Coverage for encodermap/encodermap_tf1/parameters.py: 19%
76 statements
« prev ^ index » next coverage.py v7.1.0, created at 2023-02-07 11:05 +0000
1import datetime
2import json
3import os
4from math import pi
6from .misc import run_path, search_and_replace
9class ParametersFramework:
10 def __init__(self):
11 self.main_path = os.getcwd()
13 def save(self, path=None):
14 """
15 Save parameters in json format
17 :param path: Path where parameters should be saved. If no path is given main_path/parameters.json is used.
18 :return: The path where the parameters were saved.
19 """
20 if not path:
21 path = os.path.join(self.main_path, "parameters.json")
22 if os.path.isfile(path):
23 filename, extension = os.path.splitext(path)
24 time = datetime.datetime.now().strftime("%Y-%m-%d_%H:%M:%S")
25 os.rename(path, filename + "_back_" + time + extension)
26 with open(path, "w") as file:
27 json.dump(self.__dict__, file, indent=4, sort_keys=True)
28 return path
30 @classmethod
31 def load(cls, path):
32 """
33 Loads the parameters saved in a json file into a new Parameter object.
35 :param path: path of the json parameter file
36 :return: a Parameter object
37 """
38 with open(path, "r") as file:
39 params = json.load(file)
40 if params["main_path"] != os.path.dirname(path):
41 print("seams like the directory was moved. Parameter file is updated ...")
42 search_and_replace(path, params["main_path"], os.path.dirname(path))
43 with open(path, "r") as file:
44 params = json.load(file)
46 param = cls()
47 param._setattrs(params)
48 return param
50 def _setattrs(self, dictionary):
51 for key, value in dictionary.items():
52 setattr(self, key, value)
54 def __getitem__(self, item):
55 return getattr(self, item)
58class Parameters(ParametersFramework):
59 """
60 :ivar main_path: Defines a main path where the parameters and other things might be stored.
61 :ivar n_neurons: List containing number of neurons for each layer up to the bottleneck layer.
62 For example [128, 128, 2] stands for an autoencoder with the following architecture
63 {i, 128, 128, 2, 128, 128, i} where i is the number of dimensions of the input data.
64 :ivar activation_functions: List of activation function names as implemented in TensorFlow.
65 For example: "relu", "tanh", "sigmoid" or "" to use no activation function.
66 The encoder part of the network takes the activation functions from the list starting with the second element.
67 The decoder part of the network takes the activation functions in reversed order starting with
68 the second element form the back. For example ["", "relu", "tanh", ""] would result in a autoencoder with
69 {"relu", "tanh", "", "tanh", "relu", ""} as sequence of activation functions.
70 :ivar periodicity: Defines the distance between periodic walls for the inputs.
71 For example 2pi for angular values in radians.
72 All periodic data processed by EncoderMap must be wrapped to one periodic window.
73 E.g. data with 2pi periodicity may contain values from -pi to pi or from 0 to 2pi.
74 Set the periodicity to float("inf") for non-periodic inputs.
76 :ivar learning_rate: Learning rate used by the optimizer.
77 :ivar n_steps: Number of training steps.
78 :ivar batch_size: Number of training points used in each training step
79 :ivar summary_step: A summary for TensorBoard is writen every summary_step steps.
80 :ivar checkpoint_step: A checkpoint is writen every checkpoint_step steps.
82 :ivar dist_sig_parameters: Parameters for the sigmoid functions applied to the high- and low-dimensional distances
83 in the following order (sig_h, a_h, b_h, sig_l, a_l, b_l)
84 :ivar distance_cost_scale: Adjusts how much the distance based metric is weighted in the cost function.
85 :ivar auto_cost_scale: Adjusts how much the autoencoding cost is weighted in the cost function.
86 :ivar auto_cost_variant: defines how the auto cost is calculated. Must be one of: "mean_square", "mean_abs",
87 "mean_norm"
88 :ivar center_cost_scale: Adjusts how much the centering cost is weighted in the cost function.
89 :ivar l2_reg_constant: Adjusts how much the l2 regularisation is weighted in the cost function.
91 :ivar gpu_memory_fraction: Specifies the fraction of gpu memory blocked.
92 If it is 0 memory is allocated as needed.
93 :ivar analysis_path: A path that can be used to store analysis
94 :ivar id: Can be any name for the run. Might be useful for example for specific analysis for different data sets.
97 """
99 def __init__(self):
100 super(Parameters, self).__init__()
101 self.n_neurons = [128, 128, 2]
102 self.activation_functions = ["", "tanh", "tanh", ""]
103 self.periodicity = 2 * pi
105 self.learning_rate = 0.001
106 self.n_steps = 100000
107 self.batch_size = 256
108 self.summary_step = 100
109 self.checkpoint_step = 5000
111 self.dist_sig_parameters = (4.5, 12, 6, 1, 2, 6)
112 self.distance_cost_scale = 500
113 self.auto_cost_scale = 1
114 self.auto_cost_variant = "mean_abs"
115 self.center_cost_scale = 0.0001
116 self.l2_reg_constant = 0.001
118 self.gpu_memory_fraction = 0
119 self.analysis_path = ""
120 self.id = ""
123class ADCParameters(Parameters):
124 """
125 This is the parameter object for the AngleDihedralCartesianEncoder. It holds all the parameters that the Parameters
126 object includes, plus the following parameters:
129 :ivar cartesian_pwd_start: index of the first atom to use for the pairwise distance calculation
130 :ivar cartesian_pwd_stop: index of the last atom to use for the pairwise distance calculation
131 :ivar cartesian_pwd_step: step for the calculation of paiwise distances. E.g. for a chain of atoms
132 N-C_a-C-N-C_a-C... cartesian_pwd_start=1 and cartesian_pwd_step=3 will result in using all C-alpha atoms for the
133 pairwise distance calculation.
135 :ivar use_backbone_angles: Allows to define whether backbone bond angles should be learned (True) or if instead mean
136 values should be used to generate conformations (False)
137 :ivar angle_cost_scale: Adjusts how much the angle cost is weighted in the cost function.
138 :ivar angle_cost_variant: Defines how the angle cost is calculated. Must be one of: "mean_square", "mean_abs",
139 "mean_norm"
140 :ivar angle_cost_reference: Can be used to normalize the angle cost with the cost of same reference model (dummy)
142 :ivar dihedral_cost_scale: Adjusts how much the dihedral cost is weighted in the cost function.
143 :ivar dihedral_cost_variant: Defines how the dihedral cost is calculated. Must be one of: "mean_square", "mean_abs",
144 "mean_norm"
145 :ivar dihedral_cost_reference: Can be used to normalize the dihedral cost with the cost of same reference model
146 (dummy)
148 :ivar cartesian_cost_scale: Adjusts how much the cartesian cost is weighted in the cost function.
149 :ivar cartesian_cost_scale_soft_start: Allows to slowly turn on the cartesian cost. Must be a tuple with
150 (begin, ende) or (None, None) If begin and end are given, cartesian_cost_scale will be increased linearly in the
151 given range
152 :ivar cartesian_cost_variant: Defines how the cartesian cost is calculated. Must be one of: "mean_square",
153 "mean_abs", "mean_norm"
154 :ivar cartesian_cost_reference: Can be used to normalize the cartesian cost with the cost of same reference model
155 (dummy)
157 :ivar cartesian_dist_sig_parameters: Parameters for the sigmoid functions applied to the high- and low-dimensional
158 distances in the following order (sig_h, a_h, b_h, sig_l, a_l, b_l)
159 :ivar cartesian_distance_cost_scale: Adjusts how much the cartesian distance cost is weighted in the cost function.
163 """
165 def __init__(self):
166 super(ADCParameters, self).__init__()
168 self.cartesian_pwd_start = None
169 self.cartesian_pwd_stop = None
170 self.cartesian_pwd_step = None
172 self.use_backbone_angles = False
173 self.angle_cost_scale = 0
174 self.angle_cost_variant = "mean_abs"
175 self.angle_cost_reference = 1
177 self.dihedral_cost_scale = 1
178 self.dihedral_cost_variant = "mean_abs"
179 self.dihedral_cost_reference = 1
181 self.cartesian_cost_scale = 1
182 self.cartesian_cost_scale_soft_start = (None, None) # begin, end
183 self.cartesian_cost_variant = "mean_abs"
184 self.cartesian_cost_reference = 1
186 self.cartesian_dist_sig_parameters = self.dist_sig_parameters
187 self.cartesian_distance_cost_scale = 1
189 self.auto_cost_scale = None
190 self.distance_cost_scale = None