-
Notifications
You must be signed in to change notification settings - Fork 0
/
test.py
164 lines (147 loc) · 6.39 KB
/
test.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
import argparse
import os
import torch
import numpy as np
from sklearn.neighbors import NearestNeighbors
from vocab import Vocab
from model import *
from utils import *
from batchify import get_batches
from train import evaluate
parser = argparse.ArgumentParser()
parser.add_argument('--checkpoint', metavar='DIR', required=True,
help='checkpoint directory')
parser.add_argument('--output', metavar='FILE',
help='output file name (in checkpoint directory)')
parser.add_argument('--data', metavar='FILE',
help='path to data file')
parser.add_argument('--enc', default='mu', metavar='M',
choices=['mu', 'z'],
help='encode to mean of q(z|x) or sample z from q(z|x)')
parser.add_argument('--dec', default='greedy', metavar='M',
choices=['greedy', 'sample'],
help='decoding algorithm')
parser.add_argument('--batch-size', type=int, default=256, metavar='N',
help='batch size')
parser.add_argument('--max-len', type=int, default=35, metavar='N',
help='max sequence length')
parser.add_argument('--evaluate', action='store_true',
help='evaluate on data file')
parser.add_argument('--ppl', action='store_true',
help='compute ppl by importance sampling')
parser.add_argument('--reconstruct', action='store_true',
help='reconstruct data file')
parser.add_argument('--sample', action='store_true',
help='sample sentences from prior')
parser.add_argument('--arithmetic', action='store_true',
help='compute vector offset avg(b)-avg(a) and apply to c')
parser.add_argument('--interpolate', action='store_true',
help='interpolate between pairs of sentences')
parser.add_argument('--latent-nn', action='store_true',
help='find nearest neighbor of sentences in the latent space')
parser.add_argument('--m', type=int, default=100, metavar='N',
help='num of samples for importance sampling estimate')
parser.add_argument('--n', type=int, default=5, metavar='N',
help='num of sentences to generate for sample/interpolate')
parser.add_argument('--k', type=float, default=1, metavar='R',
help='k * offset for vector arithmetic')
parser.add_argument('--seed', type=int, default=1111, metavar='N',
help='random seed')
parser.add_argument('--no-cuda', action='store_true',
help='disable CUDA')
def get_model(path):
ckpt = torch.load(path)
train_args = ckpt['args']
model = {'dae': DAE, 'vae': VAE, 'aae': AAE}[train_args.model_type](
vocab, train_args).to(device)
model.load_state_dict(ckpt['model'])
model.flatten()
model.eval()
return model
def encode(sents):
assert args.enc == 'mu' or args.enc == 'z'
batches, order = get_batches(sents, vocab, args.batch_size, device)
z = []
for inputs, _ in batches:
mu, logvar = model.encode(inputs)
if args.enc == 'mu':
zi = mu
else:
zi = reparameterize(mu, logvar)
z.append(zi.detach().cpu().numpy())
z = np.concatenate(z, axis=0)
z_ = np.zeros_like(z)
z_[np.array(order)] = z
return z_
def decode(z):
sents = []
i = 0
while i < len(z):
zi = torch.tensor(z[i: i+args.batch_size], device=device)
outputs = model.generate(zi, args.max_len, args.dec).t()
for s in outputs:
sents.append([vocab.idx2word[id] for id in s[1:]]) # skip <go>
i += args.batch_size
return strip_eos(sents)
def calc_ppl(sents, m):
batches, _ = get_batches(sents, vocab, args.batch_size, device)
total_nll = 0
with torch.no_grad():
for inputs, targets in batches:
total_nll += model.nll_is(inputs, targets, m).sum().item()
n_words = sum(len(s) + 1 for s in sents) # include <eos>
return total_nll / len(sents), np.exp(total_nll / n_words)
if __name__ == '__main__':
args = parser.parse_args()
vocab = Vocab(os.path.join(args.checkpoint, 'vocab.txt'))
set_seed(args.seed)
cuda = not args.no_cuda and torch.cuda.is_available()
device = torch.device("cuda" if cuda else "cpu")
model = get_model(os.path.join(args.checkpoint, 'model.pt'))
if args.evaluate:
sents = load_sent(args.data)
batches, _ = get_batches(sents, vocab, args.batch_size, device)
meters = evaluate(model, batches)
print(' '.join(['{} {:.2f},'.format(k, meter.avg)
for k, meter in meters.items()]))
if args.ppl:
sents = load_sent(args.data)
nll, ppl = calc_ppl(sents, args.m)
print('NLL {:.2f}, PPL {:.2f}'.format(nll, ppl))
if args.sample:
z = np.random.normal(size=(args.n, model.args.dim_z)).astype('f')
sents = decode(z)
write_sent(sents, os.path.join(args.checkpoint, args.output))
if args.reconstruct:
sents = load_sent(args.data)
z = encode(sents)
sents_rec = decode(z)
write_z(z, os.path.join(args.checkpoint, args.output+'.z'))
write_sent(sents_rec, os.path.join(args.checkpoint, args.output+'.rec'))
if args.arithmetic:
fa, fb, fc = args.data.split(',')
sa, sb, sc = load_sent(fa), load_sent(fb), load_sent(fc)
za, zb, zc = encode(sa), encode(sb), encode(sc)
zd = zc + args.k * (zb.mean(axis=0) - za.mean(axis=0))
sd = decode(zd)
write_sent(sd, os.path.join(args.checkpoint, args.output))
if args.interpolate:
f1, f2 = args.data.split(',')
s1, s2 = load_sent(f1), load_sent(f2)
z1, z2 = encode(s1), encode(s2)
zi = [interpolate(z1_, z2_, args.n) for z1_, z2_ in zip(z1, z2)]
zi = np.concatenate(zi, axis=0)
si = decode(zi)
si = list(zip(*[iter(si)]*(args.n)))
write_doc(si, os.path.join(args.checkpoint, args.output))
if args.latent_nn:
sents = load_sent(args.data)
z = encode(sents)
with open(os.path.join(args.checkpoint, args.output), 'w') as f:
nn = NearestNeighbors(n_neighbors=args.n).fit(z)
dis, idx = nn.kneighbors(z[:args.m])
for i in range(len(idx)):
f.write(' '.join(sents[i]) + '\n')
for j, d in zip(idx[i], dis[i]):
f.write(' '.join(sents[j]) + '\t%.2f\n' % d)
f.write('\n')