1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
|
"""
# -*- coding: utf-8 -*-
#
# Copyright 2021 Michael Büsch <m@bues.ch>
#
# Licensed under the Apache License version 2.0
# or the MIT license, at your option.
# SPDX-License-Identifier: Apache-2.0 OR MIT
#
"""
__all__ = [
"Parameter",
"Parameters",
]
from collections import namedtuple
from dataclasses import dataclass
from mlplib.activation import Activation
from mlplib.util import GenericIter
from typing import List, Tuple, Callable, Optional
import numpy as np
Parameter = namedtuple("Parameter", ["w", "b", "actv"])
@dataclass
class Parameters(object):
"""weights: List of np.array with shape (prev_n, n).
biases: List of np.array with shape (1, n).
actvns: List of activation functions.
"""
weights: List[np.ndarray]
biases: List[np.ndarray]
actvns: List[Activation]
@property
def layout(self) -> Tuple[int, ...]:
return tuple(w.shape[1] for w in self.weights)
@property
def nr_inputs(self) -> int:
assert len(self.weights) >= 1
return self.weights[0].shape[0]
@property
def nr_outputs(self) -> int:
assert len(self.weights) >= 1
return self.weights[-1].shape[1]
def __iter__(self):
return ParametersIter(self, len(self.weights))
def __reversed__(self):
return ParametersIter(self, len(self.weights), True, len(self.weights) - 1)
def __str__(self):
ret = []
for i, (w, b, a) in enumerate(zip(self.weights,
self.biases,
self.actvns)):
wstr = "\n ".join(str(w).splitlines())
bstr = "\n ".join(str(b).splitlines())
ret.append(f"L{i} w: {wstr}")
ret.append(f" b: {bstr}")
ret.append(f" {a.__class__.__name__}")
return "\n".join(ret)
@dataclass
class ParametersIter(GenericIter):
def __next__(self):
obj, pos = self._next()
return Parameter(
w=obj.weights[pos],
b=obj.biases[pos],
actv=obj.actvns[pos],
)
# vim: ts=4 sw=4 expandtab
|
