-
Notifications
You must be signed in to change notification settings - Fork 0
/
plotting.py
61 lines (45 loc) · 1.33 KB
/
plotting.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
import util
import click
import numpy as np
import typing as T
import matplotlib.pyplot as plt
# traces are a list of epoch->variant->loss_type
TTraces = T.List[T.Mapping[str,T.Mapping[str,T.Any]]]
def plot_traces(
traces: T.Union[TTraces, T.List[TTraces]],
metric_key='loss',
dataset_keys=['train','dev'],
out=None,
title=None):
if isinstance(traces[0], dict):
traces = [traces] # convert to List[TTraces]
def _datas(variant):
ys = np.stack([
np.array([d[variant][metric_key] for d in trace])
for trace in traces
])
ymean = np.mean(ys, axis=0)
yerr = np.sum((ys - ymean)**2, axis=0)
return ymean, yerr
xs = np.arange(len(traces[0]), dtype=int)
plt.figure(tight_layout=True)
if title: plt.title(title)
for key in dataset_keys:
ym, ye = _datas(key)
plt.errorbar(xs, ym, yerr=ye)
plt.legend(dataset_keys)
return _plotsave(out)
def _plotsave(out=None):
if out is not None:
plt.savefig(out)
else:
plt.show()
@click.command()
@util.click_helper.with_path('--out_dir')
def make_aggregate_plot_over_all_seeds(out_dir):
tracess = []
for traces_fname in out_dir.glob('traces*seed*.json'):
tracess.append(util.jsonload(traces_fname))
plot_traces(tracess, out=out_dir/'traces_aggregate.png')
if __name__ == '__main__':
make_aggregate_plot_over_all_seeds()