-
Notifications
You must be signed in to change notification settings - Fork 0
/
results_utils.py
32 lines (30 loc) · 1.19 KB
/
results_utils.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
import torch
from torch import tensor
from warcraft_shortest_path.metrics import is_valid_label_fn_new
def parse_results(fname):
f = open(fname)
lines = f.readlines()[:-1]
mode = -1 # 0 = true, 1 = pred
tensors = [None,None]
strings = ['','']
for l in lines:
if 'grad_fn=<RoundBackward>)' in l or 'True' in l or 'test' in l or 'loss' in l:
continue
if l[:6] == 'tensor':
if mode != -1 and strings[mode] != '':
cur_tensor = eval(strings[mode][:-19]+')')
tensors[mode] = cur_tensor if tensors[mode] is None else torch.cat((tensors[mode],cur_tensor))
strings[mode] = ''
mode = 1-mode if mode != -1 else 0
if mode != -1:
strings[mode] += l
cur_tensor = eval(strings[mode][:-19]+')')
tensors[mode] = cur_tensor if tensors[mode] is None else torch.cat((tensors[mode],cur_tensor))
return (tensors[0],tensors[1])
def percent_valid_paths(fname):
true_paths, pred_paths = parse_results(fname)
pred_count = 0
for i in range(pred_paths.shape[0]):
if is_valid_label_fn_new(pred_paths[i]):
pred_count += 1
return pred_count/pred_paths.shape[0]