-
Notifications
You must be signed in to change notification settings - Fork 3
/
metrics.py
314 lines (258 loc) · 9.77 KB
/
metrics.py
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
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
"""
A module to handle metrics
"""
import copy
import json
import numbers
from functools import partial
from typing import Any, Dict
import numpy as np
from utils import pairwise
def format_time(seconds):
""" Format time in h:mm:ss.ss format """
hour = 60 * 60
hours = int(seconds // hour)
minutes = int((seconds % hour) // 60)
return f"{hours}:{minutes:>02}:{seconds % 60.:>05.2f}"
def format_basic(value, format_spec=""):
""" Wrapper around format() for use in functools.partial """
return format(value, format_spec)
def format_dynamic(value, format_funcs=(format_basic,)):
"""
Wrapper around a number of format functions that chooses the shortest
output string.
"""
return sorted((f(value) for f in format_funcs), key=len)[0]
# pylint:disable=invalid-name
format_int = partial(format_basic, format_spec=".0f")
format_percent = partial(format_basic, format_spec=".1%")
format_float = partial(format_basic, format_spec=".1f")
format_scientific = partial(format_basic, format_spec=".3g")
format_dynamic_float = partial(
format_dynamic, format_funcs=(format_float, format_scientific)
)
# pylint:enable=invalid-name
FORMATTERS = {
"format_int": format_int,
"format_time": format_time,
"format_basic": format_basic,
"format_dynamic": format_dynamic,
"format_percent": format_percent,
"format_float": format_float,
"format_scientific": format_scientific,
"format_dynamic_float": format_dynamic_float,
}
class Metric:
""" Class that represents a metric """
def __init__(
self,
name,
formatter="format_basic",
default_format_str="g(a)",
max_history=None,
):
self.name = name
self.max_history = max_history
self._formatter = formatter
self.default_format_str = default_format_str
self.counts, self.values, self.min, self.max = self.reset()
@classmethod
def from_dict(cls, state: Dict[str, Any]):
"""
Create a metric from the passed in dictionary
"""
metric = Metric("")
metric.__dict__.update(state)
return metric
@property
def formatter(self):
"""
Get the formatter function for this metric
"""
return FORMATTERS[self._formatter]
def reset(self):
""" Reset the metrics """
self.counts = []
self.values = []
self.min = float("inf")
self.max = float("-inf")
return self.counts, self.values, self.min, self.max
def update(self, value, count=1):
""" Update the value and counts """
self.counts.append(count)
self.values.append(value)
average = value / count
self.min = min(self.min, average)
self.max = max(self.max, average)
if self.max_history and len(self.counts) > self.max_history:
self.counts = self.counts[1:]
self.values = self.values[1:]
def updates(self, values, counts=1):
""" Update multiple values at once """
if isinstance(counts, numbers.Number):
counts = [counts] * len(values)
self.counts.extend(counts)
self.values.extend(values)
if self.max_history:
# pylint thinks self.max_history is None...
# pylint:disable=invalid-unary-operand-type
self.counts = self.counts[-self.max_history :]
self.values = self.values[-self.max_history :]
# pylint:enable=invalid-unary-operand-type
averages = [value / count for count, value in zip(counts, values)]
self.min = min(self.min, min(averages))
self.max = max(self.max, max(averages))
@property
def last_count(self):
""" Return the last recorded count of the metric"""
# fancy way to return the last count or zero
return len(self.counts) and self.counts[-1]
@property
def last_value(self):
""" Return the last recorded value of the metric """
# fancy way to return the last value or zero
return len(self.values) and self.values[-1]
@property
def last_average(self):
""" Return the last recorded value of the metric """
# fancy way to return the last value or zero
return self.last_value / max(self.last_count, 1)
@property
def total(self):
""" Return the current total """
return sum(self.values)
@property
def total_count(self):
""" Return the current total count """
return sum(self.counts)
@property
def average(self):
""" Return the current average value """
return self.total / max(self.total_count, 1)
@property
def var(self):
""" Return the variance of the values """
# Need to use a weighted average since each value has an associated count
counts = np.array(self.counts)
values = np.array(self.values)
weights = counts / self.total_count
return np.average((values - self.average) ** 2, weights=weights)
@property
def std(self):
""" Return the standard deviation of the values """
return np.sqrt(self.var)
def __format__(self, format_str):
""" Return a formatted version of the metric """
format_str = format_str or self.default_format_str
formatted = f"{self.name}="
compact = True
paren_depth = 0
for format_spec, next_format_spec in pairwise(format_str, True):
if format_spec == "l":
compact = False
elif format_spec == "c":
compact = True
elif format_spec == "(":
formatted += "("
paren_depth += 1
elif format_spec == ")":
formatted += ")"
paren_depth -= 1
elif format_spec == "C":
if not compact:
formatted += f"last_count="
formatted += f"{self.formatter(self.last_count)}"
elif format_spec == "V":
if not compact:
formatted += f"last_value="
formatted += f"{self.formatter(self.last_value)}"
elif format_spec == "g":
if not compact:
formatted += f"last_avg="
formatted += f"{self.formatter(self.last_average)}"
elif format_spec == "a":
if not compact:
formatted += f"avg="
formatted += f"{self.formatter(self.average)}"
elif format_spec == "t":
if not compact:
formatted += f"total="
formatted += f"{self.formatter(self.total)}"
elif format_spec == "m":
if not compact:
formatted += f"min="
formatted += f"{self.formatter(self.min)}"
elif format_spec == "x":
if not compact:
formatted += f"max="
formatted += f"{self.formatter(self.max)}"
elif format_spec == "s":
if not compact:
formatted += f"std="
formatted += f"{self.formatter(self.std)}"
elif format_spec == "v":
if not compact:
formatted += f"var="
formatted += f"{self.formatter(self.var)}"
else:
raise ValueError(f"Unknown format specifier {format_spec}")
if paren_depth and format_spec != "(" and next_format_spec != ")":
formatted += ","
if not compact:
formatted += " "
return formatted
def __str__(self):
""" Return a string representation of the metric """
return self.__format__(self.default_format_str)
class MetricStore(object):
""" A collection of metrics """
def __init__(self, default_format_str="c"):
super(MetricStore, self).__init__()
self.metrics = {}
self.default_format_str = default_format_str
def keys(self):
""" Return the metrics keys """
return self.metrics.keys()
def values(self):
""" Return the metrics values """
return self.metrics.values()
def items(self):
""" Return the metrics items """
return self.metrics.items()
def __getitem__(self, key):
""" Return the requested metric """
return self.metrics[key]
def __contains__(self, key):
""" See if we are tracking the named metric """
return key in self.metrics
def __len__(self):
""" Count of the metrics being tracked """
return len(self.metrics)
def add(self, metric: Metric):
""" Adds a copy of the Metric to the store if it does not already exist """
if metric.name not in self.metrics:
self.metrics[metric.name] = copy.deepcopy(metric)
def save(self, path):
""" Save the metrics to disk """
with open(path, "wt") as metric_file:
json.dump(
self.metrics,
metric_file,
indent=2,
default=lambda obj: getattr(obj, "__dict__", {}),
)
def load(self, path):
""" Load the metrics from disk """
with open(path, "rt") as metric_file:
for name, metric_state in json.load(metric_file).items():
self.metrics[name] = Metric.from_dict(metric_state)
def __str__(self):
""" Return a string representation of the metric store """
return self.__format__(self.default_format_str)
def __format__(self, format_str):
""" Return a formatted version of the metric """
format_str = format_str or self.default_format_str
if format_str == "l":
return "\n".join(str(m) for m in self.metrics.values())
else:
return ", ".join(str(m) for m in self.metrics.values())