader) elapsed = datetime.now() - start logger.info(f"loss: {loss:.4f} - {metric}") logger.info(f"{elapsed}s elapsed, {len(dataset)/elapsed.total_seconds():.2f} Sents/s") return loss, metric def predict(self, data, pred=None, buckets=8, batch_size=5000, prob=False, **kwargs): r""" Parses the data and produces a parse tree for each sentence. Args: data (str or list[list[str]]): input to be parsed: either - a str, that will be tokenized first with the tokenizer for the parser language - a path to a file to be read, either in CoNLL-U format or in plain text if :param text: is supplied. - a list of lists of tokens text (str): optional, specifies that the input data is in plain text in the specified language code. pred (str or file): a path to a file where to write the parsed input in CoNLL-U fprmat. bucket (int): the number of buckets used to group sentences to parallelize matrix computations. batch_size (int): group sentences in batches. prob (bool): whther to return also probabilities for each arc. Return: a Dataset containing the parsed sentence trees. """ args = self.args.update(locals()) init_logger(logger, verbose=args.verbose) self.transform.eval() if args.prob: self.transform.append(Field('probs')) if isinstance(data, str) and (not conll_format(data) or args.text): self.transform.reader = Tokenizer(args.text, dir=args.cache_dir, verbose=args.verbose).reader() logger.info("Loading the data") dataset = Dataset(self.transform, data) dataset.build(args.batch_size, args.buckets) logger.info(f"\n{dataset}") logger.info("Making predictions on the dataset") start = datetime.now() preds = self._predict(dataset.loader) elapsed = datetime.now() - start for name, value in preds.items(): setattr(dataset, name, value) if pred is not None and is_master(): logger.info(f"Saving predicted results to {pred}") self.transform.save(pred, dataset.sentences) logger.info(f"{elapsed}s elapsed, {len(dataset) / elapsed.total_seconds():.2f} Sents/s") return dataset def _train(self, loader): raise NotImplementedError @torch.no_grad() def _evaluate(self, loader): raise NotImplementedError @torch.no_grad() def _predict(self, loader): raise NotImplementedError @classmethod def build(cls, path, **kwargs): raise NotImplementedError @classmethod def load(cls, name_or_path='', lang='en', cache_dir=os.path.expanduser('~/.cache/diaparser'), **kwargs): r""" Loads a parser from a pretrained model. Args: name_or_path (str): - a string with the shortcut name of a pretrained parser listed in ``resource.json`` to load from cache or download, e.g., ``'en_ptb.electra-base'``. - a path to a directory containing a pre-trained parser, e.g., `./<path>/model`. lang (str): A language code, used in alternative to ``name_or_path`` to load the default model for the given language. cache_dir (str): Directory where to cache models. The default value is `~/.cache/diaparser`. kwargs (dict): A dict holding the unconsumed arguments that can be used to update the configurations and initiate the model. Examples: >>> parser = Parser.load('en_ewt.electra-base') >>> parser = Parser.load(lang='en') >>> parser = Parser.load('./ptb.biaffine.dependency.char') """ args = Config(**locals()) args.device = 'cuda' if torch.cuda.is_available() else 'cpu' if os.path.exists(name_or_path): state = torch.load(name_or_path) else: url = select(name=name_or_path, lang=lang, **kwargs) if url is None: raise Exception(f'Could not find a model matching name {name_or_path}') verbose = kwargs.get('verbose', True) state = torch.hub.load_state_dict_from_url(url, model_dir=cache_dir, progress=verbose) cls = getattr(parsers, state['name']) args = state['args'].update(args) model = cls.MODEL(**args) model.load_pretrained(state['pretrained']) model.load_state_dict(state['state_dict'], False) model.to(args.device) transform = state['transform'] if args.feat == 'bert': tokenizer = BertField.tokenizer(args.bert) transform.FORM[1].tokenize = tokenizer.tokenize return cls(args, model, transform) def save(self, path): model = self.model if hasattr(model, 'module'): model = self.model.module args = model.args args.pop('Parser') # dont save parser class object state_dict = {k: v.cpu() for k, v in model.state_dict().items()} pretrained = state_dict.pop('pretrained.weight', None) if args.feat == 'bert': tokenize = self.transform.FORM[1].tokenize # save it self.transform.FORM[1].tokenize = None state = {'name': type(self).__name__, 'args': args, 'state_dict': state_dict, 'pretrained': pretrained, 'transform': self.transform} torch.save(state, path) if args.feat == 'bert': self.transform.FORM[1].tokenize = tokenize # restore
11,478
38.582759
125
py
diaparser
diaparser-master/diaparser/parsers/biaffine_dependency.py
# -*- coding: utf-8 -*- import os import torch import torch.nn as nn from ..models import BiaffineDependencyModel from .parser import Parser from ..utils import Config, Dataset, Embedding from ..utils.common import bos, pad, unk from ..utils.field import Field, SubwordField, BertField from ..utils.fn import ispunct from ..utils.logging import get_logger, progress_bar from ..utils.metric import AttachmentMetric from ..utils.transform import CoNLL from tokenizer.tokenizer import Tokenizer logger = get_logger(__name__) class BiaffineDependencyParser(Parser): r""" The implementation of Biaffine Dependency Parser. References: - Timothy Dozat and Christopher D. Manning. 2017. `Deep Biaffine Attention for Neural Dependency Parsing`_. .. _Deep Biaffine Attention for Neural Dependency Parsing: https://openreview.net/forum?id=Hk95PK9le """ MODEL = BiaffineDependencyModel def __init__(self, *args, **kwargs): super().__init__(*args, **kwargs) if self.args.feat in ('char', 'bert'): self.WORD, self.FEAT = self.transform.FORM else: self.WORD, self.FEAT = self.transform.FORM, self.transform.CPOS self.ARC, self.REL = self.transform.HEAD, self.transform.DEPREL self.puncts = torch.tensor([i for s, i in self.WORD.vocab.stoi.items() if ispunct(s)]).to(self.args.device) def train(self, train, dev, test, buckets=32, batch_size=5000, punct=False, tree=False, proj=False, verbose=True, **kwargs): r""" Args: train/dev/test (list[list] or str): Filenames of the train/dev/test datasets. buckets (int): The number of buckets that sentences are assigned to. Default: 32. batch_size (int): The number of tokens in each batch. Default: 5000. punct (bool): If ``False``, ignores the punctuations during evaluation. Default: ``False``. tree (bool): If ``True``, ensures to output well-formed trees. Default: ``False``. proj (bool): If ``True``, ensures to output projective trees. Default: ``False``. partial (bool): ``True`` denotes the trees are partially annotated. Default: ``False``. verbose (bool): If ``True``, increases the output verbosity. Default: ``True``. kwargs (dict): A dict holding the unconsumed arguments that can be used to update the configurations for training. """ return super().train(**Config().update(locals())) def evaluate(self, data, buckets=8, batch_size=5000, punct=False, tree=True, proj=False, partial=False, verbose=True, **kwargs): r""" Args: data (str): The data for evaluation, both list of instances and filename are allowed. buckets (int): The number of buckets that sentences are assigned to. Default: 32. batch_size (int): The number of tokens in each batch. Default: 5000. punct (bool): If ``False``, ignores the punctuations during evaluation. Default: ``False``. tree (bool): If ``True``, ensures to output well-formed trees. Default: ``False``. proj (bool): If ``True``, ensures to output projective trees. Default: ``False``. partial (bool): ``True`` denotes the trees are partially annotated. Default: ``False``. verbose (bool): If ``True``, increases the output verbosity. Default: ``True``. kwargs (dict): A dict holding the unconsumed arguments that can be used to update the configurations for evaluation. Returns: The loss scalar and evaluation results. """ return super().evaluate(**Config().update(locals())) def predict(self, data, pred=None, buckets=8, batch_size=5000, prob=False, tree=True, proj=False, verbose=False, **kwargs): r""" Args: data (list[list] or str): The data for prediction, both a list of instances and filename are allowed. pred (str): If specified, the predicted results will be saved to the file. Default: ``None``. buckets (int): The number of buckets that sentences are assigned to. Default: 32. batch_size (int): The number of tokens in each batch. Default: 5000. prob (bool): If ``True``, outputs the probabilities. Default: ``False``. tree (bool): If ``True``, ensures to output well-formed trees. Default: ``False``. proj (bool): If ``True``, ensures to output projective trees. Default: ``False``. verbose (bool): If ``True``, increases the output verbosity. Default: ``True``. kwargs (dict): A dict holding the unconsumed arguments that can be used to update the configurations for prediction. Returns: A :class:`~diaparser.utils.Dataset` object that stores the predicted results. """ return super().predict(**Config().update(locals())) def _train(self, loader): self.model.train() bar, metric = progress_bar(loader), AttachmentMetric() for words, feats, arcs, rels in bar: self.optimizer.zero_grad() mask = words.ne(self.WORD.pad_index) # ignore the first token of each sentence mask[:, 0] = 0 s_arc, s_rel = self.model(words, feats) loss = self.model.loss(s_arc, s_rel, arcs, rels, mask, self.args.partial) loss.backward() nn.utils.clip_grad_norm_(self.model.parameters(), self.args.clip) self.optimizer.step() self.scheduler.step() arc_preds, rel_preds = self.model.decode(s_arc, s_rel, mask) if self.args.partial: mask &= arcs.ge(0) # ignore all punctuation if not specified if not self.args.punct: mask &= words.unsqueeze(-1).ne(self.puncts).all(-1) metric(arc_preds, rel_preds, arcs, rels, mask) bar.set_postfix_str(f"lr: {self.scheduler.get_last_lr()[0]:.4e} - loss: {loss:.4f} - {metric}") @torch.no_grad() def _evaluate(self, loader): self.model.eval() total_loss, metric = 0, AttachmentMetric() for words, feats, arcs, rels in loader: mask = words.ne(self.WORD.pad_index) # ignore the first token of each sentence mask[:, 0] = 0 s_arc, s_rel = self.model(words, feats) loss = self.model.loss(s_arc, s_rel, arcs, rels, mask, self.args.partial) arc_preds, rel_preds = self.model.decode(s_arc, s_rel, mask, self.args.tree, self.args.proj) if self.args.partial: mask &= arcs.ge(0) # ignore all punctuation if not specified if not self.args.punct: mask &= words.unsqueeze(-1).ne(self.puncts).all(-1) total_loss += loss.item() metric(arc_preds, rel_preds, arcs, rels, mask) total_loss /= len(loader) return total_loss, metric @torch.no_grad() def _predict(self, loader): self.model.eval() preds = {} arcs, rels, probs = [], [], [] for words, feats in progress_bar(loader): mask = words.ne(self.WORD.pad_index) # ignore the first token of each sentence mask[:, 0] = 0 lens = mask.sum(1).tolist() s_arc, s_rel = self.model(words, feats) arc_preds, rel_preds = self.model.decode(s_arc, s_rel, mask, self.args.tree, self.args.proj) arcs.extend(arc_preds[mask].split(lens)) rels.extend(rel_preds[mask].split(lens)) if self.args.prob: arc_probs = s_arc.softmax(-1) probs.extend([prob[1:i+1, :i+1].cpu() for i, prob in zip(lens, arc_probs.unbind())]) arcs = [seq.tolist() for seq in arcs] rels = [self.REL.vocab[seq.tolist()] for seq in rels] preds = {'arcs': arcs, 'rels': rels} if self.args.prob: preds['probs'] = probs return preds @classmethod def build(cls, path, min_freq=2, fix_len=20, **kwargs): r""" Build a brand-new Parser, including initialization of all data fields and model parameters. Args: path (str): The path of the model to be saved. min_freq (str): The minimum frequency needed to include a token in the vocabulary. Default: 2. fix_len (int): The max length of all subword pieces. The excess part of each piece will be truncated. Required if using CharLSTM/BERT. Default: 20. kwargs (dict): A dict holding the unconsumed arguments. """ args = Config(**locals()) args.device = 'cuda' if torch.cuda.is_available() else 'cpu' os.makedirs(os.path.dirname(path), exist_ok=True) if os.path.exists(path) and not args.build: parser = cls.load(**args) parser.model = cls.MODEL(**parser.args) parser.model.load_pretrained(parser.WORD.embed).to(args.device) return parser logger.info("Building the fields") WORD = Field('words', pad=pad, unk=unk, bos=bos, lower=True) if args.feat == 'char': FEAT = SubwordField('chars', pad=pad, unk=unk, bos=bos, fix_len=args.fix_len) elif args.feat == 'bert': tokenizer = BertField.tokenizer(args.bert) args.max_len = min(args.max_len or tokenizer.max_len, tokenizer.max_len) FEAT = BertField('bert', tokenizer, fix_len=args.fix_len) WORD.bos = FEAT.bos # ensure representations have the same length else: FEAT = Field('tags', bos=bos) ARC = Field('arcs', bos=bos, use_vocab=False, fn=CoNLL.get_arcs) REL = Field('rels', bos=bos) if args.feat in ('char', 'bert'): transform = CoNLL(FORM=(WORD, FEAT), HEAD=ARC, DEPREL=REL) else: transform = CoNLL(FORM=WORD, CPOS=FEAT, HEAD=ARC, DEPREL=REL) train = Dataset(transform, args.train) WORD.build(train, args.min_freq, (Embedding.load(args.embed, args.unk) if args.embed else None)) FEAT.build(train) REL.build(train) # set parameters from data: args.update({ 'n_words': WORD.vocab.n_init, 'pad_index': WORD.pad_index, 'unk_index': WORD.unk_index, 'bos_index': WORD.bos_index, 'n_feats': len(FEAT.vocab), 'n_rels': len(REL.vocab), 'feat_pad_index': FEAT.pad_index, }) logger.info("Features:") logger.info(f" {WORD}") logger.info(f" {FEAT}\n {ARC}\n {REL}") model = cls.MODEL(**args) model.load_pretrained(WORD.embed).to(args.device) return cls(args, model, transform)
11,647
40.6
117
py
diaparser
diaparser-master/diaparser/cmds/cmd.py
# -*- coding: utf-8 -*- import torch from ..utils import Config from ..utils.logging import init_logger, logger from ..utils.parallel import init_device from ..parsers.biaffine_dependency import BiaffineDependencyParser as Parser def parse(argparser): argparser.add_argument('--conf', '-c', help='path to config file') argparser.add_argument('--path', '-p', help='model name or path to model file') argparser.add_argument('--device', '-d', default='-1', help='ID of GPU to use') argparser.add_argument('--seed', '-s', default=1, type=int, help='seed for generating random numbers') argparser.add_argument('--threads', '-t', default=16, type=int, help='max num of threads') argparser.add_argument('--batch-size', default=5000, type=int, help='batch size') argparser.add_argument("--local_rank", type=int, default=-1, help='node rank for distributed training') argparser.add_argument('--quiet', '-q', dest='verbose', action='store_false', help='suppress verbose logs') args, unknown = argparser.parse_known_args() args, _ = argparser.parse_known_args(unknown, args) args = Config(**vars(args)) torch.set_num_threads(args.threads) torch.manual_seed(args.seed) init_device(args.device, args.local_rank) init_logger(logger, f"{args.path}.{args.mode}.log", verbose=args.verbose) logger.info('Configuration:\n' + str(args)) if args.mode == 'train': parser = Parser.build(**args) parser.train(**args) elif args.mode == 'evaluate': parser = Parser.load(args.path) parser.evaluate(**args) elif args.mode == 'predict': parser = Parser.load(args.path, **args) parser.predict(**args)
1,727
43.307692
107
py
diaparser
diaparser-master/tokenizer/tokenizer.py
import stanza import torch import json import os from contextlib import contextmanager from diaparser.catalog import available_processors, download_processors # reference https://github.com/stanfordnlp/stanza/blob/master/stanza/utils/prepare_tokenizer_data.py class Tokenizer: """ Interface to Stanza tokenizers. Args. lang (str): conventional language identifier. dir (str): directory for caching models. verbose (Bool): print download progress. """ def __init__(self, lang, dir=os.path.expanduser('~/.cache/diaparser'), verbose=True): dir += '/tokenizer' # check for custom processors avail_processors = available_processors(lang, dir=dir) avail_preprocessors = avail_processors.keys() & ('tokenize', 'mwt') if avail_preprocessors: processors = {p: avail_processors[p] for p in avail_preprocessors} cached_paths = download_processors(lang, processors, dir) processors = ','.join(avail_preprocessors) try: # get Stanza resource.json which is needed by stanza.Pipeline(). stanza.download(lang='', model_dir=dir, verbose=verbose) except: pass # discard exception for unknown lang='' else: cached_paths = {} processors='tokenize' stanza.download(lang, model_dir=dir, processors=processors, verbose=verbose) try: stanza.download(lang, model_dir=dir, processors='mwt', verbose=verbose) processors += ',mwt' except: pass use_gpu = torch.cuda.is_available() self.pipeline = stanza.Pipeline(lang, dir=dir, processors=processors, verbose=verbose, use_gpu=use_gpu, **cached_paths) def predict(self, text): return self.pipeline(text).sentences def format(self, sentences): """ Convert sentences to CoNLL format. """ empty_fields = '\t_' * 8 for i, sentence in enumerate(sentences): yield f'# sent_id = {i+1}' sent_text = sentence.text.replace("\n", " ") yield f'# text = {sent_text}' for token in sentence.tokens: # multiword if len(token.words) > 1: token_range = f'{token.id[0]}-{token.id[-1]}' yield f'{token_range}\t{token.text + empty_fields}' for word in token.words: yield f'{word.id}\t{word.text + empty_fields}' else: yield f'{token.id[0]}\t{token.text + empty_fields}' yield '' def reader(self): """ Reading function that returns a generator of CoNLL-U sentences. """ @contextmanager def generator(data): """ Args: data (str): could be a filename or the text to tokenize. Returns: a context manager that can be used in a `with` contruct, yielding each line of the tokenized `data`. """ if not os.path.exists(data): yield self.format(self.predict(data)) else: with open(data) as f: yield self.format(self.predict(f.read())) return generator if __name__ == '__main__': import sys tokenizer = Tokenizer(sys.argv[1]) # language code, e.g. 'it' sentences = tokenizer.predict(sys.argv[2]) # text to tokenize. print('\n'.join(tokenizer.format(sentences)))
3,617
37.084211
100
py
diaparser
diaparser-master/docs/source/conf.py
# Configuration file for the Sphinx documentation builder. # # This file only contains a selection of the most common options. For a full # list see the documentation: # https://www.sphinx-doc.org/en/master/usage/configuration.html # -- Path setup -------------------------------------------------------------- # If extensions (or modules to document with autodoc) are in another directory, # add these directories to sys.path here. If the directory is relative to the # documentation root, use os.path.abspath to make it absolute, like shown here. # # import os # import sys # sys.path.insert(0, os.path.abspath('.')) import diaparser # -- Project information ----------------------------------------------------- project = 'DiaParser' copyright = '2020, Yu Zhang, Giuseppe Attardi' author = 'Yu Zhang, Giuseppe Attardi' # The short X.Y version version = diaparser.__version__ # The full version, including alpha/beta/rc tags release = diaparser.__version__ # -- General configuration --------------------------------------------------- # Add any Sphinx extension module names here, as strings. They can be # extensions coming with Sphinx (named 'sphinx.ext.*') or your custom # ones. extensions = ['sphinx.ext.autodoc', 'sphinx.ext.autosummary', 'sphinx.ext.intersphinx', 'sphinx.ext.mathjax', 'sphinx.ext.napoleon', 'sphinx.ext.todo', 'sphinx.ext.viewcode'] # Add any paths that contain templates here, relative to this directory. templates_path = ['_templates'] # The suffix(es) of source filenames. # You can specify multiple suffix as a list of string: # # source_suffix = ['.rst', '.md'] source_suffix = '.rst' # The master toctree document. master_doc = 'index' # The language for content autogenerated by Sphinx. Refer to documentation # for a list of supported languages. # # This is also used if you do content translation via gettext catalogs. # Usually you set "language" from the command line for these cases. language = None # List of patterns, relative to source directory, that match files and # directories to ignore when looking for source files. # This pattern also affects html_static_path and html_extra_path. exclude_patterns = [] # The name of the Pygments (syntax highlighting) style to use. pygments_style = 'sphinx' # -- Options for HTML output ------------------------------------------------- # The theme to use for HTML and HTML Help pages. See the documentation for # a list of builtin themes. # html_theme = 'sphinx_rtd_theme' # Add any paths that contain custom static files (such as style sheets) here, # relative to this directory. They are copied after the builtin static files, # so a file named "default.css" will overwrite the builtin "default.css". html_static_path = [] # Example configuration for intersphinx: refer to the Python standard library. intersphinx_mapping = { 'nltk': ('https://www.nltk.org', None), 'numpy': ('https://numpy.org/doc/stable', None), 'python': ('https://docs.python.org/3', None), 'torch': ('https://pytorch.org/docs/stable', None) } autodoc_member_order = 'bysource'
3,158
33.336957
79
py
EditNTS
EditNTS-master/main.py
#!/usr/bin/env python # coding:utf8 from __future__ import print_function import argparse import collections import logging import numpy as np import torch import torch.nn as nn import data from checkpoint import Checkpoint from editnts import EditNTS from evaluator import Evaluator PAD = 'PAD' # This has a vocab id, which is used to represent out-of-vocabulary words [0] UNK = 'UNK' # This has a vocab id, which is used to represent out-of-vocabulary words [1] KEEP = 'KEEP' # This has a vocab id, which is used for copying from the source [2] DEL = 'DEL' # This has a vocab id, which is used for deleting the corresponding word [3] START = 'START' # this has a vocab id, which is uded for indicating start of the sentence for decoding [4] STOP = 'STOP' # This has a vocab id, which is used to stop decoding [5] PAD_ID = 0 # This has a vocab id, which is used to represent out-of-vocabulary words [0] UNK_ID = 1 # This has a vocab id, which is used to represent out-of-vocabulary words [1] KEEP_ID = 2 # This has a vocab id, which is used for copying from the source [2] DEL_ID = 3 # This has a vocab id, which is used for deleting the corresponding word [3] START_ID = 4 # this has a vocab id, which is uded for indicating start of the sentence for decoding [4] STOP_ID = 5 # This has a vocab id, which is used to stop decoding [5] def sort_by_lens(seq, seq_lengths): seq_lengths_sorted, sort_order = seq_lengths.sort(descending=True) seq_sorted = seq.index_select(0, sort_order) return seq_sorted, seq_lengths_sorted, sort_order def reweigh_batch_loss(target_id_bath): pad_c = 0 unk_c = 0 keep_c = 0 del_c = 0 start_c = 0 stop_c = 0 other_c = 0 new_edits_ids_l = target_id_bath for i in new_edits_ids_l: # start_c += 1 # stop_c += 1 for ed in i: if ed == PAD_ID: pad_c += 1 elif ed == UNK_ID: unk_c += 1 elif ed == KEEP_ID: keep_c += 1 elif ed == DEL_ID: del_c += 1 elif ed == START_ID: start_c +=1 elif ed == STOP_ID: stop_c +=1 else: other_c += 1 NLL_weight = np.zeros(30006) + (1 / other_c+1) NLL_weight[PAD_ID] = 0 # pad NLL_weight[UNK_ID] = 1. / unk_c+1 NLL_weight[KEEP_ID] = 1. / keep_c+1 NLL_weight[DEL_ID] = 1. / del_c+1 NLL_weight[5] = 1. / stop_c+1 NLL_weight_t = torch.from_numpy(NLL_weight).float().cuda() # print(pad_c, unk_c, start_c, stop_c, keep_c, del_c, other_c) return NLL_weight_t def reweight_global_loss(w_add,w_keep,w_del): # keep, del, other, (0, 65304, 246768, 246768, 2781648, 3847848, 2016880) pad,start,stop,keep,del,add NLL_weight = np.ones(30006)+w_add NLL_weight[PAD_ID] = 0 # pad NLL_weight[KEEP_ID] = w_keep NLL_weight[DEL_ID] = w_del return NLL_weight def training(edit_net,nepochs, args, vocab, print_every=100, check_every=500): eval_dataset = data.Dataset(args.data_path + 'val.df.filtered.pos') # load eval dataset evaluator = Evaluator(loss= nn.NLLLoss(ignore_index=vocab.w2i['PAD'], reduction='none')) editnet_optimizer = torch.optim.Adam(edit_net.parameters(), lr=1e-3, weight_decay=1e-6) # scheduler = MultiStepLR(abstract_optimizer, milestones=[20,30,40], gamma=0.1) # abstract_scheduler = ReduceLROnPlateau(abstract_optimizer, mode='max') # uncomment this part to re-weight different operations # NLL_weight = reweight_global_loss(args.w_add, args.w_keep, args.w_del) # NLL_weight_t = torch.from_numpy(NLL_weight).float().cuda() # editnet_criterion = nn.NLLLoss(weight=NLL_weight_t, ignore_index=vocab.w2i['PAD'], reduce=False) editnet_criterion = nn.NLLLoss(ignore_index=vocab.w2i['PAD'], reduction='none') best_eval_loss = 0. # init statistics print_loss = [] # Reset every print_every for epoch in range(nepochs): # scheduler.step() #reload training for every epoch if os.path.isfile(args.data_path+'train.df.filtered.pos'): train_dataset = data.Dataset(args.data_path + 'train.df.filtered.pos') else: # iter chunks and vocab_data train_dataset = data.Datachunk(args.data_path + 'train.df.filtered.pos') for i, batch_df in train_dataset.batch_generator(batch_size=args.batch_size, shuffle=True): # time1 = time.time() prepared_batch, syn_tokens_list = data.prepare_batch(batch_df, vocab, args.max_seq_len) #comp,scpn,simp # a batch of complex tokens in vocab ids, sorted in descending order org_ids = prepared_batch[0] org_lens = org_ids.ne(0).sum(1) org = sort_by_lens(org_ids, org_lens) # inp=[inp_sorted, inp_lengths_sorted, inp_sort_order] # a batch of pos-tags in pos-tag ids for complex org_pos_ids = prepared_batch[1] org_pos_lens = org_pos_ids.ne(0).sum(1) org_pos = sort_by_lens(org_pos_ids, org_pos_lens) out = prepared_batch[2][:, :] tar = prepared_batch[2][:, 1:] simp_ids = prepared_batch[3] editnet_optimizer.zero_grad() output = edit_net(org, out, org_ids, org_pos,simp_ids) ##################calculate loss tar_lens = tar.ne(0).sum(1).float() tar_flat=tar.contiguous().view(-1) loss = editnet_criterion(output.contiguous().view(-1, vocab.count), tar_flat).contiguous() loss[tar_flat == 1] = 0 #remove loss for UNK loss = loss.view(tar.size()) loss = loss.sum(1).float() loss = loss/tar_lens loss = loss.mean() print_loss.append(loss.item()) loss.backward() torch.nn.utils.clip_grad_norm_(edit_net.parameters(), 1.) editnet_optimizer.step() if i % print_every == 0: log_msg = 'Epoch: %d, Step: %d, Loss: %.4f' % ( epoch,i, np.mean(print_loss)) print_loss = [] print(log_msg) # Checkpoint if i % check_every == 0: edit_net.eval() val_loss, bleu_score, sari, sys_out = evaluator.evaluate(eval_dataset, vocab, edit_net,args) log_msg = "epoch %d, step %d, Dev loss: %.4f, Bleu score: %.4f, Sari: %.4f \n" % (epoch, i, val_loss, bleu_score, sari) print(log_msg) if val_loss < best_eval_loss: best_eval_loss = val_loss Checkpoint(model=edit_net, opt=editnet_optimizer, epoch=epoch, step=i, ).save(args.store_dir) print("checked after %d steps"%i) edit_net.train() return edit_net dataset='newsela' def main(): torch.manual_seed(233) logging.basicConfig(level=logging.INFO, format='%(asctime)s [INFO] %(message)s') parser = argparse.ArgumentParser() parser.add_argument('--data_path', type=str,dest='data_path', default='/home/ml/ydong26/data/EditNTS_data/editnet_data/%s/'%dataset, help='Path to train vocab_data') parser.add_argument('--store_dir', action='store', dest='store_dir', default='/home/ml/ydong26/tmp_store/editNTS_%s'%dataset, help='Path to exp storage directory.') parser.add_argument('--vocab_path', type=str, dest='vocab_path', default='../vocab_data/', help='Path contains vocab, embedding, postag_set') parser.add_argument('--load_model', type=str, dest='load_model', default=None, help='Path for loading pre-trained model for further training') parser.add_argument('--vocab_size', dest='vocab_size', default=30000, type=int) parser.add_argument('--batch_size', dest='batch_size', default=32, type=int) parser.add_argument('--max_seq_len', dest='max_seq_len', default=100) parser.add_argument('--epochs', type=int, default=50) parser.add_argument('--hidden', type=int, default=200) parser.add_argument('--lr', type=float, default=1e-4) parser.add_argument('--device', type=int, default=1, help='select GPU') #train_file = '/media/vocab_data/yue/TS/editnet_data/%s/train.df.filtered.pos'%dataset # test='/media/vocab_data/yue/TS/editnet_data/%s/test.df.pos' % args.dataset args = parser.parse_args() torch.cuda.set_device(args.device) # load vocab-related files and init vocab print('*'*10) vocab = data.Vocab() vocab.add_vocab_from_file(args.vocab_path+'vocab.txt', args.vocab_size) vocab.add_embedding(gloveFile=args.vocab_path+'glove.6B.100d.txt') pos_vocab = data.POSvocab(args.vocab_path) #load pos-tags embeddings print('*' * 10) print(args) print("generating config") hyperparams=collections.namedtuple( 'hps', #hyper=parameters ['vocab_size', 'embedding_dim', 'word_hidden_units', 'sent_hidden_units', 'pretrained_embedding', 'word2id', 'id2word', 'pos_vocab_size', 'pos_embedding_dim'] ) hps = hyperparams( vocab_size=vocab.count, embedding_dim=100, word_hidden_units=args.hidden, sent_hidden_units=args.hidden, pretrained_embedding=vocab.embedding, word2id=vocab.w2i, id2word=vocab.i2w, pos_vocab_size=pos_vocab.count, pos_embedding_dim=30 ) print('init editNTS model') edit_net = EditNTS(hps, n_layers=1) edit_net.cuda() if args.load_model is not None: print("load edit_net for further training") ckpt_path = args.load_model ckpt = Checkpoint.load(ckpt_path) edit_net = ckpt.model edit_net.cuda() edit_net.train() training(edit_net, args.epochs, args, vocab) if __name__ == '__main__': import os cwd = os.getcwd() print(cwd) main()
10,186
38.792969
135
py
EditNTS
EditNTS-master/checkpoint.py
from __future__ import print_function import os import time import shutil import torch class Checkpoint(object): """ The Checkpoint class manages the saving and loading of a model during training. It allows training to be suspended and resumed at a later time (e.g. when running on a cluster using sequential jobs). To make a checkpoint, initialize a Checkpoint object with the following args; then call that object's save() method to write parameters to disk. Args: model (seq2seq): seq2seq model being trained optimizer (Optimizer): stores the state of the optimizer epoch (int): current epoch (an epoch is a loop through the full training vocab_data) step (int): number of examples seen within the current epoch input_vocab (Vocabulary): vocabulary for the input language output_vocab (Vocabulary): vocabulary for the output language Attributes: CHECKPOINT_DIR_NAME (str): name of the checkpoint directory TRAINER_STATE_NAME (str): name of the file storing trainer states MODEL_NAME (str): name of the file storing model INPUT_VOCAB_FILE (str): name of the input vocab file OUTPUT_VOCAB_FILE (str): name of the output vocab file """ CHECKPOINT_DIR_NAME = 'checkpoints' TRAINER_STATE_NAME = 'trainer_states.pt' MODEL_NAME = 'model.pt' def __init__(self, model, opt, epoch, step, path=None): self.model = model self.opt = opt self.epoch = epoch self.step = step self._path = path @property def path(self): if self._path is None: raise LookupError("The checkpoint has not been saved.") return self._path def save(self, experiment_dir): """ Saves the current model and related training parameters into a subdirectory of the checkpoint directory. The name of the subdirectory is the current local time in Y_M_D_H_M_S format. Args: experiment_dir (str): path to the experiment root directory Returns: str: path to the saved checkpoint subdirectory """ date_time = time.strftime('%Y_%m_%d_%H_%M_%S', time.localtime()) self._path = os.path.join(experiment_dir, self.CHECKPOINT_DIR_NAME, date_time) path = self._path if os.path.exists(path): shutil.rmtree(path) os.makedirs(path) torch.save({'epoch': self.epoch, 'step': self.step, 'opt': self.opt }, os.path.join(path, self.TRAINER_STATE_NAME)) torch.save(self.model, os.path.join(path, self.MODEL_NAME)) #with open(os.path.join(path, self.INPUT_VOCAB_FILE), 'wb') as fout: # dill.dump(self.input_vocab, fout) #with open(os.path.join(path, self.OUTPUT_VOCAB_FILE), 'wb') as fout: # dill.dump(self.output_vocab, fout) return path @classmethod def load(cls, path): """ Loads a Checkpoint object that was previously saved to disk. Args: path (str): path to the checkpoint subdirectory Returns: checkpoint (Checkpoint): checkpoint object with fields copied from those stored on disk """ if torch.cuda.is_available(): resume_checkpoint = torch.load(os.path.join(path, cls.TRAINER_STATE_NAME)) model = torch.load(os.path.join(path, cls.MODEL_NAME), map_location=lambda storage, loc: storage) model.cuda() # # Load all tensors onto the CPU # torch.load('tensors.pt', map_location=lambda storage, loc: storage) # # Map tensors from GPU 1 to GPU 0 # torch.load('tensors.pt', map_location={'cuda:1': 'cuda:%d'%gpu}) else: resume_checkpoint = torch.load(os.path.join(path, cls.TRAINER_STATE_NAME), map_location=lambda storage, loc: storage) model = torch.load(os.path.join(path, cls.MODEL_NAME), map_location=lambda storage, loc: storage) #model.flatten_parameters() # make RNN parameters contiguous #with open(os.path.join(path, cls.INPUT_VOCAB_FILE), 'rb') as fin: # input_vocab = dill.load(fin) #with open(os.path.join(path, cls.OUTPUT_VOCAB_FILE), 'rb') as fin: # output_vocab = dill.load(fin) opt = resume_checkpoint['opt'] print('the fking model is,', type(model)) return Checkpoint(model=model, opt=opt, epoch=resume_checkpoint['epoch'], step=resume_checkpoint['step'], path=path) @classmethod def get_latest_checkpoint(cls, experiment_path): """ Given the path to an experiment directory, returns the path to the last saved checkpoint's subdirectory. Precondition: at least one checkpoint has been made (i.e., latest checkpoint subdirectory exists). Args: experiment_path (str): path to the experiment directory Returns: str: path to the last saved checkpoint's subdirectory """ checkpoints_path = os.path.join(experiment_path, cls.CHECKPOINT_DIR_NAME) all_times = sorted(os.listdir(checkpoints_path), reverse=True) return os.path.join(checkpoints_path, all_times[0]) @classmethod def get_all_checkpoints(cls, experiment_path): """ Given the path to an experiment directory, returns the path to the last saved checkpoint's subdirectory. Precondition: at least one checkpoint has been made (i.e., latest checkpoint subdirectory exists). Args: experiment_path (str): path to the experiment directory Returns: str: path to the last saved checkpoint's subdirectory """ checkpoints_path = os.path.join(experiment_path, cls.CHECKPOINT_DIR_NAME) all_times = sorted(os.listdir(checkpoints_path)) return [os.path.join(checkpoints_path, ckpt) for ckpt in all_times]
6,074
42.705036
129
py
EditNTS
EditNTS-master/data.py
from collections import Counter import glob import random import struct import csv import pandas as pd import numpy as np import os import torch from torch.autograd import Variable import random import pickle # <s> and </s> are used in the vocab_data files to segment the abstracts into sentences. They don't receive vocab ids. PAD = 'PAD' # This has a vocab id, which is used to represent out-of-vocabulary words [0] UNK = 'UNK' # This has a vocab id, which is used to represent out-of-vocabulary words [1] KEEP = 'KEEP' # This has a vocab id, which is used for copying from the source [2] DEL = 'DEL' # This has a vocab id, which is used for deleting the corresponding word [3] START = 'START' # this has a vocab id, which is uded for indicating start of the sentence for decoding [4] STOP = 'STOP' # This has a vocab id, which is used to stop decoding [5] PAD_ID = 0 # This has a vocab id, which is used to represent out-of-vocabulary words [0] UNK_ID = 1 # This has a vocab id, which is used to represent out-of-vocabulary words [1] KEEP_ID = 2 # This has a vocab id, which is used for copying from the source [2] DEL_ID = 3 # This has a vocab id, which is used for deleting the corresponding word [3] START_ID = 4 # this has a vocab id, which is uded for indicating start of the sentence for decoding [4] STOP_ID = 5 # This has a vocab id, which is used to stop decoding [5] def sent2id(sent,vocab): """ this function transfers a sentence (in list of strings) to an np_array :param sent: sentence in list of strings :param vocab: vocab object :return: sentence in numeric token numbers """ new_sent = np.array([[vocab.w2i[i] if i in vocab.w2i.keys() else vocab.w2i[UNK] for i in sent]]) return new_sent def id2edits(ids,vocab): """ # this function transfers a id sentences of edits to actual edit actions # :param ids: list of ids indicating edits # :param vocab: vocab object # :return: list of actual edits # """ edit_list = [vocab.i2w[i] for i in ids] return edit_list def batchify(data, max_len=100): #max_len cutout defined by human bsz = len(data) try: maxlen_data = max([s.shape[0] for s in data]) except: maxlen_data = max([len(s) for s in data]) maxlen = min(maxlen_data, max_len) batch = np.zeros((bsz, maxlen), dtype=np.int) for i, s in enumerate(data): try: batch[i, :min(s.shape[0],maxlen)] = s[:min(s.shape[0],maxlen)] except: batch[i, :min(len(s), maxlen)] = s[:min(len(s), maxlen)] # batch[i, s.shape[0]:] = 3 return Variable(torch.from_numpy(batch)).cuda() def batchify_start_stop(data, max_len=100,start_id=4,stop_id=5): #max_len cutout defined by human # add start token at the beginning and stop token at the end of each sequence in a batch data = [np.append(s, [stop_id]) for s in data] # stop 3 data = [np.insert(s, 0, start_id) for s in data] # stop 3 bsz = len(data) maxlen_data = max([s.shape[0] for s in data]) maxlen = min(maxlen_data, max_len) batch = np.zeros((bsz, maxlen), dtype=np.int) for i, s in enumerate(data): batch[i, :min(s.shape[0],maxlen)] = s[:min(s.shape[0],maxlen)] # batch[i, s.shape[0]:] = 3 return Variable(torch.from_numpy(batch)).cuda() def batchify_stop(data, max_len=100,start_id=4,stop_id=5): #max_len cutout defined by human # add stop tokens at the end of the sequence in each batch data = [np.append(s, [stop_id]) for s in data] # stop 3 bsz = len(data) maxlen_data = max([s.shape[0] for s in data]) maxlen = min(maxlen_data, max_len) batch = np.zeros((bsz, maxlen), dtype=np.int) for i, s in enumerate(data): batch[i, :min(s.shape[0],maxlen)] = s[:min(s.shape[0],maxlen)] # batch[i, s.shape[0]:] = 3 return Variable(torch.from_numpy(batch)).cuda() class Vocab(): def __init__(self): self.word_list = [PAD, UNK, KEEP, DEL, START, STOP] self.w2i = {} self.i2w = {} self.count = 0 self.embedding = None def add_vocab_from_file(self, vocab_file="../vocab_data/vocab.txt",vocab_size=30000): with open(vocab_file, "rb") as f: for i,line in enumerate(f): if i >=vocab_size: break self.word_list.append(line.split()[0]) # only want the word, not the count print("read %d words from vocab file" % len(self.word_list)) for w in self.word_list: self.w2i[w] = self.count self.i2w[self.count] = w self.count += 1 def add_embedding(self, gloveFile="path_for_glove_embedding", embed_size=100): print("Loading Glove embeddings") with open(gloveFile, 'r') as f: model = {} w_set = set(self.word_list) embedding_matrix = np.zeros(shape=(len(self.word_list), embed_size)) for line in f: splitLine = line.split() word = splitLine[0] if word in w_set: # only extract embeddings in the word_list embedding = np.array([float(val) for val in splitLine[1:]]) model[word] = embedding embedding_matrix[self.w2i[word]] = embedding # if len(model) % 1000 == 0: # print("processed %d vocab_data" % len(model)) self.embedding = embedding_matrix print("%d words out of %d has embeddings in the glove file" % (len(model), len(self.word_list))) class POSvocab(): def __init__(self,vocab_path): self.word_list = [PAD,UNK,START,STOP] self.w2i = {} self.i2w = {} self.count = 0 self.embedding = None with open(vocab_path+'postag_set.p','r') as f: # postag_set is from NLTK tagdict = pickle.load(f) for w in self.word_list: self.w2i[w] = self.count self.i2w[self.count] = w self.count += 1 for w in tagdict: self.w2i[w] = self.count self.i2w[self.count] = w self.count += 1 class Datachunk(): def __init__(self,data_path): self.data_path = data_path self.listdir = os.listdir(self.data_path) random.shuffle(self.listdir) self.idx_count = 0 def example_generator(self,shuffle=True): while len(self.listdir) != 0: print("reading a new chunk with %d chunks remaining" % len(self.listdir)) df = pd.read_pickle(self.data_path + self.listdir.pop()) if shuffle: df = df.sample(frac=1).reset_index(drop=True) print('shuffling the df') for index, row in df.iterrows(): self.idx_count+=1 yield self.idx_count, row def batch_generator(self, batch_size=1, shuffle=True): while len(self.listdir) != 0: # print("reading a new chunk with %d chunks remaining" % len(self.listdir)) df = pd.read_pickle(self.data_path + self.listdir.pop()) if shuffle: df = df.sample(frac=1).reset_index(drop=True) # print('shuffling the df') list_df = [df[i:i + batch_size] for i in range(0, df.shape[0], batch_size)] for df in list_df: self.idx_count += 1 yield self.idx_count, df class Dataset(): def __init__(self,data_path): self.df = pd.read_pickle(data_path) self.idx_count = 0 def example_generator(self): for index, row in self.df.iterrows(): yield index, row def batch_generator(self, batch_size=64, shuffle=True): if shuffle: self.df = self.df.sample(frac=1).reset_index(drop=True) # print('shuffling the df') list_df = [self.df[i:i + batch_size] for i in range(0, self.df.shape[0], batch_size)] for df in list_df: self.idx_count += 1 yield self.idx_count, df def prepare_batch(batch_df,vocab, max_length=100): """ :param example: one row in pandas dataframe with feild ['comp_tokens', 'simp_tokens','comp_ids','simp_ids', 'comp_pos_ids', edit_labels','new_edit_ids'] :param vocab: vocab object for translation :return: inp: original input sentences :return: inp_pos: pos-tag ids for the input sentences :return: tgt: the target edit-labels in ids :return: inp_simp:the corresponding simple sentences in ids :return: batch_df['comp_tokens']:the complex tokens """ inp = batchify_stop(batch_df['comp_ids'], max_len=max_length) inp_pos = batchify_stop(batch_df['comp_pos_ids'], max_len=max_length) inp_simp=batchify_start_stop(batch_df['simp_id'], max_len=max_length) # tgt = batchify_start_stop(batch_df['edit_ids'], max_len=max_length) # edit ids has early stop tgt = batchify_start_stop(batch_df['new_edit_ids'], max_len=max_length) # new_edit_ids do not do early stopping # I think new edit ids do not ave early stopping return [inp, inp_pos, tgt,inp_simp], batch_df['comp_tokens']
9,193
38.459227
160
py
EditNTS
EditNTS-master/evaluator.py
import numpy as np import torch from nltk.translate.bleu_score import * smooth = SmoothingFunction() from SARI import SARIsent import nltk import data nltk.data.path.append("/media/nvme/nltk_data") from label_edits import edit2sent def sort_by_lens(seq, seq_lengths): seq_lengths_sorted, sort_order = seq_lengths.sort(descending=True) seq_sorted = seq.index_select(0, sort_order) return seq_sorted, seq_lengths_sorted, sort_order import nltk def cal_bleu_score(decoded, target): return nltk.translate.bleu_score.sentence_bleu([target], decoded, smoothing_function=nltk.translate.bleu_score.SmoothingFunction().method1) class Evaluator(): """""" def __init__(self, loss, batch_size=64): self.loss = loss self.batch_size = batch_size def evaluate(self, dataset, vocab, model, args, max_edit_steps=50): """ Evaluate a model on given dataset and return performance during training Args: dataset: an object of data.Dataset() model (editNTS model): model to evaluate vocab: an object containing data.Vocab() args: args from the main methods Returns: loss (float): loss of the given model on the given dataset evaluated with teacher forcing sari: computed based on python script """ print_loss, print_loss_tf = [], [] bleu_list = [] ter = 0. sari_list = [] sys_out=[] print('Doing tokenized evaluation') for i, batch_df in dataset.batch_generator(batch_size=1, shuffle=False): model.eval() prepared_batch, syn_tokens_list = data.prepare_batch(batch_df, vocab, args.max_seq_len) # comp,scpn,simp org_ids = prepared_batch[0] org_lens = org_ids.ne(0).sum(1) org = sort_by_lens(org_ids, org_lens) # inp=[inp_sorted, inp_lengths_sorted, inp_sort_order] org_pos_ids = prepared_batch[1] org_pos_lens = org_pos_ids.ne(0).sum(1) org_pos = sort_by_lens(org_pos_ids, org_pos_lens) # inp=[inp_sorted, inp_lengths_sorted, inp_sort_order] out = prepared_batch[2][:, :] tar = prepared_batch[2][:, 1:] simp_ids = prepared_batch[3] # best_seq_list = model.beamsearch(org, out,simp_ids, org_ids, org_pos, 5) output_without_teacher_forcing = model(org, out, org_ids, org_pos, simp_ids,0.0) #can't compute loss for this one, can only do teacher forcing output_teacher_forcing = model(org, out, org_ids, org_pos,simp_ids, 1.0) if True: # the loss on validation is computed based on teacher forcing ##################calculate loss tar_lens = tar.ne(0).sum(1).float() tar_flat = tar.contiguous().view(-1) def compute_loss(output,tar_flat): #this function computes the loss based on model outputs and target in flat loss = self.loss(output.contiguous().view(-1, vocab.count), tar_flat).contiguous() loss[tar_flat == 1] = 0 # remove loss for UNK loss = loss.view(tar.size()) loss = loss.sum(1).float() loss = loss / tar_lens loss = loss.mean() return loss loss_tf = compute_loss(output_teacher_forcing,tar_flat) print_loss_tf.append(loss_tf.item()) # the SARI and BLUE is computed based on model.eval without teacher forcing for j in range(output_without_teacher_forcing.size()[0]): ## write beam search here # try: if True: example = batch_df.iloc[j] example_out = output_without_teacher_forcing[j, :, :] ##GREEDY pred_action = torch.argmax(example_out, dim=1).view(-1).data.cpu().numpy() edit_list_in_tokens = data.id2edits(pred_action, vocab) # ###BEST BEAM # edit_list_in_tokens = vocab_data.id2edits(best_seq_list[0][1:], vocab) greedy_decoded_tokens = ' '.join(edit2sent(example['comp_tokens'], edit_list_in_tokens)) greedy_decoded_tokens = greedy_decoded_tokens.split('STOP')[0].split(' ') # tgt_tokens_translated = [vocab.i2w[i] for i in example['simp_ids']] sys_out.append(' '.join(greedy_decoded_tokens)) # prt = True if random.random() < 0.01 else False # if prt: # print('*' * 30) # # print('tgt_in_tokens_translated', ' '.join(tgt_tokens_translated)) # print('ORG', ' '.join(example['comp_tokens'])) # print('GEN', ' '.join(greedy_decoded_tokens)) # print('TGT', ' '.join(example['simp_tokens'])) # print('edit_list_in_tokens',edit_list_in_tokens) # print('gold labels', ' '.join(example['edit_labels'])) bleu_list.append(cal_bleu_score(greedy_decoded_tokens, example['simp_tokens'])) # calculate sari comp_string = ' '.join(example['comp_tokens']) simp_string = ' '.join(example['simp_tokens']) gen_string = ' '.join(greedy_decoded_tokens) sari_list.append(SARIsent(comp_string, gen_string, [simp_string])) print('loss_with_teacher_forcing', np.mean(print_loss_tf)) return np.mean(print_loss_tf), np.mean(bleu_list), np.mean(sari_list), sys_out
5,720
45.893443
154
py
EditNTS
EditNTS-master/editnts.py
from __future__ import unicode_literals, print_function, division import random import numpy as np import torch import torch.nn as nn import torch.nn.functional as F from torch.autograd import Variable use_cuda = torch.cuda.is_available() import data from torch.nn.utils.rnn import pack_padded_sequence as pack from torch.nn.utils.rnn import pad_packed_sequence as unpack MAX_LEN =100 PAD = 'PAD' # This has a vocab id, which is used to represent out-of-vocabulary words [0] UNK = 'UNK' # This has a vocab id, which is used to represent out-of-vocabulary words [1] KEEP = 'KEEP' # This has a vocab id, which is used for copying from the source [2] DEL = 'DEL' # This has a vocab id, which is used for deleting the corresponding word [3] START = 'START' # this has a vocab id, which is uded for indicating start of the sentence for decoding [4] STOP = 'STOP' # This has a vocab id, which is used to stop decoding [5] PAD_ID = 0 # This has a vocab id, which is used to represent out-of-vocabulary words [0] UNK_ID = 1 # This has a vocab id, which is used to represent out-of-vocabulary words [1] KEEP_ID = 2 # This has a vocab id, which is used for copying from the source [2] DEL_ID = 3 # This has a vocab id, which is used for deleting the corresponding word [3] START_ID = 4 # this has a vocab id, which is uded for indicating start of the sentence for decoding [4] STOP_ID = 5 # This has a vocab id, which is used to stop decoding [5] def unsort(x_sorted, sorted_order): x_unsort = torch.zeros_like(x_sorted) x_unsort[:, sorted_order,:] = x_sorted return x_unsort class EncoderRNN(nn.Module): def __init__(self, vocab_size, embedding_dim, pos_vocab_size, pos_embedding_dim,hidden_size, n_layers=1, embedding=None, embeddingPOS=None,dropout=0.3): super(EncoderRNN, self).__init__() self.n_layers = n_layers self.hidden_size = hidden_size if embedding is None: self.embedding = nn.Embedding(vocab_size, embedding_dim) else: self.embedding = embedding if embeddingPOS is None: self.embeddingPOS = nn.Embedding(pos_vocab_size, pos_embedding_dim) else: self.embeddingPOS = embeddingPOS self.rnn = nn.LSTM(embedding_dim+pos_embedding_dim, hidden_size, num_layers=n_layers, batch_first=True, bidirectional=True) self.drop = nn.Dropout(dropout) def forward(self, inp, inp_pos, hidden): #inp and inp pose should be both sorted inp_sorted=inp[0] inp_lengths_sorted=inp[1] inp_sort_order=inp[2] inp_pos_sorted = inp_pos[0] inp_pos_lengths_sorted = inp_pos[1] inp_pos_sort_order = inp_pos[2] emb = self.embedding(inp_sorted) emb_pos = self.embeddingPOS(inp_pos_sorted) embed_cat = torch.cat((emb,emb_pos),dim=2) packed_emb = pack(embed_cat, inp_lengths_sorted,batch_first=True) memory_bank, encoder_final = self.rnn(packed_emb, hidden) memory_bank = unpack(memory_bank)[0] memory_bank = unsort(memory_bank, inp_sort_order) h_unsorted=unsort(encoder_final[0], inp_sort_order) c_unsorted=unsort(encoder_final[1], inp_sort_order) return memory_bank.transpose(0,1), (h_unsorted,c_unsorted) def initHidden(self, bsz): weight = next(self.parameters()).data return Variable(weight.new(self.n_layers * 2, bsz, self.hidden_size).zero_()), \ Variable(weight.new(self.n_layers * 2, bsz, self.hidden_size).zero_()) class EditDecoderRNN(nn.Module): def __init__(self, vocab_size, embedding_dim, hidden_size, n_layers=1, embedding=None): super(EditDecoderRNN, self).__init__() self.hidden_size = hidden_size self.embedding_dim = embedding_dim self.vocab_size = vocab_size self.n_layers = n_layers if embedding is None: self.embedding = nn.Embedding(vocab_size, embedding_dim) else: self.embedding = embedding self.rnn_edits = nn.LSTM(embedding_dim, hidden_size, num_layers=n_layers, batch_first=True) self.rnn_words = nn.LSTM(embedding_dim, hidden_size, num_layers=n_layers, batch_first=True) self.attn_Projection_org = nn.Linear(hidden_size, hidden_size, bias=False) # self.attn_Projection_scpn = nn.Linear(hidden_size, hidden_size, bias=False) #hard attention here self.attn_MLP = nn.Sequential(nn.Linear(hidden_size * 4, embedding_dim), nn.Tanh()) self.out = nn.Linear(embedding_dim, self.vocab_size) self.out.weight.data = self.embedding.weight.data[:self.vocab_size] def execute(self, symbol, input, lm_state): """ :param symbol: token_id for predicted edit action (in teacher forcing mode, give the true one) :param input: the word_id being editted currently :param lm_state: last lstm state :return: """ # predicted_symbol = KEEP -> feed input to RNN_LM # predicted_symbol = DEL -> do nothing, return current lm_state # predicted_symbol = new word -> feed that word to RNN_LM is_keep = torch.eq(symbol, data.KEEP_ID) is_del = torch.eq(symbol, data.DEL_ID) if is_del: return lm_state elif is_keep: # return lstm with kept word learned in lstm _, new_lm_state = self.rnn_words(self.embedding(input.view(-1, 1)), lm_state) else: #consider as insert here # print(symbol.item()) input = self.embedding(symbol.view(-1,1)) _, new_lm_state = self.rnn_words(input,lm_state) return new_lm_state def execute_batch(self, batch_symbol, batch_input, batch_lm_state): batch_h = batch_lm_state[0] batch_c = batch_lm_state[1] bsz = batch_symbol.size(0) unbind_new_h = [] unbind_new_c = [] # unbind all batch inputs unbind_symbol = torch.unbind(batch_symbol,dim=0) unbind_input = torch.unbind(batch_input,dim=0) unbind_h = torch.unbind(batch_h,dim=1) unbind_c = torch.unbind(batch_c,dim=1) for i in range(bsz): elem=self.execute(unbind_symbol[i], unbind_input[i], (unbind_h[i].view(1,1,-1), unbind_c[i].view(1,1,-1))) unbind_new_h.append(elem[0]) unbind_new_c.append(elem[1]) new_batch_lm_h = torch.cat(unbind_new_h,dim=1) new_batch_lm_c = torch.cat(unbind_new_c,dim=1) return (new_batch_lm_h,new_batch_lm_c) def forward(self, input_edits, hidden_org,encoder_outputs_org, org_ids, simp_sent,teacher_forcing_ratio=1.): #input_edits and simp_sent need to be padded with START bsz, nsteps = input_edits.size() # revisit each word and then decide the action, for each action, do the modification and calculate rouge difference use_teacher_forcing = True if random.random() < teacher_forcing_ratio else False decoder_out = [] counter_for_keep_del = np.zeros(bsz, dtype=int) counter_for_keep_ins =np.zeros(bsz, dtype=int) # decoder in the training: if use_teacher_forcing: embedded_edits = self.embedding(input_edits) output_edits, hidden_edits = self.rnn_edits(embedded_edits, hidden_org) embedded_words = self.embedding(simp_sent) output_words, hidden_words = self.rnn_words(embedded_words, hidden_org) key_org = self.attn_Projection_org(output_edits) # bsz x nsteps x nhid MIGHT USE WORD HERE logits_org = torch.bmm(key_org, encoder_outputs_org.transpose(1, 2)) # bsz x nsteps x encsteps attn_weights_org = F.softmax(logits_org, dim=-1) # bsz x nsteps x encsteps attn_applied_org = torch.bmm(attn_weights_org, encoder_outputs_org) # bsz x nsteps x nhid for t in range(nsteps-1): # print(t) decoder_output_t = output_edits[:, t:t + 1, :] attn_applied_org_t = attn_applied_org[:, t:t + 1, :] ## find current word inds = torch.LongTensor(counter_for_keep_del) dummy = inds.view(-1, 1, 1) dummy = dummy.expand(dummy.size(0), dummy.size(1), encoder_outputs_org.size(2)).cuda() c = encoder_outputs_org.gather(1, dummy) inds = torch.LongTensor(counter_for_keep_ins) dummy = inds.view(-1, 1, 1) dummy = dummy.expand(dummy.size(0), dummy.size(1), output_words.size(2)).cuda() c_word = output_words.gather(1, dummy) output_t = torch.cat((decoder_output_t, attn_applied_org_t, c,c_word), 2) # bsz*nsteps x nhid*2 output_t = self.attn_MLP(output_t) output_t = F.log_softmax(self.out(output_t), dim=-1) decoder_out.append(output_t) # interpreter's output from lm gold_action = input_edits[:, t + 1].data.cpu().numpy() # might need to realign here because start added counter_for_keep_del = [i[0] + 1 if i[1] == 2 or i[1] == 3 else i[0] for i in zip(counter_for_keep_del, gold_action)] counter_for_keep_ins = [i[0] + 1 if i[1] != DEL_ID and i[1] != STOP_ID and i[1] != PAD_ID else i[0] for i in zip(counter_for_keep_ins, gold_action)] check1 = sum([x >= org_ids.size(1) for x in counter_for_keep_del]) check2 = sum([x >= simp_sent.size(1) for x in counter_for_keep_ins]) if check1 or check2: # print(org_ids.size(1)) # print(counter_for_keep_del) break else: # no teacher forcing decoder_input_edit = input_edits[:, :1] decoder_input_word=simp_sent[:,:1] t, tt = 0, max(MAX_LEN,input_edits.size(1)-1) # initialize embedded_edits = self.embedding(decoder_input_edit) output_edits, hidden_edits = self.rnn_edits(embedded_edits, hidden_org) embedded_words = self.embedding(decoder_input_word) output_words, hidden_words = self.rnn_words(embedded_words, hidden_org) # # # give previous word from tgt simp_sent # inds = torch.LongTensor(counter_for_keep_ins) # dummy = inds.view(-1, 1, 1) # dummy = dummy.expand(dummy.size(0), dummy.size(1), output_words.size(2)).cuda() # c_word = output_words.gather(1, dummy) while t < tt: if t>0: embedded_edits = self.embedding(decoder_input_edit) output_edits, hidden_edits = self.rnn_edits(embedded_edits, hidden_edits) key_org = self.attn_Projection_org(output_edits) # bsz x nsteps x nhid logits_org = torch.bmm(key_org, encoder_outputs_org.transpose(1, 2)) # bsz x nsteps x encsteps attn_weights_org_t = F.softmax(logits_org, dim=-1) # bsz x nsteps x encsteps attn_applied_org_t = torch.bmm(attn_weights_org_t, encoder_outputs_org) # bsz x nsteps x nhid ## find current word inds = torch.LongTensor(counter_for_keep_del) dummy = inds.view(-1, 1, 1) dummy = dummy.expand(dummy.size(0), dummy.size(1), encoder_outputs_org.size(2)).cuda() c = encoder_outputs_org.gather(1, dummy) output_t = torch.cat((output_edits, attn_applied_org_t, c, hidden_words[0]), 2) # bsz*nsteps x nhid*2 output_t = self.attn_MLP(output_t) output_t = F.log_softmax(self.out(output_t), dim=-1) decoder_out.append(output_t) decoder_input_edit=torch.argmax(output_t,dim=2) # gold_action = input[:, t + 1].vocab_data.cpu().numpy() # might need to realign here because start added pred_action= torch.argmax(output_t,dim=2) counter_for_keep_del = [i[0] + 1 if i[1] == 2 or i[1] == 3 or i[1] == 5 else i[0] for i in zip(counter_for_keep_del, pred_action)] # update rnn_words # find previous generated word # give previous word from tgt simp_sent dummy_2 = inds.view(-1, 1).cuda() org_t = org_ids.gather(1, dummy_2) hidden_words = self.execute_batch(pred_action, org_t, hidden_words) # we give the editted subsequence # hidden_words = self.execute_batch(pred_action, org_t, hidden_org) #here we only give the word t += 1 check = sum([x >= org_ids.size(1) for x in counter_for_keep_del]) if check: break return torch.cat(decoder_out, dim=1), hidden_edits def initHidden(self, bsz): weight = next(self.parameters()).data return Variable(weight.new(self.n_layers, bsz, self.hidden_size).zero_()), \ Variable(weight.new(self.n_layers, bsz, self.hidden_size).zero_()) def beam_forwad_step(self,decoder_input_edits,hidden_edits,hidden_words, org_ids,encoder_outputs_org,counter_for_keep_del,beam_size=5): #buffers: each with k elements for next step decoder_input_k=[] hidden_edits_k=[] counter_for_keep_del_k=[] prob_k=[] hidden_words_k=[] # given decoder hidden, forward one step embedded = self.embedding(decoder_input_edits) decoder_output_t, hidden_edits = self.rnn_edits(embedded, hidden_edits) key_org = self.attn_Projection_org(decoder_output_t) # bsz x nsteps x nhid logits_org = torch.bmm(key_org, encoder_outputs_org.transpose(1, 2)) # bsz x nsteps x encsteps attn_weights_org_t = F.softmax(logits_org, dim=-1) # bsz x nsteps x encsteps attn_applied_org_t = torch.bmm(attn_weights_org_t, encoder_outputs_org) # bsz x nsteps x nhid ## find current word inds = torch.LongTensor(counter_for_keep_del) dummy = inds.view(-1, 1, 1) dummy = dummy.expand(dummy.size(0), dummy.size(1), encoder_outputs_org.size(2)).cuda() c = encoder_outputs_org.gather(1, dummy) output_t = torch.cat((decoder_output_t, attn_applied_org_t, c, hidden_words[0]), 2) # bsz*nsteps x nhid*2 output_t = self.attn_MLP(output_t) output_t = F.log_softmax(self.out(output_t), dim=-1) # update rnn_words # find previous generated word # give previous word from tgt simp_sent topv, topi = torch.topk(output_t,beam_size, dim=2) for b in range(beam_size): prob_t_k=topv[:,:,b] out_id_t_k=topi[:,:,b] counter_for_keep_del = [i[0] + 1 if i[1] == 2 or i[1] == 3 or i[1] == 5 else i[0] for i in zip(counter_for_keep_del, out_id_t_k)] dummy_2 = inds.view(-1, 1).cuda() org_t = org_ids.gather(1, dummy_2) hidden_words = self.execute_batch(out_id_t_k, org_t, hidden_words) # input[:, t + 1]=gold action, decoder_input_k.append(out_id_t_k) hidden_edits_k.append(hidden_edits) prob_k.append(prob_t_k) hidden_words_k.append(hidden_words) counter_for_keep_del_k.append(counter_for_keep_del) return decoder_input_k,hidden_edits_k,hidden_words_k,prob_k,counter_for_keep_del_k def beam_forward(self, input_edits, simp_sent, hidden_org,encoder_outputs_org, org_ids, beam_size=5): # initialize for beam search bsz, nsteps = input_edits.size() # decoder_out = [] counter_for_keep_del = np.zeros(bsz, dtype=int) # decoder_input = input[:, :1] t, tt = 0, max(MAX_LEN, input_edits.size(1) - 1) # embedded = self.embedding(decoder_input) # output, hidden = self.rnn(embedded, hidden_org) # initialize for beam list best_k_seqs = [[input_edits[:, :1]]] best_k_probs = [0.] best_k_hidden_edits = [hidden_org] best_k_hidden_words=[hidden_org] best_k_counters =[counter_for_keep_del] while t < tt: # print(t) next_best_k_squared_seq = [] next_best_k_squared_probs = [] next_best_k_squared_counters = [] next_best_k_squared_hidden_edits = [] next_best_k_squared_hidden_words = [] for b in range(len(best_k_seqs)): seq = best_k_seqs[b] prob = best_k_probs[b] counter = best_k_counters[b] hidden_edits = best_k_hidden_edits[b] hidden_words = best_k_hidden_words[b] check = sum([x >= org_ids.size(1) for x in counter]) if seq[-1].item() == STOP_ID or check: # if end of token, make sure no children next_best_k_squared_seq.append(seq) next_best_k_squared_probs.append(prob) next_best_k_squared_counters.append(counter) next_best_k_squared_hidden_edits.append(hidden_edits) next_best_k_squared_hidden_words.append(hidden_words) else: # append the top k children decoder_input_k, hidden_edits_k,hidden_words_k, prob_k, counter_for_keep_del_k=self.beam_forwad_step(seq[-1], hidden_edits,hidden_words,org_ids,encoder_outputs_org,counter,beam_size) for i in range(beam_size): next_seq = seq[:] next_seq.append(decoder_input_k[i]) next_best_k_squared_seq.append(next_seq) next_best_k_squared_probs.append(prob + prob_k[i].item()) next_best_k_squared_counters.append(counter_for_keep_del_k[i]) next_best_k_squared_hidden_edits.append(hidden_edits_k[i]) next_best_k_squared_hidden_words.append(hidden_words_k[i]) # contract to the best k indexs = np.argsort(next_best_k_squared_probs)[::-1][:beam_size] best_k_seqs = [next_best_k_squared_seq[i] for i in indexs] best_k_probs = [next_best_k_squared_probs[i] for i in indexs] best_k_counters = [next_best_k_squared_counters[i] for i in indexs] best_k_hidden_edits = [next_best_k_squared_hidden_edits[i] for i in indexs] best_k_hidden_words = [next_best_k_squared_hidden_words[i] for i in indexs] t +=1 return best_k_seqs, best_k_probs, best_k_hidden_edits,best_k_hidden_words,best_k_counters class EditNTS(nn.Module): def __init__(self, config, n_layers=2): super(EditNTS, self).__init__() self.embedding = nn.Embedding(config.vocab_size, config.embedding_dim) if not(config.pretrained_embedding is None): print('load pre-trained embeddings') self.embedding.weight.data.copy_(torch.from_numpy(config.pretrained_embedding)) self.embeddingPOS = nn.Embedding(config.pos_vocab_size, config.pos_embedding_dim) self.encoder1 = EncoderRNN(config.vocab_size, config.embedding_dim, config.pos_vocab_size, config.pos_embedding_dim, config.word_hidden_units, n_layers, self.embedding, self.embeddingPOS) self.decoder = EditDecoderRNN(config.vocab_size, config.embedding_dim, config.word_hidden_units * 2, n_layers, self.embedding) def forward(self,org,output,org_ids,org_pos,simp_sent,teacher_forcing_ratio=1.0): def transform_hidden(hidden): #for bidirectional encoders h, c = hidden h = torch.cat([h[0], h[1]], dim=1)[None, :, :] c = torch.cat([c[0], c[1]], dim=1)[None, :, :] hidden = (h, c) return hidden hidden_org = self.encoder1.initHidden(org[0].size(0)) encoder_outputs_org, hidden_org = self.encoder1(org,org_pos,hidden_org) hidden_org = transform_hidden(hidden_org) logp, _ = self.decoder(output, hidden_org, encoder_outputs_org,org_ids,simp_sent,teacher_forcing_ratio) return logp def beamsearch(self, org, input_edits,simp_sent,org_ids,org_pos, beam_size=5): def transform_hidden(hidden): #for bidirectional encoders h, c = hidden h = torch.cat([h[0], h[1]], dim=1)[None, :, :] c = torch.cat([c[0], c[1]], dim=1)[None, :, :] hidden = (h, c) return hidden hidden_org = self.encoder1.initHidden(org[0].size(0)) encoder_outputs_org, hidden_org = self.encoder1(org,org_pos,hidden_org) hidden_org = transform_hidden(hidden_org) best_k_seqs, best_k_probs, best_k_hidden_edits, best_k_hidden_words, best_k_counters =\ self.decoder.beam_forward(input_edits,simp_sent, hidden_org, encoder_outputs_org,org_ids,beam_size) best_seq_list=[] for sq in best_k_seqs: best_seq_list.append([i.item() for i in sq]) # find final best output index = np.argsort(best_k_probs)[::-1][0] best_seq = best_k_seqs[index] best_seq_np=[i.item() for i in best_seq] return best_seq_list
21,683
46.344978
156
py
patchNR
patchNR-master/train_patchNR.py
# This code belongs to the paper # # F. Altekrüger, A. Denker, P. Hagemann, J. Hertrich, P. Maass and G. Steidl (2023). # PatchNR: Learning from Very Few Images by Patch Normalizing Flow Regularization. # Inverse Problems, vol. 39, no. 6. # # Please cite the paper, if you use the code. # The script trains the patchNR import torch from torch import nn import FrEIA.framework as Ff import FrEIA.modules as Fm import numpy as np import model from tqdm import tqdm import utils DEVICE = torch.device('cuda' if torch.cuda.is_available() else 'cpu') #choose the image class out of image_classes image_classes = ['material', 'lung', 'lentils'] image_class = image_classes[2] print('Image class:' + image_class) if image_class == 'material': example_img = utils.imread('input_imgs/img_learn_material.png') val_img = utils.imread('input_imgs/img_val_material.png') elif image_class == 'lung': from dival import get_standard_dataset dataset = get_standard_dataset('lodopab', impl='astra_cuda') train = dataset.create_torch_dataset(part='train', reshape=((1,1,) + dataset.space[0].shape, (1,1,) + dataset.space[1].shape)) val_set = dataset.create_torch_dataset(part='validation', reshape=((1,1,) + dataset.space[0].shape, (1,1,) + dataset.space[1].shape)) example_img = torch.cat([train[3][1],train[5][1],train[8][1], train[11][1],train[37][1],train[75][1]]).to(DEVICE) val_img = val_set[1][1].to(DEVICE) elif image_class == 'lentils': example_img = utils.imread('input_imgs/img_learn_lentils.png') val_img = utils.imread('input_imgs/img_val_lentils.png') else: print('Image class is not known') exit() if __name__ == '__main__': patch_size = 6 num_layers = 5 subnet_nodes = 512 patchNR = model.create_NF(num_layers, subnet_nodes, dimension=patch_size**2) batch_size = 32 optimizer_steps = 750000 optimizer = torch.optim.Adam(patchNR.parameters(), lr = 1e-4) im2patch = utils.patch_extractor(patch_size=patch_size) for k in tqdm(range(optimizer_steps)): #extract patches idx = np.random.randint(0,example_img.shape[0]) patch_example = im2patch(example_img[idx].unsqueeze(0),batch_size) #compute loss loss = 0 invs, jac_inv = patchNR(patch_example, rev = True) loss += torch.mean(0.5 * torch.sum(invs**2, dim=1) - jac_inv) optimizer.zero_grad() loss.backward() optimizer.step() #validation step if k%1000 ==0: with torch.no_grad(): val_patch = im2patch(val_img,batch_size) invs, jac_inv = patchNR(val_patch, rev = True) val_loss = torch.mean(0.5 * torch.sum(invs**2, dim=1) - jac_inv).item() print(k) print(loss.item()) print(val_loss) #save weights if (k+1) % 50000 == 0: it = int((k+1)/1000) #torch.save({'net_state_dict': patchNR.state_dict()}, 'patchNR_weights/weights_'+image_class + '_'+str(it) + '.pth') torch.save({'net_state_dict': patchNR.state_dict()}, 'patchNR_weights/weights_'+image_class + '.pth')
3,300
36.089888
137
py
patchNR
patchNR-master/patchNR_CT.py
# This code belongs to the paper # # F. Altekrüger, A. Denker, P. Hagemann, J. Hertrich, P. Maass and G. Steidl (2023). # PatchNR: Learning from Very Few Images by Patch Normalizing Flow Regularization. # Inverse Problems, vol. 39, no. 6. # # Please cite the paper, if you use the code. # # The script reproduces the numerical example CT in the paper. import torch from torch import nn import numpy as np import torch.nn.functional as F import model from tqdm import tqdm import utils import dival from dival import get_standard_dataset import odl from odl.contrib.torch import OperatorModule from dival.util.torch_losses import poisson_loss from functools import partial DEVICE = torch.device('cuda' if torch.cuda.is_available() else 'cpu') print(DEVICE) def patchNR(img, lam, patch_size, n_patches_out, patchNR, n_iter_max, operator): """ Defines the reconstruction using patchNR as regularizer """ # fixed parameters obs = img.to(DEVICE) operator = operator center = False init = fbp(obs) pad_size = 4 #pad the image before extracting patches to avoid boundary effects pad = [pad_size]*4 # create patch extractors input_im2pat = utils.patch_extractor(patch_size, pad=False, center=center) # intialize optimizer for image fake_img = torch.tensor(init.clone(),dtype=torch.float,device=DEVICE,requires_grad = True) optim_img = torch.optim.Adam([fake_img], lr=0.005) #define the poisson loss photons_per_pixel = 4096 mu_max = 81.35858 criterion = partial(poisson_loss,photons_per_pixel=photons_per_pixel,mu_max=mu_max) # Main loop for it in tqdm(range(n_iter_max)): optim_img.zero_grad() tmp = nn.functional.pad(fake_img,pad,mode='reflect') fake_data = input_im2pat(tmp,n_patches_out) #patchNR pred_inv, log_det_inv = patchNR(fake_data,rev=True) reg = torch.mean(torch.sum(pred_inv**2,dim=1)/2) - torch.mean(log_det_inv) #data fidelity data_fid = criterion(operator(fake_img),obs) #loss loss = data_fid + lam*reg loss.backward() optim_img.step() return fake_img if __name__ == '__main__': #choose the number of angles angle_types = ['full','limited'] angle_type = angle_types[1] #input parameters patch_size = 6 num_layers = 5 subnet_nodes = 512 #load model net = model.create_NF(num_layers, subnet_nodes, dimension=patch_size**2) weights = torch.load('patchNR_weights/weights_lung.pth') net.load_state_dict(weights['net_state_dict']) #load images dataset = get_standard_dataset('lodopab', impl='astra_cuda') test = dataset.create_torch_dataset(part='test', reshape=((1,1,) + dataset.space[0].shape, (1,1,) + dataset.space[1].shape)) ray_trafo = dataset.ray_trafo if angle_type == 'limited': lim_dataset = dival.datasets.angle_subset_dataset.AngleSubsetDataset(dataset, slice(100,900),impl='astra_cuda') test = lim_dataset.create_torch_dataset(part='test', reshape=((1,1,) + lim_dataset.space[0].shape, (1,1,) + lim_dataset.space[1].shape)) ray_trafo = lim_dataset.ray_trafo gt = test[64][1] obs = test[64][0] gt = test[39][1] obs = test[39][0] #load operator and FBP operator = OperatorModule(ray_trafo).to(DEVICE) fbp = odl.tomo.analytic.filtered_back_projection.fbp_op(ray_trafo, filter_type = 'Hann', frequency_scaling = 0.641025641025641) fbp = OperatorModule(fbp) lam = 700 n_pat = 40000 iteration = 300 if angle_type == 'limited': iteration = 3000 rec = patchNR(obs,lam = lam, patch_size = patch_size, n_patches_out = n_pat, patchNR = net, n_iter_max = iteration, operator = operator) utils.save_img(rec,'results/patchNR_'+angle_type+'_angleCT') #torch.save(rec,'results/patchNR_'+angle_type+'_angleCT_tens.pt')
4,204
32.110236
94
py
patchNR
patchNR-master/utils.py
# This code belongs to the paper # # F. Altekrüger, A. Denker, P. Hagemann, J. Hertrich, P. Maass and G. Steidl (2023). # PatchNR: Learning from Very Few Images by Patch Normalizing Flow Regularization. # Inverse Problems, vol. 39, no. 6. # # Please cite the paper, if you use the code. # The functions are adapted from # # J. Hertrich, A. Houdard and C. Redenbach (2022). # Wasserstein Patch Prior for Image Superresolution. # IEEE Transactions on Computational Imaging. # (https://github.com/johertrich/Wasserstein_Patch_Prior) # # and # # A. Houdard, A. Leclaire, N. Papadakis and J. Rabin. # Wasserstein Generative Models for Patch-based Texture Synthesis. # ArXiv Preprint#2007.03408 # (https://github.com/ahoudard/wgenpatex) import torch from torch import nn import skimage.io as io import numpy as np import math DEVICE = torch.device('cuda' if torch.cuda.is_available() else 'cpu') def imread(img_name): """ loads an image as torch.tensor on the selected device """ np_img = io.imread(img_name) tens_img = torch.tensor(np_img, dtype=torch.float, device=DEVICE) if torch.max(tens_img) > 1: tens_img/=255 if len(tens_img.shape) < 3: tens_img = tens_img.unsqueeze(2) if tens_img.shape[2] > 3: tens_img = tens_img[:,:,:3] tens_img = tens_img.permute(2,0,1) return tens_img.unsqueeze(0) def save_img(tensor_img, name): ''' save img (tensor form) with the name ''' img = np.clip(tensor_img.squeeze().detach().cpu().numpy(),0,1) io.imsave(str(name)+'.png', img) return class gaussian_downsample(nn.Module): """ Downsampling module with Gaussian filtering """ def __init__(self, kernel_size, sigma, stride, pad=False): super(gaussian_downsample, self).__init__() self.gauss = nn.Conv2d(1, 1, kernel_size, stride=stride, groups=1, bias=False) gaussian_weights = self.init_weights(kernel_size, sigma) self.gauss.weight.data = gaussian_weights.to(DEVICE) self.gauss.weight.requires_grad_(False) self.pad = pad self.padsize = kernel_size-1 def forward(self, x): if self.pad: x = torch.cat((x, x[:,:,:self.padsize,:]), 2) x = torch.cat((x, x[:,:,:,:self.padsize]), 3) return self.gauss(x) def init_weights(self, kernel_size, sigma): x_cord = torch.arange(kernel_size) x_grid = x_cord.repeat(kernel_size).view(kernel_size, kernel_size) y_grid = x_grid.t() xy_grid = torch.stack([x_grid, y_grid], dim=-1) mean = (kernel_size - 1)/2. variance = sigma**2. gaussian_kernel = (1./(2.*math.pi*variance))*torch.exp(-torch.sum((xy_grid - mean)**2., dim=-1)/(2*variance)) gaussian_kernel = gaussian_kernel / torch.sum(gaussian_kernel) return gaussian_kernel.view(1, 1, kernel_size, kernel_size) class patch_extractor(nn.Module): """ Module for creating custom patch extractor """ def __init__(self, patch_size, pad=False,center=False): super(patch_extractor, self).__init__() self.im2pat = nn.Unfold(kernel_size=patch_size) self.pad = pad self.padsize = patch_size-1 self.center=center self.patch_size=patch_size def forward(self, input, batch_size=0): if self.pad: input = torch.cat((input, input[:,:,:self.padsize,:]), 2) input = torch.cat((input, input[:,:,:,:self.padsize]), 3) patches = self.im2pat(input).squeeze(0).transpose(1,0) if batch_size > 0: idx = torch.randperm(patches.size(0))[:batch_size] patches = patches[idx,:] if self.center: patches = patches - torch.mean(patches,-1).unsqueeze(-1) return patches
3,773
33.623853
117
py
patchNR
patchNR-master/model.py
# This code belongs to the paper # # F. Altekrüger, A. Denker, P. Hagemann, J. Hertrich, P. Maass and G. Steidl (2023). # PatchNR: Learning from Very Few Images by Patch Normalizing Flow Regularization. # Inverse Problems, vol. 39, no. 6. # # Please cite the paper, if you use the code. import torch from torch import nn import FrEIA.framework as Ff import FrEIA.modules as Fm import numpy as np DEVICE = torch.device('cuda' if torch.cuda.is_available() else 'cpu') def create_NF(num_layers, sub_net_size, dimension): """ Creates the patchNR network """ def subnet_fc(c_in, c_out): return nn.Sequential(nn.Linear(c_in, sub_net_size), nn.ReLU(), nn.Linear(sub_net_size, sub_net_size), nn.ReLU(), nn.Linear(sub_net_size, c_out)) nodes = [Ff.InputNode(dimension, name='input')] for k in range(num_layers): nodes.append(Ff.Node(nodes[-1], Fm.GLOWCouplingBlock, {'subnet_constructor':subnet_fc, 'clamp':1.6}, name=F'coupling_{k}')) nodes.append(Ff.Node(nodes[-1], Fm.PermuteRandom, {'seed':(k+1)}, name=F'permute_flow_{k}')) nodes.append(Ff.OutputNode(nodes[-1], name='output')) model = Ff.ReversibleGraphNet(nodes, verbose=False).to(DEVICE) return model
1,431
33.926829
84
py
patchNR
patchNR-master/patchNR_zeroshot_material.py
# This code belongs to the paper # # F. Altekrüger, A. Denker, P. Hagemann, J. Hertrich, P. Maass and G. Steidl (2023). # PatchNR: Learning from Very Few Images by Patch Normalizing Flow Regularization. # Inverse Problems, vol. 39, no. 6. # # Please cite the paper, if you use the code. # # The script reproduces the zero-shot superresolution example with SiC in the paper. import torch from torch import nn import numpy as np import random from model import create_NF from tqdm import tqdm from utils import * DEVICE = torch.device('cuda' if torch.cuda.is_available() else 'cpu') # downsample operator def Downsample(input_img, scale = 0.25): gaussian_std = 2. kernel_size = 16 gaussian_down = gaussian_downsample(kernel_size,gaussian_std,int(1/scale),pad=False) out = gaussian_down(input_img) return out def train_patchNR(patchNR, img, patch_size, steps, batch_size, center): """ Train the patchNR for the given img (low resolution) """ batch_size = batch_size optimizer_steps = steps center = center optimizer = torch.optim.Adam(patchNR.parameters(), lr = 1e-4) im2patch = patch_extractor(patch_size=patch_size, center = center) patches = torch.empty(0,device=DEVICE) #enlarge training patches by rotation and mirroring for j in range(2): if j == 0: tmp = img elif j == 1: tmp = torch.flip(img,[1]) for i in range(4): patches = torch.cat([patches,im2patch(torch.rot90(tmp,i,[0,1]))]) for k in tqdm(range(optimizer_steps)): #extract patches idx = torch.tensor(random.sample(range(patches.shape[0]),batch_size)) patch_example = patches[idx,:] #compute loss loss = 0 invs, jac_inv = patchNR(patch_example, rev = True) loss += torch.mean(0.5 * torch.sum(invs**2, dim=1) - jac_inv) optimizer.zero_grad() loss.backward() optimizer.step() weights = dict() weights['batch_size'] = batch_size weights['optimizer_steps'] = optimizer_steps weights['patch_size'] = patch_size weights['net_state_dict'] = patchNR.state_dict() #torch.save(weights, 'patchNR_zeroshot_SiC_weights.pt') def patchNR(img, lam, patch_size, n_patches_out, flow, n_iter_max, center, operator): """ Defines the reconstruction using patchNR as regularizer """ # fixed parameters operator = operator center = center init = torch.nn.functional.interpolate(img,scale_factor=4,mode='bicubic') #save_img(init,'bicubic') # create patch extractors input_im2pat = patch_extractor(patch_size, pad=False, center=center) # intialize optimizer for image fake_img = init.clone().detach().requires_grad_(True).to(DEVICE) optim_img = torch.optim.Adam([fake_img], lr=0.01) # Main loop for it in tqdm(range(n_iter_max)): optim_img.zero_grad() tmp = torch.nn.functional.pad(fake_img, pad = (6,6,6,6), mode= 'reflect') fake_data = input_im2pat(tmp,n_patches_out) #patchNR pred_inv, log_det_inv = flow(fake_data,rev=True) reg = torch.mean(torch.sum(pred_inv**2,dim=1)/2) - torch.mean(log_det_inv) #data fidelity data_fid = torch.sum((operator(tmp) - img)**2) #loss loss = data_fid + lam*reg loss.backward() optim_img.step() return fake_img def run_ZeroShot_patchNR(img, load_model = False): patch_size = 6 center = False #params for training train_steps = 10000 batch_size = 128 #params for reco n_pat = 50000 lam = 0.25 n_iter = 60 model = create_NF(num_layers = 5, sub_net_size = 512, dimension=patch_size**2) if load_model: weights = torch.load('patchNR_zeroshot_SiC_weights.pt') patch_size = weights['patch_size'] model.load_state_dict(weights['net_state_dict']) else: train_patchNR(model, img, patch_size, train_steps, batch_size, center = center) reco = patchNR(img, lam, patch_size, n_pat, model, n_iter, center = center, operator = Downsample) return reco if __name__ == '__main__': hr = imread('input_imgs/img_test_material.png') lr = Downsample(hr) lr += 0.01 * torch.randn_like(lr) #save_img(lr,'lr_img') pred = run_ZeroShot_patchNR(lr, load_model = False) save_img(pred,'results/patchNR_zeroshot_material')
4,456
32.261194
102
py
patchNR
patchNR-master/patchNR_deblurring.py
# This code belongs to the paper # # F. Altekrüger, A. Denker, P. Hagemann, J. Hertrich, P. Maass and G. Steidl (2023). # PatchNR: Learning from Very Few Images by Patch Normalizing Flow Regularization. # Inverse Problems, vol. 39, no. 6. # # Please cite the paper, if you use the code. # # The script reproduces the numerical example Deblurring in the paper. import torch from torch import nn import numpy as np import torch.nn.functional as F import model import scipy.io from tqdm import tqdm import utils DEVICE = torch.device('cuda' if torch.cuda.is_available() else 'cpu') print(DEVICE) class Blur(nn.Module): def __init__(self): super().__init__() data_dict = scipy.io.loadmat('im05_flit01.mat') kernels = data_dict['f'] self.kernel = np.array(kernels) self.blur = nn.Conv2d(1,1, self.kernel.shape[0], bias=False, padding='same' ,padding_mode='reflect') self.blur.weight.data = torch.from_numpy(self.kernel).float().unsqueeze(0).unsqueeze(0) def forward(self, x): return self.blur(x) def patchNR(img, lam, patch_size, n_patches_out, patchNR, n_iter_max): """ Defines the reconstruction using patchNR as regularizer """ # fixed parameters obs = img.to(DEVICE) operator = Blur().to(DEVICE) center = False init = obs pad_size = 4 #pad the image before extracting patches to avoid boundary effects pad = [pad_size]*4 # create patch extractors input_im2pat = utils.patch_extractor(patch_size, pad=False, center=center) # intialize optimizer for image fake_img = torch.tensor(init.clone(),dtype=torch.float,device=DEVICE,requires_grad = True) optim_img = torch.optim.Adam([fake_img], lr=0.005) # Main loop for it in tqdm(range(n_iter_max)): optim_img.zero_grad() tmp = nn.functional.pad(fake_img,pad,mode='reflect') fake_data = input_im2pat(tmp,n_patches_out) #patchNR pred_inv, log_det_inv = patchNR(fake_data,rev=True) reg = torch.mean(torch.sum(pred_inv**2,dim=1)/2) - torch.mean(log_det_inv) #data fidelity data_fid = torch.sum((operator(fake_img) - obs)**2) #loss loss = data_fid + lam*reg loss.backward() optim_img.step() return fake_img if __name__ == '__main__': #input parameters patch_size = 6 num_layers = 5 subnet_nodes = 512 net = model.create_NF(num_layers, subnet_nodes, dimension=patch_size**2) weights = torch.load('patchNR_weights/weights_lentils.pth') net.load_state_dict(weights['net_state_dict']) operator = Blur().to(DEVICE) gt = utils.imread('input_imgs/img_test_lentils.png') with torch.no_grad(): noisy = operator(gt) noisy = noisy + 5/255*torch.randn(noisy.shape,device=DEVICE) lam = 0.87 n_pat = 40000 iteration = 600 rec = patchNR(noisy,lam = lam, patch_size = patch_size, n_patches_out = n_pat, patchNR = net, n_iter_max = iteration) utils.save_img(rec,'results/patchNR_lentils')
3,142
29.813725
108
py
patchNR
patchNR-master/patchNR_zeroshot.py
# This code belongs to the paper # # F. Altekrüger, A. Denker, P. Hagemann, J. Hertrich, P. Maass and G. Steidl (2023). # PatchNR: Learning from Very Few Images by Patch Normalizing Flow Regularization. # Inverse Problems, vol. 39, no. 6. # # Please cite the paper, if you use the code. # # The script reproduces the zero-shot superresolution example in the paper. import torch from torch import nn import numpy as np import random from model import create_NF from tqdm import tqdm from utils import * DEVICE = torch.device('cuda' if torch.cuda.is_available() else 'cpu') # downsample operator def Downsample(input_img, scale = 0.5): gaussian_std = 1. kernel_size = 16 gaussian_down = gaussian_downsample(kernel_size,gaussian_std,int(1/scale),pad=False) out = gaussian_down(input_img) return out def train_patchNR(patchNR, img, patch_size, steps, batch_size, center): """ Train the patchNR for the given img (low resolution) """ batch_size = batch_size optimizer_steps = steps center = center optimizer = torch.optim.Adam(patchNR.parameters(), lr = 1e-4) im2patch = patch_extractor(patch_size=patch_size, center = center) patches = torch.empty(0,device=DEVICE) #enlarge training patches by rotation and mirroring for j in range(2): if j == 0: tmp = img elif j == 1: tmp = torch.flip(img,[1]) for i in range(4): patches = torch.cat([patches,im2patch(torch.rot90(tmp,i,[0,1]))]) for k in tqdm(range(optimizer_steps)): #extract patches idx = torch.tensor(random.sample(range(patches.shape[0]),batch_size)) patch_example = patches[idx,:] #compute loss loss = 0 invs, jac_inv = patchNR(patch_example, rev = True) loss += torch.mean(0.5 * torch.sum(invs**2, dim=1) - jac_inv) optimizer.zero_grad() loss.backward() optimizer.step() weights = dict() weights['batch_size'] = batch_size weights['optimizer_steps'] = optimizer_steps weights['patch_size'] = patch_size weights['net_state_dict'] = patchNR.state_dict() #torch.save(weights, 'patchNR_zeroshot_weights.pt') def patchNR(img, lam, patch_size, n_patches_out, flow, n_iter_max, center, operator): """ Defines the reconstruction using patchNR as regularizer """ # fixed parameters operator = operator center = center init = torch.nn.functional.interpolate(img,scale_factor=2,mode='bicubic') #save_img(init,'bicubic') # create patch extractors input_im2pat = patch_extractor(patch_size, pad=False, center=center) # intialize optimizer for image fake_img = init.clone().detach().requires_grad_(True).to(DEVICE) optim_img = torch.optim.Adam([fake_img], lr=0.01) # Main loop for it in tqdm(range(n_iter_max)): optim_img.zero_grad() tmp = torch.nn.functional.pad(fake_img, pad = (7,7,7,7), mode= 'reflect') fake_data = input_im2pat(tmp,n_patches_out) #patchNR pred_inv, log_det_inv = flow(fake_data,rev=True) reg = torch.mean(torch.sum(pred_inv**2,dim=1)/2) - torch.mean(log_det_inv) #data fidelity data_fid = torch.sum((operator(tmp) - img)**2) #loss loss = data_fid + lam*reg loss.backward() optim_img.step() return fake_img def run_ZeroShot_patchNR(img, load_model = False): patch_size = 6 center = False #params for training train_steps = 10000 batch_size = 128 #params for reconstruction n_pat = 80000 lam = 0.25 n_iter = 60 model = create_NF(num_layers = 5, sub_net_size = 512, dimension=patch_size**2) if load_model: weights = torch.load('patchNR_zeroshot_weights.pt') patch_size = weights['patch_size'] model.load_state_dict(weights['net_state_dict']) else: train_patchNR(model, img, patch_size, train_steps, batch_size, center = center) reco = patchNR(img, lam, patch_size, n_pat, model, n_iter, center = center, operator = Downsample) return reco if __name__ == '__main__': hr = imread('input_imgs/img_test_bsd.png') lr = Downsample(hr) lr += 0.01 * torch.randn_like(lr) #save_img(lr,'lr_img') pred = run_ZeroShot_patchNR(lr, load_model = False) save_img(pred,'results/patchNR_zeroshot')
4,439
31.888889
102
py
patchNR
patchNR-master/patchNR_superres.py
# This code belongs to the paper # # F. Altekrüger, A. Denker, P. Hagemann, J. Hertrich, P. Maass and G. Steidl (2023). # PatchNR: Learning from Very Few Images by Patch Normalizing Flow Regularization. # Inverse Problems, vol. 39, no. 6. # # Please cite the paper, if you use the code. # # The script reproduces the numerical example Superresolution in the paper. import torch from torch import nn import numpy as np import torch.nn.functional as F import model from tqdm import tqdm import utils DEVICE = torch.device('cuda' if torch.cuda.is_available() else 'cpu') print(DEVICE) def Downsample(input_img, scale = 0.25): ''' downsamples an img by factor 4 using gaussian downsample from wgenpatex.py ''' if scale > 1: print('Error. Scale factor is larger than 1.') return gaussian_std = 2 kernel_size = 16 gaussian_down = utils.gaussian_downsample(kernel_size,gaussian_std,int(1/scale),pad=True) #gaussian downsample with zero padding out = gaussian_down(input_img).to(DEVICE) return out def patchNR(img, lam, patch_size, n_patches_out, patchNR, n_iter_max): """ Defines the reconstruction using patchNR as regularizer """ # fixed parameters lr_img = img.to(DEVICE) operator = Downsample center = False init = F.interpolate(img,scale_factor=4,mode='bicubic') # create patch extractors input_im2pat = utils.patch_extractor(patch_size, pad=False, center=center) # intialize optimizer for image fake_img = torch.tensor(init.clone(),dtype=torch.float,device=DEVICE,requires_grad = True) optim_img = torch.optim.Adam([fake_img], lr=0.03) # Main loop for it in tqdm(range(n_iter_max)): optim_img.zero_grad() fake_data = input_im2pat(fake_img,n_patches_out) #patchNR pred_inv, log_det_inv = patchNR(fake_data,rev=True) reg = torch.mean(torch.sum(pred_inv**2,dim=1)/2) - torch.mean(log_det_inv) #data fidelity data_fid = torch.sum((operator(fake_img) - lr_img)**2) #loss loss = data_fid + lam*reg loss.backward() optim_img.step() return fake_img if __name__ == '__main__': #input parameters patch_size = 6 num_layers = 5 subnet_nodes = 512 net = model.create_NF(num_layers, subnet_nodes, dimension=patch_size**2) weights = torch.load('patchNR_weights/weights_material.pth') net.load_state_dict(weights['net_state_dict']) hr = utils.imread('input_imgs/img_test_material.png') lr = Downsample(hr) lr = lr + 0.01*torch.randn(lr.shape,device=DEVICE) lam = 0.15 n_pat = 130000 iteration = 300 rec = patchNR(lr,lam = lam, patch_size = patch_size, n_patches_out = n_pat, patchNR = net, n_iter_max = iteration) utils.save_img(rec,'results/patchNR_material')
2,928
28.887755
132
py
AxiCLASS
AxiCLASS-master/external_fz/sphinx-documentation/conf.py
# -*- coding: utf-8 -*- # # DarkAges documentation build configuration file, created by # sphinx-quickstart on Tue Jun 20 22:50:36 2017. # # This file is execfile()d with the current directory set to its # containing dir. # # Note that not all possible configuration values are present in this # autogenerated file. # # All configuration values have a default; values that are commented out # serve to show the default. # If extensions (or modules to document with autodoc) are in another directory, # add these directories to sys.path here. If the directory is relative to the # documentation root, use os.path.abspath to make it absolute, like shown here. # import os import sys sys.path.insert(0, os.path.abspath('..')) # Mock the some modules used from mock import Mock as MagicMock class Mock(MagicMock): @classmethod def __getattr__(cls, name): return Mock() MOCK_MODULES = ['numpy', 'scipy','scipy.interpolate','scipy.integrate','DarkAges'] #sys.modules.update((mod_name, Mock()) for mod_name in MOCK_MODULES) # -- General configuration ------------------------------------------------ # If your documentation needs a minimal Sphinx version, state it here. # # needs_sphinx = '1.0' # Add any Sphinx extension module names here, as strings. They can be # extensions coming with Sphinx (named 'sphinx.ext.*') or your custom # ones. extensions = ['sphinx.ext.autodoc','sphinx.ext.autosummary', #'sphinx.ext.todo', #'sphinx.ext.coverage', #'sphinx.ext.inheritance_diagram', 'sphinx.ext.napoleon', 'sphinx.ext.viewcode', 'sphinx.ext.extlinks', 'sphinx.ext.mathjax'] # Add any paths that contain templates here, relative to this directory. templates_path = ['_templates'] # The suffix(es) of source filenames. # You can specify multiple suffix as a list of string: # # source_suffix = ['.rst', '.md'] source_suffix = '.rst' # The master toctree document. master_doc = 'index' # General information about the project. project = u'DarkAges' copyright = u'2017, Patrick Stöcker' author = u'Patrick Stöcker' # The version info for the project you're documenting, acts as replacement for # |version| and |release|, also used in various other places throughout the # built documents. # with open('../VERSION', 'r') as version_file: # The full version, including alpha/beta/rc tags. release = version_file.readline() # The short X.Y version. version = '.'.join(release.split('.')[:-1]) ## The short X.Y version. #version = u'1.0' ## The full version, including alpha/beta/rc tags. #release = u'1.0' # The language for content autogenerated by Sphinx. Refer to documentation # for a list of supported languages. # # This is also used if you do content translation via gettext catalogs. # Usually you set "language" from the command line for these cases. language = None # List of patterns, relative to source directory, that match files and # directories to ignore when looking for source files. # This patterns also effect to html_static_path and html_extra_path exclude_patterns = ['_build', 'Thumbs.db', '.DS_Store'] # The name of the Pygments (syntax highlighting) style to use. pygments_style = 'sphinx' #show_authors = True # If true, `todo` and `todoList` produce output, else they produce nothing. #todo_include_todos = True # -- Options for HTML output ---------------------------------------------- # The theme to use for HTML and HTML Help pages. See the documentation for # a list of builtin themes. # #html_theme = 'classic' html_theme = 'pyramid' #html_theme = 'nature' #html_theme = 'scrolls' # Theme options are theme-specific and customize the look and feel of a theme # further. For a list of options available for each theme, see the # documentation. # # html_theme_options = {} # Add any paths that contain custom static files (such as style sheets) here, # relative to this directory. They are copied after the builtin static files, # so a file named "default.css" will overwrite the builtin "default.css". html_static_path = ['_static'] # -- Options for HTMLHelp output ------------------------------------------ # Output file base name for HTML help builder. htmlhelp_basename = 'DarkAgesdoc' # -- Options for LaTeX output --------------------------------------------- latex_elements = { # The paper size ('letterpaper' or 'a4paper'). # 'papersize': 'a4paper', # The font size ('10pt', '11pt' or '12pt'). # # 'pointsize': '11pt', # Additional stuff for the LaTeX preamble. # # 'preamble': '', # Latex figure (float) alignment # # 'figure_align': 'htbp', } # Grouping the document tree into LaTeX files. List of tuples # (source start file, target name, title, # author, documentclass [howto, manual, or own class]). latex_documents = [ (master_doc, 'DarkAges.tex', u'DarkAges Documentation', u'Patrick Stöcker', 'manual'), ] # -- Options for manual page output --------------------------------------- # One entry per manual page. List of tuples # (source start file, name, description, authors, manual section). man_pages = [ (master_doc, 'darkages', u'DarkAges Documentation', [author], 1) ] # -- Options for Texinfo output ------------------------------------------- # Grouping the document tree into Texinfo files. List of tuples # (source start file, target name, title, author, # dir menu entry, description, category) texinfo_documents = [ (master_doc, 'DarkAges', u'DarkAges Documentation', author, 'DarkAges', 'One line description of project.', 'Miscellaneous'), ] # Allow duplicate toc entries. #epub_tocdup = True autoclass_content = 'both' autodoc_member_order = 'bysource' autodoc_mock_imports = ['scipy','scipy.interpolate','scipy.integrate'] # Napoleon settings # go to http://sphinxcontrib-napoleon.readthedocs.org/en/latest/ # to see what all this is about napoleon_google_docstring = False napoleon_numpy_docstring = True napoleon_include_init_with_doc = False napoleon_include_private_with_doc = False napoleon_include_special_with_doc = True napoleon_use_admonition_for_examples = False napoleon_use_admonition_for_notes = False napoleon_use_admonition_for_references = False napoleon_use_ivar = True napoleon_use_param = True napoleon_use_rtype = True
6,269
29.585366
82
py
bluebert
bluebert-master/mt-bluebert/mt_bluebert/blue_strip_model.py
""" Usage: blue_strip_model <src> <dest> """ import logging import os import docopt import torch def main(): args = docopt.docopt(__doc__) print(args) if not os.path.exists(args['<src>']): logging.error('%s: Cannot find the model', args['<src>']) map_location = 'cpu' if not torch.cuda.is_available() else None state_dict = torch.load(args['<src>'], map_location=map_location) config = state_dict['config'] if config['ema_opt'] > 0: state = state_dict['ema'] else: state = state_dict['state'] my_state = {k: v for k, v in state.items() if not k.startswith('scoring_list.')} my_config = {k: config[k] for k in ('vocab_size', 'hidden_size', 'num_hidden_layers', 'num_attention_heads', 'hidden_act', 'intermediate_size', 'hidden_dropout_prob', 'attention_probs_dropout_prob', 'max_position_embeddings', 'type_vocab_size', 'initializer_range')} torch.save({'state': my_state, 'config': my_config}, args['<dest>']) if __name__ == '__main__': main()
1,149
29.263158
117
py
bluebert
bluebert-master/mt-bluebert/mt_bluebert/blue_prepro_std.py
""" Preprocessing BLUE dataset. Usage: blue_prepro_std [options] --vocab=<file> --root_dir=<dir> --task_def=<file> --datasets=<str> Options: --do_lower_case --max_seq_len=<int> [default: 128] --overwrite """ # Copyright (c) Microsoft. All rights reserved. # Modified by Yifan Peng import json import logging import os import docopt import numpy as np from pytorch_pretrained_bert.tokenization import BertTokenizer from mt_bluebert.data_utils.log_wrapper import create_logger from mt_bluebert.data_utils.task_def import TaskType, DataFormat, EncoderModelType from mt_bluebert.data_utils.vocab import Vocabulary from mt_bluebert.experiments.squad import squad_utils from mt_bluebert.blue_exp_def import BlueTaskDefs from mt_bluebert.mt_dnn.batcher import BatchGen MAX_SEQ_LEN = 512 def load_data(file_path, data_format: DataFormat, task_type: TaskType, label_dict: Vocabulary = None): """ Args: label_dict: map string label to numbers. only valid for Classification task or ranking task. For ranking task, better label should have large number """ if task_type == TaskType.Ranking: assert data_format == DataFormat.PremiseAndMultiHypothesis rows = [] for line in open(file_path, encoding="utf-8"): fields = line.strip().split("\t") if data_format == DataFormat.PremiseOnly: assert len(fields) == 3 row = {"uid": fields[0], "label": fields[1], "premise": fields[2]} elif data_format == DataFormat.PremiseAndOneHypothesis: assert len(fields) == 4 row = {"uid": fields[0], "label": fields[1], "premise": fields[2], "hypothesis": fields[3]} elif data_format == DataFormat.PremiseAndMultiHypothesis: assert len(fields) > 5 row = {"uid": fields[0], "ruid": fields[1].split(","), "label": fields[2], "premise": fields[3], "hypothesis": fields[4:]} elif data_format == DataFormat.Sequence: assert len(fields) == 4 row = {"uid": fields[0], "label": json.loads(fields[1]), "premise": json.loads(fields[2]), "offset": json.loads(fields[3])} else: raise ValueError(data_format) if task_type == TaskType.Classification: if label_dict is not None: row["label"] = label_dict[row["label"]] else: row["label"] = int(row["label"]) elif task_type == TaskType.Regression: row["label"] = float(row["label"]) elif task_type == TaskType.Ranking: labels = row["label"].split(",") if label_dict is not None: labels = [label_dict[label] for label in labels] else: labels = [float(label) for label in labels] row["label"] = int(np.argmax(labels)) row["olabel"] = labels elif task_type == TaskType.Span: pass # don't process row label elif task_type == TaskType.SequenceLabeling: if label_dict is not None: row["label"] = [label_dict[l] for l in row["label"]] else: row["label"] = [int(l) for l in row["label"]] assert row["label"] is not None, \ '%s: %s: label is None. label_dict: %s' % (file_path, row['uid'], label_dict.tok2ind) rows.append(row) return rows def truncate_seq_pair(tokens_a, tokens_b, max_length): """Truncates a sequence pair in place to the maximum length. Copyed from https://github.com/huggingface/pytorch-pretrained-BERT """ # This is a simple heuristic which will always truncate the longer sequence # one token at a time. This makes more sense than truncating an equal percent # of tokens from each, since if one sequence is very short then each token # that's truncated likely contains more information than a longer sequence. logger = logging.getLogger(__name__) truncate_tokens_a = False truncate_tokens_b = False while True: total_length = len(tokens_a) + len(tokens_b) if total_length <= max_length: break if len(tokens_a) > len(tokens_b): truncate_tokens_a = True tokens_a.pop() else: truncate_tokens_b = True tokens_b.pop() if truncate_tokens_a: logger.debug('%s: longer than %s', tokens_a, max_length) if truncate_tokens_b: logger.debug('%s: longer than %s', tokens_b, max_length) def bert_feature_extractor(text_a, text_b=None, max_seq_length=512, tokenize_fn=None): logger = logging.getLogger(__name__) tokens_a = tokenize_fn.tokenize(text_a) tokens_b = None if text_b: tokens_b = tokenize_fn.tokenize(text_b) if tokens_b: # Modifies `tokens_a` and `tokens_b` in place so that the total # length is less than the specified length. # Account for [CLS], [SEP], [SEP] with "- 3" truncate_seq_pair(tokens_a, tokens_b, max_seq_length - 3) else: # Account for one [SEP] & one [CLS] with "- 2" if len(tokens_a) > max_seq_length - 2: logger.debug('%s: longer than %s', text_a, max_seq_length) tokens_a = tokens_a[:max_seq_length - 2] # The convention in BERT is: # (a) For sequence pairs: # tokens: [CLS] is this jack ##son ##ville ? [SEP] no it is not . [SEP] # type_ids: 0 0 0 0 0 0 0 0 1 1 1 1 1 1 # (b) For single sequences: # tokens: [CLS] the dog is hairy . [SEP] # type_ids: 0 0 0 0 0 0 0 # # Where "type_ids" are used to indicate whether this is the first # sequence or the second sequence. The embedding vectors for `type=0` and # `type=1` were learned during pre-training and are added to the wordpiece # embedding vector (and position vector). This is not *strictly* necessary # since the [SEP] token unambiguously separates the sequences, but it makes # it easier for the model to learn the concept of sequences. # # For classification tasks, the first vector (corresponding to [CLS]) is # used as as the "sentence vector". Note that this only makes sense because # the entire model is fine-tuned. if tokens_b: input_ids = tokenize_fn.convert_tokens_to_ids(['[CLS]'] + tokens_a + ['[SEP]'] + tokens_b + ['[SEP]']) segment_ids = [0] * (len(tokens_a) + 2) + [1] * (len(tokens_b) + 1) if tokens_b: input_ids = tokenize_fn.convert_tokens_to_ids(['[CLS]'] + tokens_b + ['[SEP]'] + tokens_a + ['[SEP]']) segment_ids = [0] * (len(tokens_b) + 2) + [1] * (len(tokens_a) + 1) else: input_ids = tokenize_fn.convert_tokens_to_ids(['[CLS]'] + tokens_a + ['[SEP]']) segment_ids = [0] * len(input_ids) input_mask = None return input_ids, input_mask, segment_ids def split_if_longer(data, label_mapper, max_seq_len=30): rows = [] for idx, sample in enumerate(data): # uid = sample['uid'] docid = sample['uid'].split('.')[0] premise = sample['premise'] label = sample['label'] offset = sample['offset'] while len(premise) > max_seq_len: tmplabel = label[:max_seq_len] for iidx in range(len(tmplabel)): if label_mapper[tmplabel.pop()] == 'O': break p = premise[:len(tmplabel) + 1] l = label[:len(tmplabel) + 1] o = offset[:len(tmplabel) + 1] uid = '{}.{}'.format(docid, o[0].split(';')[0]) rows.append({"uid": uid, "label": l, "premise": p, "offset": o}) premise = premise[len(tmplabel) + 1:] label = label[len(tmplabel) + 1:] offset = offset[len(tmplabel) + 1:] if len(premise) == 0: continue uid = '{}.{}'.format(docid, offset[0].split(';')[0]) rows.append({"uid": uid, "label": label, "premise": premise, "offset": offset}) return rows def build_data_sequence(data, dump_path, max_seq_len=MAX_SEQ_LEN, tokenizer=None, label_mapper=None): with open(dump_path, 'w', encoding='utf-8') as writer: for idx, sample in enumerate(data): ids = sample['uid'] premise = sample['premise'] ne_labels = sample['label'] premise_offset = sample['offset'] tokens = [] labels = [] offsets = [] for i, word in enumerate(premise): subwords = tokenizer.tokenize(word) tokens.extend(subwords) for j in range(len(subwords)): if j == 0: labels.append(ne_labels[i]) offsets.append(premise_offset[i]) else: labels.append(label_mapper['X']) offsets.append('X') if len(premise) > max_seq_len - 2: tokens = tokens[:max_seq_len - 2] labels = labels[:max_seq_len - 2] offsets = offsets[:max_seq_len - 2] label = [label_mapper['CLS']] + labels + [label_mapper['SEP']] offsets = ['X'] + offsets + ['X'] input_ids = tokenizer.convert_tokens_to_ids(['[CLS]'] + tokens + ['[SEP]']) assert len(label) == len(input_ids) type_ids = [0] * len(input_ids) features = {'uid': ids, 'label': label, 'token_id': input_ids, 'type_id': type_ids, 'offset': offsets} writer.write('{}\n'.format(json.dumps(features))) def build_data_premise_only(data, dump_path, max_seq_len=MAX_SEQ_LEN, tokenizer=None, encoderModelType=EncoderModelType.BERT): """Build data of single sentence tasks """ with open(dump_path, 'w', encoding='utf-8') as writer: for idx, sample in enumerate(data): ids = sample['uid'] premise = sample['premise'] label = sample['label'] assert encoderModelType == EncoderModelType.BERT input_ids, _, type_ids = bert_feature_extractor(premise, max_seq_length=max_seq_len, tokenize_fn=tokenizer) features = {'uid': ids, 'label': label, 'token_id': input_ids, 'type_id': type_ids} writer.write('{}\n'.format(json.dumps(features))) def build_data_premise_and_one_hypo( data, dump_path, task_type, max_seq_len=MAX_SEQ_LEN, tokenizer=None, encoderModelType=EncoderModelType.BERT): """Build data of sentence pair tasks """ with open(dump_path, 'w', encoding='utf-8') as writer: for idx, sample in enumerate(data): ids = sample['uid'] premise = sample['premise'] hypothesis = sample['hypothesis'] label = sample['label'] assert encoderModelType == EncoderModelType.BERT input_ids, _, type_ids = bert_feature_extractor( premise, hypothesis, max_seq_length=max_seq_len, tokenize_fn=tokenizer) if task_type == TaskType.Span: seg_a_start = len(type_ids) - sum(type_ids) seg_a_end = len(type_ids) answer_start, answer_end, answer, is_impossible = squad_utils.parse_squad_label(label) span_start, span_end = squad_utils.calc_tokenized_span_range(premise, hypothesis, answer, answer_start, answer_end, tokenizer, encoderModelType) span_start = seg_a_start + span_start span_end = min(seg_a_end, seg_a_start + span_end) answer_tokens = tokenizer.convert_ids_to_tokens(input_ids[span_start:span_end]) if span_start >= span_end: span_start = -1 span_end = -1 features = { 'uid': ids, 'label': is_impossible, 'answer': answer, "answer_tokens": answer_tokens, "token_start": span_start, "token_end": span_end, 'token_id': input_ids, 'type_id': type_ids} else: features = { 'uid': ids, 'label': label, 'token_id': input_ids, 'type_id': type_ids} writer.write('{}\n'.format(json.dumps(features))) def build_data_premise_and_multi_hypo( data, dump_path, max_seq_len=MAX_SEQ_LEN, tokenizer=None, encoderModelType=EncoderModelType.BERT): """Build QNLI as a pair-wise ranking task """ with open(dump_path, 'w', encoding='utf-8') as writer: for idx, sample in enumerate(data): ids = sample['uid'] premise = sample['premise'] hypothesis_1 = sample['hypothesis'][0] hypothesis_2 = sample['hypothesis'][1] label = sample['label'] assert encoderModelType == EncoderModelType.BERT input_ids_1, _, type_ids_1 = bert_feature_extractor( premise, hypothesis_1, max_seq_length=max_seq_len, tokenize_fn=tokenizer) input_ids_2, _, type_ids_2 = bert_feature_extractor( premise, hypothesis_2, max_seq_length=max_seq_len, tokenize_fn=tokenizer) features = { 'uid': ids, 'label': label, 'token_id': [ input_ids_1, input_ids_2], 'type_id': [ type_ids_1, type_ids_2], 'ruid': sample['ruid'], 'olabel': sample['olabel']} writer.write('{}\n'.format(json.dumps(features))) def build_data(data, dump_path, tokenizer, data_format=DataFormat.PremiseOnly, max_seq_len=MAX_SEQ_LEN, encoderModelType=EncoderModelType.BERT, task_type=None, lab_dict=None): # We only support BERT based MRC for now if task_type == TaskType.Span: assert data_format == DataFormat.PremiseAndOneHypothesis assert encoderModelType == EncoderModelType.BERT if data_format == DataFormat.PremiseOnly: build_data_premise_only(data, dump_path, max_seq_len, tokenizer) elif data_format == DataFormat.PremiseAndOneHypothesis: build_data_premise_and_one_hypo(data, dump_path, task_type, max_seq_len, tokenizer) elif data_format == DataFormat.PremiseAndMultiHypothesis: build_data_premise_and_multi_hypo(data, dump_path, max_seq_len, tokenizer) elif data_format == DataFormat.Sequence: build_data_sequence(data, dump_path, max_seq_len, tokenizer, lab_dict) else: raise ValueError(data_format) def main(args): root = args['--root_dir'] assert os.path.exists(root) max_seq_len = int(args['--max_seq_len']) log_file = os.path.join(root, 'blue_prepro_std_{}.log'.format(max_seq_len)) logger = create_logger(__name__, to_disk=True, log_file=log_file) is_uncased = False if 'uncased' in args['--vocab']: is_uncased = True do_lower_case = args['--do_lower_case'] mt_dnn_suffix = 'bert' encoder_model = EncoderModelType.BERT tokenizer = BertTokenizer.from_pretrained(args['--vocab'], do_lower_case=do_lower_case) if is_uncased: mt_dnn_suffix = '{}_uncased'.format(mt_dnn_suffix) else: mt_dnn_suffix = '{}_cased'.format(mt_dnn_suffix) if do_lower_case: mt_dnn_suffix = '{}_lower'.format(mt_dnn_suffix) mt_dnn_root = os.path.join(root, mt_dnn_suffix) if not os.path.isdir(mt_dnn_root): os.mkdir(mt_dnn_root) task_defs = BlueTaskDefs(args['--task_def']) if args['--datasets'] == 'all': tasks = task_defs.tasks else: tasks = args['--datasets'].split(',') for task in tasks: logger.info("Task %s" % task) data_format = task_defs.data_format_map[task] task_type = task_defs.task_type_map[task] label_mapper = task_defs.label_mapper_map[task] split_names = task_defs.split_names_map[task] for split_name in split_names: dump_path = os.path.join(mt_dnn_root, f"{task}_{split_name}.json") if os.path.exists(dump_path) and not args['--overwrite']: logger.warning('%s: Not overwrite %s: %s', task, split_name, dump_path) continue rows = load_data(os.path.join(root, f"{task}_{split_name}.tsv"), data_format, task_type, label_mapper) logger.info('%s: Loaded %s %s samples', task, len(rows), split_name) if task_type == TaskType.SequenceLabeling: rows = split_if_longer(rows, label_mapper, 30) build_data(rows, dump_path, tokenizer, data_format, max_seq_len=max_seq_len, encoderModelType=encoder_model, task_type=task_type, lab_dict=label_mapper) logger.info('%s: Done', task) # combine train and dev for task in tasks: logger.info("Task %s: combine train and dev" % task) dump_path = os.path.join(mt_dnn_root, f"{task}_train+dev.json") if os.path.exists(dump_path) and not args['--overwrite']: logger.warning('%s: Not overwrite train+dev: %s', task, dump_path) continue task_type = task_defs.task_type_map[task] train_path = os.path.join(mt_dnn_root, f"{task}_train.json") train_rows = BatchGen.load(train_path, task_type=task_type) dev_path = os.path.join(mt_dnn_root, f"{task}_dev.json") dev_rows = BatchGen.load(dev_path, task_type=task_type) with open(dump_path, 'w', encoding='utf-8') as fp: for features in train_rows + dev_rows: fp.write('{}\n'.format(json.dumps(features))) logger.info('%s: Done', task) if __name__ == '__main__': args = docopt.docopt(__doc__) main(args)
17,889
41.901679
119
py
bluebert
bluebert-master/mt-bluebert/mt_bluebert/blue_train.py
# Copyright (c) Microsoft. All rights reserved. # Modified by Yifan Peng import argparse import copy import json import os import random from datetime import datetime import numpy as np import torch from pytorch_pretrained_bert.modeling import BertConfig from tensorboardX import SummaryWriter # from experiments.glue.glue_utils import submit, eval_model from mt_bluebert.blue_exp_def import BlueTaskDefs from mt_bluebert.blue_inference import eval_model from mt_bluebert.data_utils.log_wrapper import create_logger from mt_bluebert.data_utils.task_def import EncoderModelType from mt_bluebert.data_utils.utils import set_environment # from torch.utils.tensorboard import SummaryWriter from mt_bluebert.mt_dnn.batcher import BatchGen from mt_bluebert.mt_dnn.model import MTDNNModel def model_config(parser: argparse.ArgumentParser) -> argparse.ArgumentParser: parser.add_argument('--update_bert_opt', default=0, type=int) parser.add_argument('--multi_gpu_on', action='store_true') parser.add_argument('--mem_cum_type', type=str, default='simple', help='bilinear/simple/defualt') parser.add_argument('--answer_num_turn', type=int, default=5) parser.add_argument('--answer_mem_drop_p', type=float, default=0.1) parser.add_argument('--answer_att_hidden_size', type=int, default=128) parser.add_argument('--answer_att_type', type=str, default='bilinear', help='bilinear/simple/defualt') parser.add_argument('--answer_rnn_type', type=str, default='gru', help='rnn/gru/lstm') parser.add_argument('--answer_sum_att_type', type=str, default='bilinear', help='bilinear/simple/defualt') parser.add_argument('--answer_merge_opt', type=int, default=1) parser.add_argument('--answer_mem_type', type=int, default=1) parser.add_argument('--answer_dropout_p', type=float, default=0.1) parser.add_argument('--answer_weight_norm_on', action='store_true') parser.add_argument('--dump_state_on', action='store_true') parser.add_argument('--answer_opt', type=int, default=0, help='0,1') parser.add_argument('--label_size', type=str, default='3') parser.add_argument('--mtl_opt', type=int, default=0) parser.add_argument('--ratio', type=float, default=0) parser.add_argument('--mix_opt', type=int, default=0) parser.add_argument('--max_seq_len', type=int, default=512) parser.add_argument('--init_ratio', type=float, default=1) parser.add_argument('--encoder_type', type=int, default=EncoderModelType.BERT) return parser def data_config(parser: argparse.ArgumentParser) -> argparse.ArgumentParser: parser.add_argument('--log_file', default='mt-dnn-train.log', help='path for log file.') parser.add_argument('--tensorboard', action='store_true') parser.add_argument('--tensorboard_logdir', default='tensorboard_logdir') parser.add_argument("--init_checkpoint", default='mt_dnn_models/bert_model_base.pt', type=str) parser.add_argument('--data_dir', default='blue_data/canonical_data/bert_uncased_lower') parser.add_argument('--data_sort_on', action='store_true') parser.add_argument('--name', default='farmer') parser.add_argument('--task_def', type=str, default="experiments/blue/blue_task_def.yml") parser.add_argument('--train_datasets', default='mnli') parser.add_argument('--test_datasets', default='mnli_mismatched,mnli_matched') return parser def train_config(parser: argparse.ArgumentParser) -> argparse.ArgumentParser: parser.add_argument('--cuda', type=bool, default=torch.cuda.is_available(), help='whether to use GPU acceleration.') parser.add_argument('--log_per_updates', type=int, default=500) parser.add_argument('--save_per_updates', type=int, default=10000) parser.add_argument('--save_per_updates_on', action='store_true') parser.add_argument('--epochs', type=int, default=5) parser.add_argument('--batch_size', type=int, default=8) parser.add_argument('--batch_size_eval', type=int, default=8) parser.add_argument('--optimizer', default='adamax', help='supported optimizer: adamax, sgd, adadelta, adam') parser.add_argument('--grad_clipping', type=float, default=0) parser.add_argument('--global_grad_clipping', type=float, default=1.0) parser.add_argument('--weight_decay', type=float, default=0) parser.add_argument('--learning_rate', type=float, default=5e-5) parser.add_argument('--momentum', type=float, default=0) parser.add_argument('--warmup', type=float, default=0.1) parser.add_argument('--warmup_schedule', type=str, default='warmup_linear') parser.add_argument('--adam_eps', type=float, default=1e-6) parser.add_argument('--vb_dropout', action='store_false') parser.add_argument('--dropout_p', type=float, default=0.1) parser.add_argument('--dropout_w', type=float, default=0.000, help='Randomly drop a fraction drooput_w of training instances.') parser.add_argument('--bert_dropout_p', type=float, default=0.1) # loading parser.add_argument("--model_ckpt", default='checkpoints/model_0.pt', type=str) parser.add_argument("--resume", action='store_true') # EMA parser.add_argument('--ema_opt', type=int, default=0) parser.add_argument('--ema_gamma', type=float, default=0.995) # scheduler parser.add_argument('--have_lr_scheduler', dest='have_lr_scheduler', action='store_false') parser.add_argument('--multi_step_lr', type=str, default='10,20,30') parser.add_argument('--freeze_layers', type=int, default=-1) parser.add_argument('--embedding_opt', type=int, default=0) parser.add_argument('--lr_gamma', type=float, default=0.5) parser.add_argument('--bert_l2norm', type=float, default=0.0) parser.add_argument('--scheduler_type', type=str, default='ms', help='ms/rop/exp') parser.add_argument('--output_dir', default='checkpoint') parser.add_argument('--seed', type=int, default=2018, help='random seed for data shuffling, embedding init, etc.') parser.add_argument('--grad_accumulation_step', type=int, default=1) # fp 16 parser.add_argument('--fp16', action='store_true', help="Whether to use 16-bit (mixed) precision (through NVIDIA apex) instead of 32-bit") parser.add_argument('--fp16_opt_level', type=str, default='O1', help="For fp16: Apex AMP optimization level selected in ['O0', 'O1', 'O2', and 'O3']." "See details at https://nvidia.github.io/apex/amp.html") # save parser.add_argument('--not_save', action='store_true', help="Don't save the model") return parser def dump(path, data): with open(path, 'w') as f: json.dump(data, f) def dump2(path, uids, scores, predictions): with open(path, 'w') as f: for uid, score, pred in zip(uids, scores, predictions): s = json.dumps({'uid': uid, 'score': score, 'prediction': pred}) f.write(s + '\n') def generate_decoder_opt(enable_san, max_opt): opt_v = 0 if enable_san and max_opt < 3: opt_v = max_opt return opt_v def get_args() -> argparse.Namespace: parser = argparse.ArgumentParser() parser = data_config(parser) parser = model_config(parser) parser = train_config(parser) args = parser.parse_args() return args def main(): args = get_args() args.train_datasets = args.train_datasets.split(',') args.test_datasets = args.test_datasets.split(',') args.output_dir = os.path.abspath(args.output_dir) os.makedirs(args.output_dir, exist_ok=True) # log_path = args.log_file logger = create_logger(__name__, to_disk=True, log_file=args.log_file) logger.info('args: %s', json.dumps(vars(args), indent=2)) set_environment(args.seed, args.cuda) task_defs = BlueTaskDefs(args.task_def) encoder_type = task_defs.encoder_type assert encoder_type == EncoderModelType.BERT, '%s: only support BERT' % encoder_type args.encoder_type = encoder_type logger.info('Launching the MT-DNN training') # update data dir train_data_list = [] tasks = {} tasks_class = {} nclass_list = [] decoder_opts = [] task_types = [] dropout_list = [] for dataset in args.train_datasets: task = dataset.split('_')[0] if task in tasks: logger.warning('Skipping: %s in %s', task, tasks) continue assert task in task_defs.n_class_map, \ '%s not in n_class_map: %s' % (task, task_defs.n_class_map) assert task in task_defs.data_format_map, \ '%s not in data_format_map: %s' % (task, task_defs.data_format_map) data_type = task_defs.data_format_map[task] nclass = task_defs.n_class_map[task] task_id = len(tasks) if args.mtl_opt > 0: task_id = tasks_class[nclass] if nclass in tasks_class else len(tasks_class) task_type = task_defs.task_type_map[task] dopt = generate_decoder_opt(task_defs.enable_san_map[task], args.answer_opt) if task_id < len(decoder_opts): decoder_opts[task_id] = min(decoder_opts[task_id], dopt) else: decoder_opts.append(dopt) task_types.append(task_type) if task not in tasks: tasks[task] = len(tasks) if args.mtl_opt < 1: nclass_list.append(nclass) if nclass not in tasks_class: tasks_class[nclass] = len(tasks_class) if args.mtl_opt > 0: nclass_list.append(nclass) dropout_p = task_defs.dropout_p_map.get(task, args.dropout_p) dropout_list.append(dropout_p) # use train and dev train_path = os.path.join(args.data_dir, f'{dataset}_train+dev.json') logger.info('Loading %s as task %s', task, task_id) train_data = BatchGen( BatchGen.load(train_path, True, task_type=task_type, maxlen=args.max_seq_len), batch_size=args.batch_size, dropout_w=args.dropout_w, gpu=args.cuda, task_id=task_id, maxlen=args.max_seq_len, data_type=data_type, task_type=task_type, encoder_type=encoder_type) train_data_list.append(train_data) dev_data_list = [] test_data_list = [] for dataset in args.test_datasets: task = dataset.split('_')[0] task_id = tasks_class[task_defs.n_class_map[task]] if args.mtl_opt > 0 else tasks[task] task_type = task_defs.task_type_map[task] data_type = task_defs.data_format_map[task] dev_path = os.path.join(args.data_dir, f'{dataset}_dev.json') dev_data = BatchGen( BatchGen.load(dev_path, False, task_type=task_type, maxlen=args.max_seq_len), batch_size=args.batch_size_eval, gpu=args.cuda, is_train=False, task_id=task_id, maxlen=args.max_seq_len, data_type=data_type, task_type=task_type, encoder_type=encoder_type) dev_data_list.append(dev_data) test_path = os.path.join(args.data_dir, f'{dataset}_test.json') test_data = BatchGen( BatchGen.load(test_path, False, task_type=task_type, maxlen=args.max_seq_len), batch_size=args.batch_size_eval, gpu=args.cuda, is_train=False, task_id=task_id, maxlen=args.max_seq_len, data_type=data_type, task_type=task_type, encoder_type=encoder_type) test_data_list.append(test_data) opt = copy.deepcopy(vars(args)) opt['answer_opt'] = decoder_opts opt['task_types'] = task_types opt['tasks_dropout_p'] = dropout_list label_size = ','.join([str(l) for l in nclass_list]) opt['label_size'] = label_size logger.info('#' * 20) logger.info('opt: %s', json.dumps(opt, indent=2)) logger.info('#' * 20) bert_model_path = args.init_checkpoint state_dict = None if os.path.exists(bert_model_path): state_dict = torch.load(bert_model_path) config = state_dict['config'] config['attention_probs_dropout_prob'] = args.bert_dropout_p config['hidden_dropout_prob'] = args.bert_dropout_p opt.update(config) else: logger.error('#' * 20) logger.error('Could not find the init model!\n' 'The parameters will be initialized randomly!') logger.error('#' * 20) config = BertConfig(vocab_size_or_config_json_file=30522).to_dict() opt.update(config) all_iters = [iter(item) for item in train_data_list] all_lens = [len(bg) for bg in train_data_list] # div number of grad accumulation. num_all_batches = args.epochs * sum(all_lens) // args.grad_accumulation_step logger.info('############# Gradient Accumulation Info #############') logger.info('number of step: %s', args.epochs * sum(all_lens)) logger.info('number of grad grad_accumulation step: %s', args.grad_accumulation_step) logger.info('adjusted number of step: %s', num_all_batches) logger.info('############# Gradient Accumulation Info #############') if len(train_data_list) > 1 and args.ratio > 0: num_all_batches = int(args.epochs * (len(train_data_list[0]) * (1 + args.ratio))) model = MTDNNModel(opt, state_dict=state_dict, num_train_step=num_all_batches) if args.resume and args.model_ckpt: logger.info('loading model from %s', args.model_ckpt) model.load(args.model_ckpt) # model meta str headline = '############# Model Arch of MT-DNN #############' # print network logger.debug('\n{}\n{}\n'.format(headline, model.network)) # dump config config_file = os.path.join(args.output_dir, 'config.json') with open(config_file, 'a', encoding='utf-8') as writer: writer.write('{}\n'.format(json.dumps(opt))) writer.write('\n{}\n{}\n'.format(headline, model.network)) logger.info("Total number of params: %s", model.total_param) # tensorboard if args.tensorboard: args.tensorboard_logdir = os.path.join(args.output_dir, args.tensorboard_logdir) tensorboard = SummaryWriter(log_dir=args.tensorboard_logdir) for epoch in range(0, args.epochs): logger.warning('At epoch %s', epoch) for train_data in train_data_list: train_data.reset() start = datetime.now() all_indices = [] if len(train_data_list) > 1 and args.ratio > 0: main_indices = [0] * len(train_data_list[0]) extra_indices = [] for i in range(1, len(train_data_list)): extra_indices += [i] * len(train_data_list[i]) random_picks = int(min(len(train_data_list[0]) * args.ratio, len(extra_indices))) extra_indices = np.random.choice(extra_indices, random_picks, replace=False) if args.mix_opt > 0: extra_indices = extra_indices.tolist() random.shuffle(extra_indices) all_indices = extra_indices + main_indices else: all_indices = main_indices + extra_indices.tolist() else: for i in range(1, len(train_data_list)): all_indices += [i] * len(train_data_list[i]) if args.mix_opt > 0: random.shuffle(all_indices) all_indices += [0] * len(train_data_list[0]) if args.mix_opt < 1: random.shuffle(all_indices) for i in range(len(all_indices)): task_id = all_indices[i] batch_meta, batch_data = next(all_iters[task_id]) model.update(batch_meta, batch_data) if model.local_updates % (args.log_per_updates * args.grad_accumulation_step) == 0 \ or model.local_updates == 1: remaining_time = str( (datetime.now() - start) / (i + 1) * (len(all_indices) - i - 1) ).split('.')[0] logger.info('Task [%2d] updates[%6d] train loss[%.5f] remaining[%s]', task_id, model.updates, model.train_loss.avg, remaining_time) if args.tensorboard: tensorboard.add_scalar('train/loss', model.train_loss.avg, global_step=model.updates) if args.save_per_updates_on \ and (model.local_updates % ( args.save_per_updates * args.grad_accumulation_step) == 0): model_file = os.path.join(args.output_dir, f'model_{epoch}_{model.updates}.pt') logger.info('Saving mt-dnn model to %s', model_file) model.save(model_file) for idx, dataset in enumerate(args.test_datasets): task = dataset.split('_')[0] label_mapper = task_defs.label_mapper_map[task] metric_meta = task_defs.metric_meta_map[task] # dev data = dev_data_list[idx] with torch.no_grad(): metrics, predictions, scores, golds, ids = eval_model( model, data, metric_meta, args.cuda, True, label_mapper) for key, val in metrics.items(): if args.tensorboard: tensorboard.add_scalar(f'dev/{dataset}/{key}', val, global_step=epoch) logger.warning('Task %s - epoch %s - Dev %s: %s', dataset, epoch, key, val) path = os.path.join(args.output_dir, f'{dataset}_dev_scores_{epoch}.json') result = {'metrics': metrics, 'predictions': predictions, 'uids': ids, 'scores': scores} dump(path, result) path = os.path.join(args.output_dir, f'{dataset}_dev_scores_{epoch}_2.json') dump2(path, ids, scores, predictions) # test data = test_data_list[idx] with torch.no_grad(): metrics, predictions, scores, golds, ids = eval_model( model, data, metric_meta, args.cuda, True, label_mapper) for key, val in metrics.items(): if args.tensorboard: tensorboard.add_scalar(f'test/{dataset}/{key}', val, global_step=epoch) logger.warning('Task %s - epoch %s - Test %s: %s', dataset, epoch, key, val) path = os.path.join(args.output_dir, f'{dataset}_test_scores_{epoch}.json') result = {'metrics': metrics, 'predictions': predictions, 'uids': ids, 'scores': scores} dump(path, result) path = os.path.join(args.output_dir, f'{dataset}_test_scores_{epoch}_2.json') dump2(path, ids, scores, predictions) logger.info('[new test scores saved.]') if not args.not_save: model_file = os.path.join(args.output_dir, f'model_{epoch}.pt') model.save(model_file) if args.tensorboard: tensorboard.close() if __name__ == '__main__': main()
19,152
42.332579
111
py
bluebert
bluebert-master/mt-bluebert/mt_bluebert/mt_dnn/batcher.py
# coding=utf-8 # Copyright (c) Microsoft. All rights reserved. import sys import json import torch import random from shutil import copyfile from mt_bluebert.data_utils.task_def import TaskType, DataFormat from mt_bluebert.data_utils.task_def import EncoderModelType UNK_ID=100 BOS_ID=101 class BatchGen: def __init__(self, data, batch_size=32, gpu=True, is_train=True, maxlen=128, dropout_w=0.005, do_batch=True, weighted_on=False, task_id=0, task=None, task_type=TaskType.Classification, data_type=DataFormat.PremiseOnly, soft_label=False, encoder_type=EncoderModelType.BERT): self.batch_size = batch_size self.maxlen = maxlen self.is_train = is_train self.gpu = gpu self.weighted_on = weighted_on self.data = data self.task_id = task_id self.pairwise_size = 1 self.data_type = data_type self.task_type=task_type self.encoder_type = encoder_type # soft label used for knowledge distillation self.soft_label_on = soft_label if do_batch: if is_train: indices = list(range(len(self.data))) random.shuffle(indices) data = [self.data[i] for i in indices] self.data = BatchGen.make_baches(data, batch_size) self.offset = 0 self.dropout_w = dropout_w @staticmethod def make_baches(data, batch_size=32): return [data[i:i + batch_size] for i in range(0, len(data), batch_size)] @staticmethod def load(path, is_train=True, maxlen=128, factor=1.0, task_type=None): assert task_type is not None with open(path, 'r', encoding='utf-8') as reader: data = [] cnt = 0 for line in reader: sample = json.loads(line) sample['factor'] = factor cnt += 1 if is_train: if (task_type == TaskType.Ranking) and (len(sample['token_id'][0]) > maxlen or len(sample['token_id'][1]) > maxlen): continue if (task_type != TaskType.Ranking) and (len(sample['token_id']) > maxlen): continue data.append(sample) print('Loaded {} samples out of {}'.format(len(data), cnt)) return data def reset(self): if self.is_train: indices = list(range(len(self.data))) random.shuffle(indices) self.data = [self.data[i] for i in indices] self.offset = 0 def __random_select__(self, arr): if self.dropout_w > 0: return [UNK_ID if random.uniform(0, 1) < self.dropout_w else e for e in arr] else: return arr def __len__(self): return len(self.data) def patch(self, v): v = v.cuda(non_blocking=True) return v @staticmethod def todevice(v, device): v = v.to(device) return v def rebacth(self, batch): newbatch = [] for sample in batch: size = len(sample['token_id']) self.pairwise_size = size assert size == len(sample['type_id']) for idx in range(0, size): token_id = sample['token_id'][idx] type_id = sample['type_id'][idx] uid = sample['ruid'][idx] olab = sample['olabel'][idx] newbatch.append({'uid': uid, 'token_id': token_id, 'type_id': type_id, 'label':sample['label'], 'true_label': olab}) return newbatch def __if_pair__(self, data_type): return data_type in [DataFormat.PremiseAndOneHypothesis, DataFormat.PremiseAndMultiHypothesis] def __iter__(self): while self.offset < len(self): batch = self.data[self.offset] if self.task_type == TaskType.Ranking: batch = self.rebacth(batch) # prepare model input batch_data, batch_info = self._prepare_model_input(batch) batch_info['task_id'] = self.task_id # used for select correct decoding head batch_info['input_len'] = len(batch_data) # used to select model inputs # select different loss function and other difference in training and testing batch_info['task_type'] = self.task_type batch_info['pairwise_size'] = self.pairwise_size # need for ranking task if self.gpu: for i, item in enumerate(batch_data): batch_data[i] = self.patch(item.pin_memory()) # add label labels = [sample['label'] for sample in batch] if self.is_train: # in training model, label is used by Pytorch, so would be tensor if self.task_type == TaskType.Regression: batch_data.append(torch.FloatTensor(labels)) batch_info['label'] = len(batch_data) - 1 elif self.task_type in (TaskType.Classification, TaskType.Ranking): batch_data.append(torch.LongTensor(labels)) batch_info['label'] = len(batch_data) - 1 elif self.task_type == TaskType.Span: start = [sample['token_start'] for sample in batch] end = [sample['token_end'] for sample in batch] batch_data.extend([torch.LongTensor(start), torch.LongTensor(end)]) batch_info['start'] = len(batch_data) - 2 batch_info['end'] = len(batch_data) - 1 elif self.task_type == TaskType.SequenceLabeling: batch_size = self._get_batch_size(batch) tok_len = self._get_max_len(batch, key='token_id') tlab = torch.LongTensor(batch_size, tok_len).fill_(-1) for i, label in enumerate(labels): ll = len(label) tlab[i, : ll] = torch.LongTensor(label) batch_data.append(tlab) batch_info['label'] = len(batch_data) - 1 # soft label generated by ensemble models for knowledge distillation if self.soft_label_on and (batch[0].get('softlabel', None) is not None): assert self.task_type != TaskType.Span # Span task doesn't support soft label yet. sortlabels = [sample['softlabel'] for sample in batch] sortlabels = torch.FloatTensor(sortlabels) batch_info['soft_label'] = self.patch(sortlabels.pin_memory()) if self.gpu else sortlabels else: # in test model, label would be used for evaluation batch_info['label'] = labels if self.task_type == TaskType.Ranking: batch_info['true_label'] = [sample['true_label'] for sample in batch] batch_info['uids'] = [sample['uid'] for sample in batch] # used in scoring self.offset += 1 yield batch_info, batch_data def _get_max_len(self, batch, key='token_id'): tok_len = max(len(x[key]) for x in batch) return tok_len def _get_batch_size(self, batch): return len(batch) def _prepare_model_input(self, batch): batch_size = self._get_batch_size(batch) tok_len = self._get_max_len(batch, key='token_id') #tok_len = max(len(x['token_id']) for x in batch) hypothesis_len = max(len(x['type_id']) - sum(x['type_id']) for x in batch) if self.encoder_type == EncoderModelType.ROBERTA: token_ids = torch.LongTensor(batch_size, tok_len).fill_(1) type_ids = torch.LongTensor(batch_size, tok_len).fill_(0) masks = torch.LongTensor(batch_size, tok_len).fill_(0) else: token_ids = torch.LongTensor(batch_size, tok_len).fill_(0) type_ids = torch.LongTensor(batch_size, tok_len).fill_(0) masks = torch.LongTensor(batch_size, tok_len).fill_(0) if self.__if_pair__(self.data_type): premise_masks = torch.ByteTensor(batch_size, tok_len).fill_(1) hypothesis_masks = torch.ByteTensor(batch_size, hypothesis_len).fill_(1) for i, sample in enumerate(batch): select_len = min(len(sample['token_id']), tok_len) tok = sample['token_id'] if self.is_train: tok = self.__random_select__(tok) token_ids[i, :select_len] = torch.LongTensor(tok[:select_len]) type_ids[i, :select_len] = torch.LongTensor(sample['type_id'][:select_len]) masks[i, :select_len] = torch.LongTensor([1] * select_len) if self.__if_pair__(self.data_type): hlen = len(sample['type_id']) - sum(sample['type_id']) hypothesis_masks[i, :hlen] = torch.LongTensor([0] * hlen) for j in range(hlen, select_len): premise_masks[i, j] = 0 if self.__if_pair__(self.data_type): batch_info = { 'token_id': 0, 'segment_id': 1, 'mask': 2, 'premise_mask': 3, 'hypothesis_mask': 4 } batch_data = [token_ids, type_ids, masks, premise_masks, hypothesis_masks] else: batch_info = { 'token_id': 0, 'segment_id': 1, 'mask': 2 } batch_data = [token_ids, type_ids, masks] return batch_data, batch_info
9,667
42.54955
136
py
bluebert
bluebert-master/mt-bluebert/mt_bluebert/mt_dnn/matcher.py
# coding=utf-8 # Copyright (c) Microsoft. All rights reserved. import torch.nn as nn from pytorch_pretrained_bert.modeling import BertConfig, BertLayerNorm, BertModel from mt_bluebert.module.dropout_wrapper import DropoutWrapper from mt_bluebert.module.san import SANClassifier from mt_bluebert.data_utils.task_def import EncoderModelType, TaskType class LinearPooler(nn.Module): def __init__(self, hidden_size): super(LinearPooler, self).__init__() self.dense = nn.Linear(hidden_size, hidden_size) self.activation = nn.Tanh() def forward(self, hidden_states): first_token_tensor = hidden_states[:, 0] pooled_output = self.dense(first_token_tensor) pooled_output = self.activation(pooled_output) return pooled_output class SANBertNetwork(nn.Module): def __init__(self, opt, bert_config=None): super(SANBertNetwork, self).__init__() self.dropout_list = nn.ModuleList() self.encoder_type = opt['encoder_type'] if opt['encoder_type'] == EncoderModelType.ROBERTA: from fairseq.models.roberta import RobertaModel self.bert = RobertaModel.from_pretrained(opt['init_checkpoint']) hidden_size = self.bert.args.encoder_embed_dim self.pooler = LinearPooler(hidden_size) else: self.bert_config = BertConfig.from_dict(opt) self.bert = BertModel(self.bert_config) hidden_size = self.bert_config.hidden_size if opt.get('dump_feature', False): self.opt = opt return if opt['update_bert_opt'] > 0: for p in self.bert.parameters(): p.requires_grad = False self.decoder_opt = opt['answer_opt'] self.task_types = opt["task_types"] self.scoring_list = nn.ModuleList() labels = [int(ls) for ls in opt['label_size'].split(',')] task_dropout_p = opt['tasks_dropout_p'] for task, lab in enumerate(labels): decoder_opt = self.decoder_opt[task] task_type = self.task_types[task] dropout = DropoutWrapper(task_dropout_p[task], opt['vb_dropout']) self.dropout_list.append(dropout) if task_type == TaskType.Span: assert decoder_opt != 1 out_proj = nn.Linear(hidden_size, 2) elif task_type == TaskType.SequenceLabeling: out_proj = nn.Linear(hidden_size, lab) else: if decoder_opt == 1: out_proj = SANClassifier(hidden_size, hidden_size, lab, opt, prefix='answer', dropout=dropout) else: out_proj = nn.Linear(hidden_size, lab) self.scoring_list.append(out_proj) self.opt = opt self._my_init() def _my_init(self): def init_weights(module): if isinstance(module, (nn.Linear, nn.Embedding)): # Slightly different from the TF version which uses truncated_normal for initialization # cf https://github.com/pytorch/pytorch/pull/5617 module.weight.data.normal_(mean=0.0, std=0.02 * self.opt['init_ratio']) elif isinstance(module, BertLayerNorm): # Slightly different from the BERT pytorch version, which should be a bug. # Note that it only affects on training from scratch. For detailed discussions, please contact xiaodl@. # Layer normalization (https://arxiv.org/abs/1607.06450) # support both old/latest version if 'beta' in dir(module) and 'gamma' in dir(module): module.beta.data.zero_() module.gamma.data.fill_(1.0) else: module.bias.data.zero_() module.weight.data.fill_(1.0) if isinstance(module, nn.Linear): module.bias.data.zero_() self.apply(init_weights) def forward(self, input_ids, token_type_ids, attention_mask, premise_mask=None, hyp_mask=None, task_id=0): if self.encoder_type == EncoderModelType.ROBERTA: sequence_output = self.bert.extract_features(input_ids) pooled_output = self.pooler(sequence_output) else: all_encoder_layers, pooled_output = self.bert(input_ids, token_type_ids, attention_mask) sequence_output = all_encoder_layers[-1] decoder_opt = self.decoder_opt[task_id] task_type = self.task_types[task_id] if task_type == TaskType.Span: assert decoder_opt != 1 sequence_output = self.dropout_list[task_id](sequence_output) logits = self.scoring_list[task_id](sequence_output) start_scores, end_scores = logits.split(1, dim=-1) start_scores = start_scores.squeeze(-1) end_scores = end_scores.squeeze(-1) return start_scores, end_scores elif task_type == TaskType.SequenceLabeling: pooled_output = all_encoder_layers[-1] pooled_output = self.dropout_list[task_id](pooled_output) pooled_output = pooled_output.contiguous().view(-1, pooled_output.size(2)) logits = self.scoring_list[task_id](pooled_output) return logits else: if decoder_opt == 1: max_query = hyp_mask.size(1) assert max_query > 0 assert premise_mask is not None assert hyp_mask is not None hyp_mem = sequence_output[:, :max_query, :] logits = self.scoring_list[task_id](sequence_output, hyp_mem, premise_mask, hyp_mask) else: pooled_output = self.dropout_list[task_id](pooled_output) logits = self.scoring_list[task_id](pooled_output) return logits
5,873
44.890625
119
py
bluebert
bluebert-master/mt-bluebert/mt_bluebert/mt_dnn/model.py
# coding=utf-8 # Copyright (c) Microsoft. All rights reserved. import logging import sys import numpy as np import torch import torch.nn as nn import torch.nn.functional as F import torch.optim as optim from torch.optim.lr_scheduler import * from mt_bluebert.data_utils.utils import AverageMeter from pytorch_pretrained_bert import BertAdam as Adam from mt_bluebert.module.bert_optim import Adamax, RAdam from mt_bluebert.module.my_optim import EMA from .matcher import SANBertNetwork from mt_bluebert.data_utils.task_def import TaskType logger = logging.getLogger(__name__) class MTDNNModel(object): def __init__(self, opt, state_dict=None, num_train_step=-1): self.config = opt self.updates = state_dict['updates'] if state_dict and 'updates' in state_dict else 0 self.local_updates = 0 self.train_loss = AverageMeter() self.network = SANBertNetwork(opt) if state_dict: self.network.load_state_dict(state_dict['state'], strict=False) self.mnetwork = nn.DataParallel(self.network) if opt['multi_gpu_on'] else self.network self.total_param = sum([p.nelement() for p in self.network.parameters() if p.requires_grad]) if opt['cuda']: self.network.cuda() no_decay = ['bias', 'gamma', 'beta', 'LayerNorm.bias', 'LayerNorm.weight'] optimizer_parameters = [ {'params': [p for n, p in self.network.named_parameters() if not any(nd in n for nd in no_decay)], 'weight_decay': 0.01}, {'params': [p for n, p in self.network.named_parameters() if any(nd in n for nd in no_decay)], 'weight_decay': 0.0} ] # note that adamax are modified based on the BERT code if opt['optimizer'] == 'sgd': self.optimizer = optim.SGD(optimizer_parameters, opt['learning_rate'], weight_decay=opt['weight_decay']) elif opt['optimizer'] == 'adamax': self.optimizer = Adamax(optimizer_parameters, opt['learning_rate'], warmup=opt['warmup'], t_total=num_train_step, max_grad_norm=opt['grad_clipping'], schedule=opt['warmup_schedule'], weight_decay=opt['weight_decay']) if opt.get('have_lr_scheduler', False): opt['have_lr_scheduler'] = False elif opt['optimizer'] == 'radam': self.optimizer = RAdam(optimizer_parameters, opt['learning_rate'], warmup=opt['warmup'], t_total=num_train_step, max_grad_norm=opt['grad_clipping'], schedule=opt['warmup_schedule'], eps=opt['adam_eps'], weight_decay=opt['weight_decay']) if opt.get('have_lr_scheduler', False): opt['have_lr_scheduler'] = False # The current radam does not support FP16. opt['fp16'] = False elif opt['optimizer'] == 'adadelta': self.optimizer = optim.Adadelta(optimizer_parameters, opt['learning_rate'], rho=0.95) elif opt['optimizer'] == 'adam': self.optimizer = Adam(optimizer_parameters, lr=opt['learning_rate'], warmup=opt['warmup'], t_total=num_train_step, max_grad_norm=opt['grad_clipping'], schedule=opt['warmup_schedule'], weight_decay=opt['weight_decay']) if opt.get('have_lr_scheduler', False): opt['have_lr_scheduler'] = False else: raise RuntimeError('Unsupported optimizer: %s' % opt['optimizer']) if state_dict and 'optimizer' in state_dict: self.optimizer.load_state_dict(state_dict['optimizer']) if opt['fp16']: try: from apex import amp global amp except ImportError: raise ImportError("Please install apex from https://www.github.com/nvidia/apex to use fp16 training.") model, optimizer = amp.initialize(self.network, self.optimizer, opt_level=opt['fp16_opt_level']) self.network = model self.optimizer = optimizer if opt.get('have_lr_scheduler', False): if opt.get('scheduler_type', 'rop') == 'rop': self.scheduler = ReduceLROnPlateau(self.optimizer, mode='max', factor=opt['lr_gamma'], patience=3) elif opt.get('scheduler_type', 'rop') == 'exp': self.scheduler = ExponentialLR(self.optimizer, gamma=opt.get('lr_gamma', 0.95)) else: milestones = [int(step) for step in opt.get('multi_step_lr', '10,20,30').split(',')] self.scheduler = MultiStepLR(self.optimizer, milestones=milestones, gamma=opt.get('lr_gamma')) else: self.scheduler = None self.ema = None if opt['ema_opt'] > 0: self.ema = EMA(self.config['ema_gamma'], self.network) if opt['cuda']: self.ema.cuda() self.para_swapped = False # zero optimizer grad self.optimizer.zero_grad() def setup_ema(self): if self.config['ema_opt']: self.ema.setup() def update_ema(self): if self.config['ema_opt']: self.ema.update() def eval(self): if self.config['ema_opt']: self.ema.swap_parameters() self.para_swapped = True def train(self): if self.para_swapped: self.ema.swap_parameters() self.para_swapped = False def update(self, batch_meta, batch_data): self.network.train() labels = batch_data[batch_meta['label']] soft_labels = None if self.config.get('mkd_opt', 0) > 0 and ('soft_label' in batch_meta): soft_labels = batch_meta['soft_label'] task_type = batch_meta['task_type'] if task_type == TaskType.Span: start = batch_data[batch_meta['start']] end = batch_data[batch_meta['end']] if self.config["cuda"]: start = start.cuda(non_blocking=True) end = end.cuda(non_blocking=True) start.requires_grad = False end.requires_grad = False else: y = labels if task_type == TaskType.Ranking: y = y.contiguous().view(-1, batch_meta['pairwise_size'])[:, 0] if self.config['cuda']: y = y.cuda(non_blocking=True) y.requires_grad = False task_id = batch_meta['task_id'] inputs = batch_data[:batch_meta['input_len']] if len(inputs) == 3: inputs.append(None) inputs.append(None) inputs.append(task_id) if self.config.get('weighted_on', False): if self.config['cuda']: weight = batch_data[batch_meta['factor']].cuda(non_blocking=True) else: weight = batch_data[batch_meta['factor']] if task_type == TaskType.Span: start_logits, end_logits = self.mnetwork(*inputs) ignored_index = start_logits.size(1) start.clamp_(0, ignored_index) end.clamp_(0, ignored_index) if self.config.get('weighted_on', False): loss = torch.mean(F.cross_entropy(start_logits, start, reduce=False) * weight) + \ torch.mean(F.cross_entropy(end_logits, end, reduce=False) * weight) else: loss = F.cross_entropy(start_logits, start, ignore_index=ignored_index) + \ F.cross_entropy(end_logits, end, ignore_index=ignored_index) loss = loss / 2 elif task_type == TaskType.SequenceLabeling: y = y.view(-1) logits = self.mnetwork(*inputs) loss = F.cross_entropy(logits, y, ignore_index=-1) else: logits = self.mnetwork(*inputs) if task_type == TaskType.Ranking: logits = logits.view(-1, batch_meta['pairwise_size']) if self.config.get('weighted_on', False): if task_type == TaskType.Regression: loss = torch.mean(F.mse_loss(logits.squeeze(), y, reduce=False) * weight) else: loss = torch.mean(F.cross_entropy(logits, y, reduce=False) * weight) if soft_labels is not None: # compute KL label_size = soft_labels.size(1) kd_loss = F.kl_div(F.log_softmax(logits.view(-1, label_size).float(), 1), soft_labels, reduction='batchmean') loss = loss + kd_loss else: if task_type == TaskType.Regression: loss = F.mse_loss(logits.squeeze(), y) else: loss = F.cross_entropy(logits, y) if soft_labels is not None: # compute KL label_size = soft_labels.size(1) kd_loss = F.kl_div(F.log_softmax(logits.view(-1, label_size).float(), 1), soft_labels, reduction='batchmean') loss = loss + kd_loss self.train_loss.update(loss.item(), logits.size(0)) # scale loss loss = loss / self.config.get('grad_accumulation_step', 1) if self.config['fp16']: with amp.scale_loss(loss, self.optimizer) as scaled_loss: scaled_loss.backward() else: loss.backward() self.local_updates += 1 if self.local_updates % self.config.get('grad_accumulation_step', 1) == 0: if self.config['global_grad_clipping'] > 0: if self.config['fp16']: torch.nn.utils.clip_grad_norm_(amp.master_params(self.optimizer), self.config['global_grad_clipping']) else: torch.nn.utils.clip_grad_norm_(self.network.parameters(), self.config['global_grad_clipping']) self.updates += 1 # reset number of the grad accumulation self.optimizer.step() self.optimizer.zero_grad() self.update_ema() def predict(self, batch_meta, batch_data): self.network.eval() task_id = batch_meta['task_id'] task_type = batch_meta['task_type'] inputs = batch_data[:batch_meta['input_len']] if len(inputs) == 3: inputs.append(None) inputs.append(None) inputs.append(task_id) score = self.mnetwork(*inputs) if task_type == TaskType.Ranking: score = score.contiguous().view(-1, batch_meta['pairwise_size']) assert task_type == TaskType.Ranking score = F.softmax(score, dim=1) score = score.data.cpu() score = score.numpy() predict = np.zeros(score.shape, dtype=int) positive = np.argmax(score, axis=1) for idx, pos in enumerate(positive): predict[idx, pos] = 1 predict = predict.reshape(-1).tolist() score = score.reshape(-1).tolist() return score, predict, batch_meta['true_label'] elif task_type == TaskType.SequenceLabeling: mask = batch_data[batch_meta['mask']] score = score.contiguous() score = score.data.cpu() score = score.numpy() predict = np.argmax(score, axis=1).reshape(mask.size()).tolist() valied_lenght = mask.sum(1).tolist() final_predict = [] for idx, p in enumerate(predict): final_predict.append(p[: valied_lenght[idx]]) score = score.reshape(-1).tolist() return score, final_predict, batch_meta['label'] else: if task_type == TaskType.Classification: score = F.softmax(score, dim=1) score = score.data.cpu() score = score.numpy() predict = np.argmax(score, axis=1).tolist() score = score.reshape(-1).tolist() return score, predict, batch_meta['label'] def extract(self, batch_meta, batch_data): self.network.eval() # 'token_id': 0; 'segment_id': 1; 'mask': 2 inputs = batch_data[:3] all_encoder_layers, pooled_output = self.mnetwork.bert(*inputs) return all_encoder_layers, pooled_output def save(self, filename): network_state = dict([(k, v.cpu()) for k, v in self.network.state_dict().items()]) ema_state = dict( [(k, v.cpu()) for k, v in self.ema.model.state_dict().items()]) if self.ema is not None else dict() params = { 'state': network_state, 'optimizer': self.optimizer.state_dict(), 'ema': ema_state, 'config': self.config, } torch.save(params, filename) logger.info('model saved to {}'.format(filename)) def load(self, checkpoint): model_state_dict = torch.load(checkpoint) if model_state_dict['config']['init_checkpoint'].rsplit('/', 1)[1] != \ self.config['init_checkpoint'].rsplit('/', 1)[1]: logger.error( '*** SANBert network is pretrained on a different Bert Model. Please use that to fine-tune for other tasks. ***') sys.exit() self.network.load_state_dict(model_state_dict['state'], strict=False) self.optimizer.load_state_dict(model_state_dict['optimizer']) self.config = model_state_dict['config'] if self.ema: self.ema.model.load_state_dict(model_state_dict['ema']) def cuda(self): self.network.cuda() if self.config['ema_opt']: self.ema.cuda()
14,345
43.006135
133
py
bluebert
bluebert-master/mt-bluebert/mt_bluebert/module/my_optim.py
# Copyright (c) Microsoft. All rights reserved. from copy import deepcopy import torch from torch.nn import Parameter from functools import wraps class EMA: def __init__(self, gamma, model): super(EMA, self).__init__() self.gamma = gamma self.shadow = {} self.model = model self.setup() def setup(self): for name, para in self.model.named_parameters(): if para.requires_grad: self.shadow[name] = para.clone() def cuda(self): for k, v in self.shadow.items(): self.shadow[k] = v.cuda() def update(self): for name,para in self.model.named_parameters(): if para.requires_grad: self.shadow[name] = (1.0 - self.gamma) * para + self.gamma * self.shadow[name] def swap_parameters(self): for name, para in self.model.named_parameters(): if para.requires_grad: temp_data = para.data para.data = self.shadow[name].data self.shadow[name].data = temp_data def state_dict(self): return self.shadow # Adapted from # https://github.com/pytorch/pytorch/blob/master/torch/nn/utils/weight_norm.py # and https://github.com/salesforce/awd-lstm-lm/blob/master/weight_drop.py def _norm(p, dim): """Computes the norm over all dimensions except dim""" if dim is None: return p.norm() elif dim == 0: output_size = (p.size(0),) + (1,) * (p.dim() - 1) return p.contiguous().view(p.size(0), -1).norm(dim=1).view(*output_size) elif dim == p.dim() - 1: output_size = (1,) * (p.dim() - 1) + (p.size(-1),) return p.contiguous().view(-1, p.size(-1)).norm(dim=0).view(*output_size) else: return _norm(p.transpose(0, dim), 0).transpose(0, dim) def _dummy(*args, **kwargs): # We need to replace flatten_parameters with a nothing function return class WeightNorm(torch.nn.Module): def __init__(self, weights, dim): super(WeightNorm, self).__init__() self.weights = weights self.dim = dim def compute_weight(self, module, name): g = getattr(module, name + '_g') v = getattr(module, name + '_v') return v * (g / _norm(v, self.dim)) @staticmethod def apply(module, weights, dim): # Terrible temporary solution to an issue regarding compacting weights # re: CUDNN RNN if issubclass(type(module), torch.nn.RNNBase): module.flatten_parameters = _dummy if weights is None: # do for all weight params weights = [w for w in module._parameters.keys() if 'weight' in w] fn = WeightNorm(weights, dim) for name in weights: if hasattr(module, name): print('Applying weight norm to {} - {}'.format(str(module), name)) weight = getattr(module, name) del module._parameters[name] module.register_parameter( name + '_g', Parameter(_norm(weight, dim).data)) module.register_parameter(name + '_v', Parameter(weight.data)) setattr(module, name, fn.compute_weight(module, name)) module.register_forward_pre_hook(fn) return fn def remove(self, module): for name in self.weights: weight = self.compute_weight(module) delattr(module, name) del module._parameters[name + '_g'] del module._parameters[name + '_v'] module.register_parameter(name, Parameter(weight.data)) def __call__(self, module, inputs): for name in self.weights: setattr(module, name, self.compute_weight(module, name)) def weight_norm(module, weights=None, dim=0): WeightNorm.apply(module, weights, dim) return module
3,836
33.258929
94
py
bluebert
bluebert-master/mt-bluebert/mt_bluebert/module/bert_optim.py
# Copyright (c) Microsoft. All rights reserved. import math import torch from torch.optim import Optimizer from torch.nn.utils import clip_grad_norm_ from pytorch_pretrained_bert.optimization import warmup_constant, warmup_cosine, warmup_linear def warmup_linear_xdl(x, warmup=0.002): if x < warmup: return x/warmup return (1.0 - x)/(1.0 - warmup) def schedule_func(sch): try: f = eval(sch) except: f = warmup_linear return f class Adamax(Optimizer): """Implements BERT version of Adam algorithm with weight decay fix (and no ). Params: lr: learning rate warmup: portion of t_total for the warmup, -1 means no warmup. Default: -1 t_total: total number of training steps for the learning rate schedule, -1 means constant learning rate. Default: -1 schedule: schedule to use for the warmup (see above). Default: 'warmup_linear' b1: Adams b1. Default: 0.9 b2: Adams b2. Default: 0.999 e: Adams epsilon. Default: 1e-6 weight_decay: Weight decay. Default: 0.01 max_grad_norm: Maximum norm for the gradients (-1 means no clipping). Default: 1.0 by xiaodl """ def __init__(self, params, lr, warmup=-1, t_total=-1, schedule='warmup_linear', betas=(0.9, 0.999), eps=1e-6, weight_decay=0.01, max_grad_norm=1.0): if not lr >= 0.0: raise ValueError("Invalid learning rate: {} - should be >= 0.0".format(lr)) if not 0.0 <= warmup < 1.0 and not warmup == -1: raise ValueError("Invalid warmup: {} - should be in [0.0, 1.0[ or -1".format(warmup)) if not 0.0 <= eps: raise ValueError("Invalid epsilon value: {}".format(eps)) if not 0.0 <= betas[0] < 1.0: raise ValueError("Invalid beta parameter at index 0: {}".format(betas[0])) if not 0.0 <= betas[1] < 1.0: raise ValueError("Invalid beta parameter at index 1: {}".format(betas[1])) defaults = dict(lr=lr, schedule=schedule, warmup=warmup, t_total=t_total, betas=betas, eps=eps, weight_decay=weight_decay, max_grad_norm=max_grad_norm) super(Adamax, self).__init__(params, defaults) def get_lr(self): lr = [] for group in self.param_groups: for p in group['params']: state = self.state[p] if len(state) == 0: return [0] if group['t_total'] != -1: schedule_fct = schedule_func(group['schedule']) lr_scheduled = group['lr'] * schedule_fct(state['step']/group['t_total'], group['warmup']) else: lr_scheduled = group['lr'] lr.append(lr_scheduled) return lr def to(self, device): """ Move the optimizer state to a specified device""" for state in self.state.values(): state['exp_avg'].to(device) state['exp_inf'].to(device) def initialize_step(self, initial_step): """Initialize state with a defined step (but we don't have stored averaged). Arguments: initial_step (int): Initial step number. """ for group in self.param_groups: for p in group['params']: state = self.state[p] # State initialization state['step'] = initial_step # Exponential moving average of gradient values state['exp_avg'] = torch.zeros_like(p.data) # Exponential moving average of squared gradient values state['exp_inf'] = torch.zeros_like(p.data) def step(self, closure=None): loss = None if closure is not None: loss = closure() for group in self.param_groups: for p in group['params']: if p.grad is None: continue grad = p.grad.data if grad.is_sparse: raise RuntimeError('Adam does not support sparse gradients, please consider SparseAdam instead') state = self.state[p] # State initialization if len(state) == 0: state['step'] = 0 # Exponential moving average of gradient values state['exp_avg'] = torch.zeros_like(p.data) state['exp_inf'] = torch.zeros_like(p.data) exp_avg, exp_inf = state['exp_avg'], state['exp_inf'] beta1, beta2 = group['betas'] eps = group['eps'] # Add grad clipping if group['max_grad_norm'] > 0: clip_grad_norm_(p, group['max_grad_norm']) # Update biased first moment estimate. exp_avg.mul_(beta1).add_(1 - beta1, grad) # Update the exponentially weighted infinity norm. norm_buf = torch.cat([ exp_inf.mul_(beta2).unsqueeze(0), grad.abs().add_(eps).unsqueeze_(0) ], 0) torch.max(norm_buf, 0, keepdim=False, out=(exp_inf, exp_inf.new().long())) update = exp_avg / (exp_inf + eps) if group['weight_decay'] > 0.0: update += group['weight_decay'] * p.data if group['t_total'] != -1: schedule_fct = schedule_func(group['schedule']) lr_scheduled = group['lr'] * schedule_fct(state['step']/group['t_total'], group['warmup']) else: lr_scheduled = group['lr'] update_with_lr = lr_scheduled * update p.data.add_(-update_with_lr) state['step'] += 1 return loss class RAdam(Optimizer): """Modified from: https://github.com/LiyuanLucasLiu/RAdam/blob/master/radam.py """ def __init__(self, params, lr, warmup=-1, t_total=-1, schedule='warmup_linear', betas=(0.9, 0.999), eps=1e-6, weight_decay=0.001, max_grad_norm=1.0): if not lr >= 0.0: raise ValueError("Invalid learning rate: {} - should be >= 0.0".format(lr)) if not 0.0 <= warmup < 1.0 and not warmup == -1: raise ValueError("Invalid warmup: {} - should be in [0.0, 1.0[ or -1".format(warmup)) if not 0.0 <= eps: raise ValueError("Invalid epsilon value: {}".format(eps)) if not 0.0 <= betas[0] < 1.0: raise ValueError("Invalid beta parameter at index 0: {}".format(betas[0])) if not 0.0 <= betas[1] < 1.0: raise ValueError("Invalid beta parameter at index 1: {}".format(betas[1])) defaults = dict(lr=lr, schedule=schedule, warmup=warmup, t_total=t_total, betas=betas, eps=eps, weight_decay=weight_decay, max_grad_norm=max_grad_norm) self.buffer = [[None, None, None] for ind in range(10)] super(RAdam, self).__init__(params, defaults) def get_lr(self): lr = [] for group in self.param_groups: for p in group['params']: state = self.state[p] if len(state) == 0: return [0] if group['t_total'] != -1: schedule_fct = schedule_func(group['schedule']) lr_scheduled = group['lr'] * schedule_fct(state['step']/group['t_total'], group['warmup']) else: lr_scheduled = group['lr'] lr.append(lr_scheduled) return lr def to(self, device): """ Move the optimizer state to a specified device""" for state in self.state.values(): state['exp_avg'].to(device) state['exp_avg_sq'].to(device) def initialize_step(self, initial_step): """Initialize state with a defined step (but we don't have stored averaged). Arguments: initial_step (int): Initial step number. """ for group in self.param_groups: for p in group['params']: state = self.state[p] # State initialization state['step'] = initial_step # Exponential moving average of gradient values state['exp_avg'] = torch.zeros_like(p.data) # Exponential moving average of squared gradient values state['exp_avg_sq'] = torch.zeros_like(p.data) def step(self, closure=None): loss = None if closure is not None: loss = closure() # set_trace() for group in self.param_groups: for p in group['params']: if p.grad is None: continue grad = p.grad.data.float() if grad.is_sparse: raise RuntimeError('RAdam does not support sparse gradients') p_data_fp32 = p.data.float() state = self.state[p] # State initialization if len(state) == 0: state['step'] = 0 # Exponential moving average of gradient values state['exp_avg'] = torch.zeros_like(p_data_fp32) state['exp_avg_sq'] = torch.zeros_like(p_data_fp32) else: state['exp_avg'] = state['exp_avg'].type_as(p_data_fp32) state['exp_avg_sq'] = state['exp_avg_sq'].type_as(p_data_fp32) exp_avg, exp_avg_sq = state['exp_avg'], state['exp_avg_sq'] beta1, beta2 = group['betas'] eps = group['eps'] # Add grad clipping if group['max_grad_norm'] > 0: clip_grad_norm_(p, group['max_grad_norm']) # Update biased first moment estimate. exp_avg.mul_(beta1).add_(1 - beta1, grad) exp_avg_sq.mul_(beta2).addcmul_(1 - beta2, grad, grad) state['step'] += 1 if group['t_total'] != -1: schedule_fct = schedule_func(group['schedule']) lr_scheduled = group['lr'] * schedule_fct(state['step']/group['t_total'], group['warmup']) else: lr_scheduled = group['lr'] buffered = self.buffer[int(state['step'] % 10)] if state['step'] == buffered[0]: N_sma, step_size = buffered[1], buffered[2] else: buffered[0] = state['step'] beta2_t = beta2 ** state['step'] N_sma_max = 2 / (1 - beta2) - 1 N_sma = N_sma_max - 2 * state['step'] * beta2_t / (1 - beta2_t) buffered[1] = N_sma # more conservative since it's an approximated value if N_sma >= 5: step_size = lr_scheduled * math.sqrt((1 - beta2_t) * (N_sma - 4) / (N_sma_max - 4) * (N_sma - 2) / N_sma * N_sma_max / (N_sma_max - 2)) / (1 - beta1 ** state['step']) else: step_size = lr_scheduled / (1 - beta1 ** state['step']) buffered[2] = step_size if N_sma >= 5: denom = exp_avg_sq.sqrt().add_(group['eps']) p_data_fp32.addcdiv_(-step_size, exp_avg, denom) else: p_data_fp32.add_(-step_size, exp_avg) if group['weight_decay'] != 0: p_data_fp32.add_(-group['weight_decay'] * lr_scheduled, p_data_fp32) p.data.copy_(p_data_fp32) return loss
11,825
41.847826
190
py
bluebert
bluebert-master/mt-bluebert/mt_bluebert/module/similarity.py
# Copyright (c) Microsoft. All rights reserved. import torch import torch.nn as nn import torch.nn.functional as F import numpy from torch.nn.utils import weight_norm from torch.nn.parameter import Parameter from .common import activation, init_wrapper from .dropout_wrapper import DropoutWrapper class DotProduct(nn.Module): def __init__(self, x1_dim, x2_dim, prefix='sim', opt={}, dropout=None): super(DotProduct, self).__init__() assert x1_dim == x2_dim self.opt = opt self.prefix = prefix self.scale_on = opt.get('{}_scale'.format(self.prefix), False) self.scalor = 1.0 / numpy.power(x2_dim, 0.5) def forward(self, x1, x2): assert x1.size(2) == x2.size(2) scores = x1.bmm(x2.transpose(1, 2)) if self.scale_on: scores *= self.scalor return scores class DotProductProject(nn.Module): def __init__(self, x1_dim, x2_dim, prefix='sim', opt={}, dropout=None): super(DotProductProject, self).__init__() self.prefix = prefix self.opt = opt self.hidden_size = opt.get('{}_hidden_size'.format(self.prefix), 64) self.residual_on = opt.get('{}_residual_on'.format(self.prefix), False) self.layer_norm_on = opt.get('{}_norm_on'.format(self.prefix), False) self.share = opt.get('{}_share'.format(self.prefix), False) self.f = activation(opt.get('{}_activation'.format(self.prefix), 'relu')) self.scale_on = opt.get('{}_scale_on'.format(self.prefix), False) self.dropout = dropout x1_in_dim = x1_dim x2_in_dim = x2_dim out_dim = self.hidden_size self.proj_1 = nn.Linear(x1_in_dim, out_dim, bias=False) if self.layer_norm_on: self.proj_1 = weight_norm(self.proj_1) if self.share and x1_in_dim == x2_in_dim: self.proj_2 = self.proj_1 else: self.proj_2 = nn.Linear(x2_in_dim, out_dim) if self.layer_norm_on: self.proj_2 = weight_norm(self.proj_2) if self.scale_on: self.scalar = Parameter(torch.ones(1,1,1) / (self.hidden_size ** 0.5), requires_grad=False) else: self.sclalar = Parameter(torch.ones(1,1, self.hidden_size), requires_grad=True) def forward(self, x1, x2): assert x1.size(2) == x2.size(2) if self.dropout: x1 = self.dropout(x1) x2 = self.dropout(x2) x1_flat = x1.contiguous().view(-1, x1.size(2)) x2_flat = x2.contiguous().view(-1, x2.size(2)) x1_o = self.f(self.proj_1(x1_flat)).view(x1.size(0), x1.size(1), -1) # x2_o = self.f(self.proj_1(x2_flat)).view(x2.size(0), x2.size(1), -1) x2_o = self.f(self.proj_2(x2_flat)).view(x2.size(0), x2.size(1), -1) if self.scale_on: scalar = self.scalar.expand_as(x2_o) x2_o = scalar * x2_o scores = x1_o.bmm(x2_o.transpose(1, 2)) return scores class Bilinear(nn.Module): def __init__(self, x1_dim, x2_dim, prefix='sim', opt={}, dropout=None): super(Bilinear, self).__init__() self.opt = opt self.layer_norm_on = opt.get('{}_norm_on'.format(self.prefix), False) self.transform_on = opt.get('{}_proj_on'.format(self.prefix), False) # self.init = init_wrapper(opt.get('{}_init'.format(self.prefix), '')) self.dropout = dropout if self.transform_on: self.proj = nn.Linear(x1_dim, x2_dim) # self.init(self.proj.weight) if self.layer_norm_on: self.proj = weight_norm(self.proj) def forward(self, x, y): """ x = batch * len * h1 y = batch * h2 x_mask = batch * len """ if self.dropout: x = self.dropout(x) y = self.dropout(y) proj = self.proj(y) if self.transform_on else y if self.dropout: proj = self.dropout(proj) scores = x.bmm(proj.unsqueeze(2)).squeeze(2) return scores class BilinearSum(nn.Module): def __init__(self, x1_dim, x2_dim, prefix='sim', opt={}, dropout=None): super(BilinearSum, self).__init__() self.x_linear = nn.Linear(x1_dim, 1, bias=False) self.y_linear = nn.Linear(x2_dim, 1, bias=False) self.layer_norm_on = opt.get('{}_norm_on'.format(self.prefix), False) self.init = init_wrapper(opt.get('{}_init'.format(self.prefix), False)) if self.layer_norm_on: self.x_linear = weight_norm(self.x_linear) self.y_linear = weight_norm(self.y_linear) self.init(self.x_linear.weight) self.init(self.y_linear.weight) self.dropout = dropout def forward(self, x1, x2): """ x1: batch * len1 * input_size x2: batch * len2 * input_size score: batch * len1 * len2 """ if self.dropout: x1 = self.dropout(x1) x2 = self.dropout(x2) x1_logits = self.x_linear(x1.contiguous().view(-1, x1.size(-1))).view(x1.size(0), -1, 1) x2_logits = self.y_linear(x2.contiguous().view(-1, x2.size(-1))).view(x2.size(0), 1, -1) shape = (x1.size(0), x1.size(1), x2.size()) scores = x1_logits.expand_as(shape) + x2_logits.expand_as(shape) return scores class Trilinear(nn.Module): """Function used in BiDAF""" def __init__(self, x1_dim, x2_dim, prefix='sim', opt={}, dropout=None): super(Trilinear, self).__init__() self.prefix = prefix self.x_linear = nn.Linear(x1_dim, 1, bias=False) self.x_dot_linear = nn.Linear(x1_dim, 1, bias=False) self.y_linear = nn.Linear(x2_dim, 1, bias=False) self.layer_norm_on = opt.get('{}_norm_on'.format(self.prefix), False) self.init = init_wrapper(opt.get('{}_init'.format(self.prefix), 'xavier_uniform')) if self.layer_norm_on: self.x_linear = weight_norm(self.x_linear) self.x_dot_linear = weight_norm(self.x_dot_linear) self.y_linear = weight_norm(self.y_linear) self.init(self.x_linear.weight) self.init(self.x_dot_linear.weight) self.init(self.y_linear.weight) self.dropout = dropout def forward(self, x1, x2): """ x1: batch * len1 * input_size x2: batch * len2 * input_size score: batch * len1 * len2 """ if self.dropout: x1 = self.dropout(x1) x2 = self.dropout(x2) x1_logits = self.x_linear(x1.contiguous().view(-1, x1.size(-1))).view(x1.size(0), -1, 1) x2_logits = self.y_linear(x2.contiguous().view(-1, x2.size(-1))).view(x2.size(0), 1, -1) x1_dot = self.x_dot_linear(x1.contiguous().view(-1, x1.size(-1))).view(x1.size(0), -1, 1).expand_as(x1) x1_dot = x1 * x1_dot scores = x1_dot.bmm(x2.transpose(1, 2)) scores += x1_logits.expand_as(scores) + x2_logits.expand_as(scores) return scores class SimilarityWrapper(nn.Module): def __init__(self, x1_dim, x2_dim, prefix='attention', opt={}, dropout=None): super(SimilarityWrapper, self).__init__() self.score_func_str = opt.get('{}_sim_func'.format(prefix), 'dotproductproject').lower() self.score_func = None if self.score_func_str == 'dotproduct': self.score_func = DotProduct(x1_dim, x2_dim, prefix=prefix, opt=opt, dropout=dropout) elif self.score_func_str == 'dotproductproject': self.score_func = DotProductProject(x1_dim, x2_dim, prefix=prefix, opt=opt, dropout=dropout) elif self.score_func_str == 'bilinear': self.score_func = Bilinear(x1_dim, x2_dim, prefix=prefix, opt=opt, dropout=dropout) elif self.score_func_str == 'bilinearsum': self.score_func = BilinearSum(x1_dim, x2_dim, prefix=prefix, opt=opt, dropout=dropout) elif self.score_func_str == 'trilinear': self.score_func = Trilinear(x1_dim, x2_dim, prefix=prefix, opt=opt, dropout=dropout) else: raise NotImplementedError def forward(self, x1, x2): scores = self.score_func(x1, x2) return scores class AttentionWrapper(nn.Module): def __init__(self, x1_dim, x2_dim, x3_dim=None, prefix='attention', opt={}, dropout=None): super(AttentionWrapper, self).__init__() self.prefix = prefix self.att_dropout = opt.get('{}_att_dropout'.format(self.prefix), 0) self.score_func = SimilarityWrapper(x1_dim, x2_dim, prefix=prefix, opt=opt, dropout=dropout) self.drop_diagonal = opt.get('{}_drop_diagonal'.format(self.prefix), False) self.output_size = x2_dim if x3_dim is None else x3_dim def forward(self, query, key, value, key_padding_mask=None, return_scores=False): logits = self.score_func(query, key) key_mask = key_padding_mask.unsqueeze(1).expand_as(logits) logits.data.masked_fill_(key_mask.data, -float('inf')) if self.drop_diagonal: assert logits.size(1) == logits.size(2) diag_mask = torch.diag(logits.data.new(logits.size(1)).zero_() + 1).byte().unsqueeze(0).expand_as(logits) logits.data.masked_fill_(diag_mask, -float('inf')) prob = F.softmax(logits.view(-1, key.size(1)), 1) prob = prob.view(-1, query.size(1), key.size(1)) if self.att_dropout > 0: prob = self.dropout(prob) if value is None: value = key attn = prob.bmm(value) if return_scores: return attn, prob, logits else: return attn class LinearSelfAttn(nn.Module): """Self attention over a sequence: * o_i = softmax(Wx_i) for x_i in X. """ def __init__(self, input_size, dropout=None): super(LinearSelfAttn, self).__init__() self.linear = nn.Linear(input_size, 1) self.dropout = dropout def forward(self, x, x_mask): x = self.dropout(x) x_flat = x.contiguous().view(-1, x.size(-1)) scores = self.linear(x_flat).view(x.size(0), x.size(1)) scores.data.masked_fill_(x_mask.data, -float('inf')) alpha = F.softmax(scores, 1) return alpha.unsqueeze(1).bmm(x).squeeze(1) class MLPSelfAttn(nn.Module): def __init__(self, input_size, opt={}, prefix='attn_sum', dropout=None): super(MLPSelfAttn, self).__init__() self.prefix = prefix self.FC = nn.Linear(input_size, input_size) self.linear = nn.Linear(input_size, 1) self.layer_norm_on = opt.get('{}_norm_on'.format(self.prefix), False) self.f = activation(opt.get('{}_activation'.format(self.prefix), 'relu')) if dropout is None: self.dropout = DropoutWrapper(opt.get('{}_dropout_p'.format(self.prefix), 0)) else: self.dropout = dropout if self.layer_norm_on: self.FC = weight_norm(self.FC) def forward(self, x, x_mask): x = self.dropout(x) x_flat = x.contiguous().view(-1, x.size(-1)) scores = self.linear(self.f(self.FC(x_flat))).view(x.size(0), x.size(1)) scores.data.masked_fill_(x_mask.data, -float('inf')) alpha = F.softmax(scores) return alpha.unsqueeze(1).bmm(x).squeeze(1) class SelfAttnWrapper(nn.Module): def __init__(self, input_size, prefix='attn_sum', opt={}, dropout=None): super(SelfAttnWrapper, self).__init__() """ Self att wrapper, support linear and MLP """ attn_type = opt.get('{}_type'.format(prefix), 'linear') if attn_type == 'mlp': self.att = MLPSelfAttn(input_size, prefix, opt, dropout) else: self.att = LinearSelfAttn(input_size, dropout) def forward(self, x, x_mask): return self.att(x, x_mask) class DeepAttentionWrapper(nn.Module): def __init__(self, x1_dim, x2_dim, x3_dims, att_cnt, prefix='deep_att', opt=None, dropout=None): super(DeepAttentionWrapper, self).__init__() self.opt = {} if opt is None else opt self.prefix = prefix self.x1_dim = x1_dim self.x2_dim = x2_dim self.x3_dims = x3_dims if dropout is None: self.dropout = DropoutWrapper(opt.get('{}_dropout_p'.format(self.prefix), 0)) else: self.dropout = dropout self.attn_list = nn.ModuleList() for i in range(0, att_cnt): if opt['multihead_on']: attention = MultiheadAttentionWrapper(self.x1_dim, self.x2_dim, self.x3_dims[i], prefix, opt, dropout=dropout) else: attention = AttentionWrapper(self.x1_dim, self.x2_dim, self.x3_dims[i], prefix, opt, self.dropout) self.attn_list.append(attention) def forward(self, x1, x2, x3, x2_mask): rvl = [] for i in range(0, len(x3)): hiddens = self.attn_list[i](x1, x2, x3[i], x2_mask) rvl.append(hiddens) return torch.cat(rvl, 2) class BilinearFlatSim(nn.Module): """A bilinear attention layer over a sequence X w.r.t y: * o_i = x_i'Wy for x_i in X. """ def __init__(self, x_size, y_size, opt={}, prefix='seqatt', dropout=None): super(BilinearFlatSim, self).__init__() self.opt = opt self.weight_norm_on = opt.get('{}_weight_norm_on'.format(prefix), False) self.linear = nn.Linear(y_size, x_size) if self.weight_norm_on: self.linear = weight_norm(self.linear) if dropout is None: self.dropout = DropoutWrapper(opt.get('{}_dropout_p'.format(self.prefix), 0)) else: self.dropout = dropout def forward(self, x, y, x_mask): """ x = batch * len * h1 y = batch * h2 x_mask = batch * len """ x = self.dropout(x) y = self.dropout(y) Wy = self.linear(y) xWy = x.bmm(Wy.unsqueeze(2)).squeeze(2) xWy.data.masked_fill_(x_mask.data, -float('inf')) return xWy class SimpleFlatSim(nn.Module): def __init__(self, x_size, y_size, opt={}, prefix='seqatt', dropout=None): super(SimpleFlatSim, self).__init__() self.opt = opt self.weight_norm_on = opt.get('{}_norm_on'.format(prefix), False) self.linear = nn.Linear(y_size + x_size, 1) if self.weight_norm_on: self.linear = weight_norm(self.linear) if dropout is None: self.dropout = DropoutWrapper(opt.get('{}_dropout_p'.format(self.prefix), 0)) else: self.dropout = dropout def forward(self, x, y, x_mask): """ x = batch * len * h1 y = batch * h2 x_mask = batch * len """ x = self.dropout(x) y = self.dropout(y) y = y.unsqueeze(1).expand_as(x) flat_x = torch.cat([x, y], 2).contiguous().view(x.size(0) * x.size(1), -1) flat_scores = self.linear(flat_x) scores = flat_scores.contiguous().view(x.size(0), -1) scores.data.masked_fill_(x_mask.data, -float('inf')) return scores class FlatSim(nn.Module): def __init__(self, x_size, y_size, opt={}, prefix='seqatt', dropout=None): super(FlatSim, self).__init__() assert x_size == y_size self.opt = opt self.weight_norm_on = opt.get('{}_weight_norm_on'.format(prefix), False) self.linear = nn.Linear(x_size * 3, 1) if self.weight_norm_on: self.linear = weight_norm(self.linear) if dropout is None: self.dropout = DropoutWrapper(opt.get('{}_dropout_p'.format(self.prefix), 0)) else: self.dropout = dropout def forward(self, x, y, x_mask): """ x = batch * len * h1 y = batch * h2 x_mask = batch * len """ x = self.dropout(x) y = self.dropout(y) y = y.unsqueeze(1).expand_as(x) flat_x = torch.cat([x, y, x * y], 2).contiguous().view(x.size(0) * x.size(1), -1) flat_scores = self.linear(flat_x) scores = flat_scores.contiguous().view(x.size(0), -1) scores.data.masked_fill_(x_mask.data, -float('inf')) return scores class FlatSimV2(nn.Module): def __init__(self, x_size, y_size, opt={}, prefix='seqatt', dropout=None): super(FlatSimV2, self).__init__() assert x_size == y_size self.opt = opt self.weight_norm_on = opt.get('{}_weight_norm_on'.format(prefix), False) self.linear = nn.Linear(x_size * 4, 1) if self.weight_norm_on: self.linear = weight_norm(self.linear) if dropout is None: self.dropout = DropoutWrapper(opt.get('{}_dropout_p'.format(self.prefix), 0)) else: self.dropout = dropout def forward(self, x, y, x_mask): """ x = batch * len * h1 y = batch * h2 x_mask = batch * len """ x = self.dropout(x) y = self.dropout(y) y = y.unsqueeze(1).expand_as(x) flat_x = torch.cat([x, y, x * y, torch.abs(x - y)], 2).contiguous().view(x.size(0) * x.size(1), -1) flat_scores = self.linear(flat_x) scores = flat_scores.contiguous().view(x.size(0), -1) scores.data.masked_fill_(x_mask.data, -float('inf')) return scores class FlatSimilarityWrapper(nn.Module): def __init__(self, x1_dim, x2_dim, prefix='attention', opt={}, dropout=None): super(FlatSimilarityWrapper, self).__init__() self.score_func_str = opt.get('{}_att_type'.format(prefix), 'none').lower() self.att_dropout = DropoutWrapper(opt.get('{}_att_dropout'.format(prefix), 0)) self.score_func = None if self.score_func_str == 'bilinear': self.score_func = BilinearFlatSim(x1_dim, x2_dim, prefix=prefix, opt=opt, dropout=dropout) elif self.score_func_str == 'simple': self.score_func = SimpleFlatSim(x1_dim, x2_dim, prefix=prefix, opt=opt, dropout=dropout) elif self.score_func_str == 'flatsim': self.score_func = FlatSim(x1_dim, x2_dim, prefix=prefix, opt=opt, dropout=dropout) else: self.score_func = FlatSimV2(x1_dim, x2_dim, prefix=prefix, opt=opt, dropout=dropout) def forward(self, x1, x2, mask): scores = self.score_func(x1, x2, mask) return scores class MultiheadAttentionWrapper(nn.Module): """Multi-headed attention. See "Attention Is All You Need" for more details. """ def __init__(self, query_dim, key_dim, value_dim, prefix='attention', opt={}, dropout=None): super().__init__() self.prefix = prefix self.num_heads = opt.get('{}_head'.format(self.prefix), 1) self.dropout = DropoutWrapper(opt.get('{}_dropout'.format(self.prefix), 0)) if dropout is None else dropout self.qkv_dim = [query_dim, key_dim, value_dim] assert query_dim == key_dim, "query dim must equal with key dim" self.hidden_size = opt.get('{}_hidden_size'.format(self.prefix), 64) self.proj_on = opt.get('{}_proj_on'.format(prefix), False) self.share = opt.get('{}_share'.format(self.prefix), False) self.layer_norm_on = opt.get('{}_norm_on'.format(self.prefix), False) self.scale_on = opt.get('{}_scale_on'.format(self.prefix), False) if self.proj_on: self.proj_modules = nn.ModuleList([nn.Linear(dim, self.hidden_size) for dim in self.qkv_dim[0:2]]) if self.layer_norm_on: for proj in self.proj_modules: proj = weight_norm(proj) if self.share and self.qkv_dim[0] == self.qkv_dim[1]: self.proj_modules[1] = self.proj_modules[0] self.f = activation(opt.get('{}_activation'.format(self.prefix), 'relu')) self.qkv_head_dim = [self.hidden_size // self.num_heads] * 3 self.qkv_head_dim[2] = value_dim // self.num_heads assert self.qkv_head_dim[0] * self.num_heads == self.hidden_size, "hidden size must be divisible by num_heads" assert self.qkv_head_dim[2] * self.num_heads == value_dim, "value size must be divisible by num_heads" else: self.qkv_head_dim = [emb // self.num_heads for emb in self.qkv_dim] #import pdb; pdb.set_trace() assert self.qkv_head_dim[0] * self.num_heads == self.qkv_dim[0], "query size must be divisible by num_heads" assert self.qkv_head_dim[1] * self.num_heads == self.qkv_dim[1], "key size must be divisible by num_heads" assert self.qkv_head_dim[2] * self.num_heads == self.qkv_dim[2], "value size must be divisible by num_heads" if self.scale_on: self.scaling = self.qkv_head_dim[0]**-0.5 self.drop_diagonal = opt.get('{}_drop_diagonal'.format(self.prefix), False) self.output_size = self.qkv_dim[2] def forward(self, query, key, value, key_padding_mask=None): query = query.transpose(0, 1) key = key.transpose(0, 1) value = value.transpose(0, 1) tgt_len, bsz, embed_dim = query.size() assert embed_dim == self.qkv_dim[0] q, k, v = query, key, value if self.proj_on: if self.dropout: q, k = self.dropout(q), self.dropout(k) q, k = [self.f(proj(input)) for input, proj in zip([query, key], self.proj_modules)] src_len = k.size(0) if key_padding_mask is not None: assert key_padding_mask.size(0) == bsz assert key_padding_mask.size(1) == src_len if self.scale_on: q *= self.scaling q = q.contiguous().view(tgt_len, bsz*self.num_heads, self.qkv_head_dim[0]).transpose(0, 1) k = k.contiguous().view(src_len, bsz*self.num_heads, self.qkv_head_dim[1]).transpose(0, 1) v = v.contiguous().view(src_len, bsz*self.num_heads, self.qkv_head_dim[2]).transpose(0, 1) attn_weights = torch.bmm(q, k.transpose(1, 2)) assert list(attn_weights.size()) == [bsz * self.num_heads, tgt_len, src_len] if key_padding_mask is not None: # don't attend to padding symbols attn_weights = attn_weights.view(bsz, self.num_heads, tgt_len, src_len) attn_weights = attn_weights.float().masked_fill( key_padding_mask.unsqueeze(1).unsqueeze(2), float('-inf'), ).type_as(attn_weights) # FP16 support: cast to float and back attn_weights = attn_weights.view(bsz * self.num_heads, tgt_len, src_len) if self.drop_diagonal: assert attn_weights.size(1) == attn_weights.size(2) diag_mask = torch.diag(attn_weights.data.new(attn_weights.size(1)).zero_() + 1).byte().unsqueeze(0).expand_as(attn_weights) attn_weights.data.masked_fill_(diag_mask, -float('inf')) attn_weights = F.softmax(attn_weights.float(), dim=-1).type_as(attn_weights) attn_weights = self.dropout(attn_weights) attn = torch.bmm(attn_weights, v) assert list(attn.size()) == [bsz * self.num_heads, tgt_len, self.qkv_head_dim[2]] attn = attn.transpose(0, 1).contiguous().view(tgt_len, bsz, -1) # output_shape: Batch * Time * Channel attn = attn.transpose(0, 1) return attn
23,222
40.030035
135
py
bluebert
bluebert-master/mt-bluebert/mt_bluebert/module/dropout_wrapper.py
# Copyright (c) Microsoft. All rights reserved. import torch import torch.nn as nn import torch.nn.functional as F class DropoutWrapper(nn.Module): """ This is a dropout wrapper which supports the fix mask dropout """ def __init__(self, dropout_p=0, enable_vbp=True): super(DropoutWrapper, self).__init__() """variational dropout means fix dropout mask ref: https://discuss.pytorch.org/t/dropout-for-rnns/633/11 """ self.enable_variational_dropout = enable_vbp self.dropout_p = dropout_p def forward(self, x): """ :param x: batch * len * input_size """ if self.training == False or self.dropout_p == 0: return x if len(x.size()) == 3: mask = 1.0 / (1-self.dropout_p) * torch.bernoulli((1-self.dropout_p) * (x.data.new(x.size(0), x.size(2)).zero_() + 1)) mask.requires_grad = False return mask.unsqueeze(1).expand_as(x) * x else: return F.dropout(x, p=self.dropout_p, training=self.training)
1,074
33.677419
130
py
bluebert
bluebert-master/mt-bluebert/mt_bluebert/module/common.py
# Copyright (c) Microsoft. All rights reserved. import torch import math from torch.nn.functional import tanh, relu, prelu, leaky_relu, sigmoid, elu, selu from torch.nn.init import uniform, normal, eye, xavier_uniform, xavier_normal, kaiming_uniform, kaiming_normal, orthogonal def linear(x): return x def swish(x): return x * sigmoid(x) def bertgelu(x): return x * 0.5 * (1.0 + torch.erf(x / math.sqrt(2.0))) def gptgelu(x): return 0.5 * x * (1 + torch.tanh(math.sqrt(2 / math.pi) * (x + 0.044715 * torch.pow(x, 3)))) # default gelue gelu = bertgelu def activation(func_a): """Activation function wrapper """ try: f = eval(func_a) except: f = linear return f def init_wrapper(init='xavier_uniform'): return eval(init)
782
22.727273
122
py
bluebert
bluebert-master/mt-bluebert/mt_bluebert/module/sub_layers.py
# Copyright (c) Microsoft. All rights reserved. import torch import torch.nn as nn import torch.nn.functional as F from torch.nn.parameter import Parameter class LayerNorm(nn.Module): #ref: https://github.com/pytorch/pytorch/issues/1959 # :https://arxiv.org/pdf/1607.06450.pdf def __init__(self, hidden_size, eps=1e-4): super(LayerNorm, self).__init__() self.alpha = Parameter(torch.ones(1,1,hidden_size)) # gain g self.beta = Parameter(torch.zeros(1,1,hidden_size)) # bias b self.eps = eps def forward(self, x): """ Args: :param x: batch * len * input_size Returns: normalized x """ mu = torch.mean(x, 2, keepdim=True).expand_as(x) sigma = torch.std(x, 2, keepdim=True).expand_as(x) return (x - mu) / (sigma + self.eps) * self.alpha.expand_as(x) + self.beta.expand_as(x)
905
32.555556
95
py
bluebert
bluebert-master/mt-bluebert/mt_bluebert/module/san.py
# Copyright (c) Microsoft. All rights reserved. import torch import random import torch.nn as nn from torch.nn.utils import weight_norm from torch.nn.parameter import Parameter import torch.nn.functional as F from mt_bluebert.module.dropout_wrapper import DropoutWrapper from mt_bluebert.module.similarity import FlatSimilarityWrapper, SelfAttnWrapper from mt_bluebert.module.my_optim import weight_norm as WN SMALL_POS_NUM=1.0e-30 def generate_mask(new_data, dropout_p=0.0, is_training=False): if not is_training: dropout_p = 0.0 new_data = (1-dropout_p) * (new_data.zero_() + 1) for i in range(new_data.size(0)): one = random.randint(0, new_data.size(1)-1) new_data[i][one] = 1 mask = 1.0/(1 - dropout_p) * torch.bernoulli(new_data) mask.requires_grad = False return mask class Classifier(nn.Module): def __init__(self, x_size, y_size, opt, prefix='decoder', dropout=None): super(Classifier, self).__init__() self.opt = opt if dropout is None: self.dropout = DropoutWrapper(opt.get('{}_dropout_p'.format(prefix), 0)) else: self.dropout = dropout self.merge_opt = opt.get('{}_merge_opt'.format(prefix), 0) self.weight_norm_on = opt.get('{}_weight_norm_on'.format(prefix), False) if self.merge_opt == 1: self.proj = nn.Linear(x_size * 4, y_size) else: self.proj = nn.Linear(x_size * 2, y_size) if self.weight_norm_on: self.proj = weight_norm(self.proj) def forward(self, x1, x2, mask=None): if self.merge_opt == 1: x = torch.cat([x1, x2, (x1 - x2).abs(), x1 * x2], 1) else: x = torch.cat([x1, x2], 1) x = self.dropout(x) scores = self.proj(x) return scores class SANClassifier(nn.Module): """Implementation of Stochastic Answer Networks for Natural Language Inference, Xiaodong Liu, Kevin Duh and Jianfeng Gao https://arxiv.org/abs/1804.07888 """ def __init__(self, x_size, h_size, label_size, opt={}, prefix='decoder', dropout=None): super(SANClassifier, self).__init__() if dropout is None: self.dropout = DropoutWrapper(opt.get('{}_dropout_p'.format(self.prefix), 0)) else: self.dropout = dropout self.prefix = prefix self.query_wsum = SelfAttnWrapper(x_size, prefix='mem_cum', opt=opt, dropout=self.dropout) self.attn = FlatSimilarityWrapper(x_size, h_size, prefix, opt, self.dropout) self.rnn_type = '{}{}'.format(opt.get('{}_rnn_type'.format(prefix), 'gru').upper(), 'Cell') self.rnn =getattr(nn, self.rnn_type)(x_size, h_size) self.num_turn = opt.get('{}_num_turn'.format(prefix), 5) self.opt = opt self.mem_random_drop = opt.get('{}_mem_drop_p'.format(prefix), 0) self.mem_type = opt.get('{}_mem_type'.format(prefix), 0) self.weight_norm_on = opt.get('{}_weight_norm_on'.format(prefix), False) self.label_size = label_size self.dump_state = opt.get('dump_state_on', False) self.alpha = Parameter(torch.zeros(1, 1), requires_grad=False) if self.weight_norm_on: self.rnn = WN(self.rnn) self.classifier = Classifier(x_size, self.label_size, opt, prefix=prefix, dropout=self.dropout) def forward(self, x, h0, x_mask=None, h_mask=None): h0 = self.query_wsum(h0, h_mask) if type(self.rnn) is nn.LSTMCell: c0 = h0.new(h0.size()).zero_() scores_list = [] for turn in range(self.num_turn): att_scores = self.attn(x, h0, x_mask) x_sum = torch.bmm(F.softmax(att_scores, 1).unsqueeze(1), x).squeeze(1) scores = self.classifier(x_sum, h0) scores_list.append(scores) # next turn if self.rnn is not None: h0 = self.dropout(h0) if type(self.rnn) is nn.LSTMCell: h0, c0 = self.rnn(x_sum, (h0, c0)) else: h0 = self.rnn(x_sum, h0) if self.mem_type == 1: mask = generate_mask(self.alpha.data.new(x.size(0), self.num_turn), self.mem_random_drop, self.training) mask = [m.contiguous() for m in torch.unbind(mask, 1)] tmp_scores_list = [mask[idx].view(x.size(0), 1).expand_as(inp) * F.softmax(inp, 1) for idx, inp in enumerate(scores_list)] scores = torch.stack(tmp_scores_list, 2) scores = torch.mean(scores, 2) scores = torch.log(scores) else: scores = scores_list[-1] if self.dump_state: return scores, scores_list else: return scores
4,721
41.540541
134
py
bluebert
bluebert-master/mt-bluebert/mt_bluebert/experiments/squad/verify_calc_span.py
from pytorch_pretrained_bert import BertTokenizer from data_utils.task_def import EncoderModelType from experiments.squad.squad_utils import calc_tokenized_span_range, parse_squad_label model = "bert-base-uncased" do_lower_case = True tokenizer = BertTokenizer.from_pretrained(model, do_lower_case=do_lower_case) for no, line in enumerate(open(r"data\canonical_data\squad_v2_train.tsv", encoding="utf-8")): if no % 1000 == 0: print(no) uid, label, context, question = line.strip().split("\t") answer_start, answer_end, answer, is_impossible = parse_squad_label(label) calc_tokenized_span_range(context, question, answer, answer_start, answer_end, tokenizer, EncoderModelType.BERT, verbose=True)
751
46
116
py