-
Notifications
You must be signed in to change notification settings - Fork 0
/
eval.py
204 lines (157 loc) · 8.15 KB
/
eval.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
import torch.backends.cudnn as cudnn
import torch.optim
import torch.utils.data
import torchvision.transforms as transforms
from datasets import *
from utils import *
from new_utils import *
from nltk.translate.bleu_score import corpus_bleu
import torch.nn.functional as F
from tqdm import tqdm
import json
import argparse
# Parameters
data_folder = 'path_to_data_files' # folder with data files saved by create_input_files.py
data_name = 'flickr8k_5_cap_per_img_5_min_word_freq' # base name shared by data files
checkpoint = 'BEST_checkpoint_flickr8k_5_cap_per_img_5_min_word_freq.pth.tar' # model checkpoint
word_map_file = 'path_to_data_files' + '/WORDMAP_flickr8k_5_cap_per_img_5_min_word_freq.json'
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
cudnn.benchmark = True
captions_dump=True
checkpoint = torch.load(checkpoint)
decoder = checkpoint['decoder']
decoder = decoder.to(device)
decoder.eval()
encoder = checkpoint['encoder']
encoder = encoder.to(device)
encoder.eval()
with open(word_map_file, 'r') as j:
word_map = json.load(j)
rev_word_map = {v: k for k, v in word_map.items()}
vocab_size = len(word_map)
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])
parser = argparse.ArgumentParser(description = 'Evaluation of IC model')
parser.add_argument('beam_size', type=int, help = 'Beam size for evaluation')
args = parser.parse_args()
def evaluate(beam_size):
global captions_dump, data_name
empty_hypo = 0
empty_hypo_r = 0
loader = torch.utils.data.DataLoader(
CaptionDataset(data_folder, data_name, 'TEST', transform=transforms.Compose([normalize])),
batch_size=1, shuffle=True, num_workers=1, pin_memory=True)
references = list()
hypotheses = list()
captions_dict=dict()
image_names = list()
for i, (image, caps, caplens, allcaps, image_name) in enumerate(
tqdm(loader, desc="EVALUATING AT BEAM SIZE " + str(beam_size))):
k = beam_size
image = image.to(device)
encoder_out = encoder(image)
enc_image_size = encoder_out.size(1)
encoder_dim = encoder_out.size(-1)
encoder_out = encoder_out.view(1, -1, encoder_dim)
num_pixels = encoder_out.size(1)
encoder_out = encoder_out.expand(k, num_pixels, encoder_dim)
k_prev_words = torch.LongTensor([[word_map['<start>']]] * k).to(device)
seqs = k_prev_words
top_k_scores = torch.zeros(k, 1).to(device)
complete_seqs = list()
complete_seqs_scores = list()
step = 1
global_img = decoder.get_global_image(encoder_out)
h, c = torch.zeros_like(global_img), torch.zeros_like(global_img)
encoder_out_small = decoder.image_feat_small(encoder_out)
while True:
embeddings = decoder.embedding(k_prev_words).squeeze(1)
x_t = torch.cat([embeddings, global_img], dim = 1)
g_t = decoder.sigmoid(decoder.w_x(x_t) + decoder.w_h(h))
h, c = decoder.decode_step(x_t,(h, c))
s_t = g_t * decoder.tanh(c)
h_new = decoder.w_g(h).unsqueeze(-1)
ones_matrix = torch.ones(k, 1, num_pixels).to(device)
z_t = decoder.w_h_t(decoder.tanh(decoder.w_v(encoder_out_small) + torch.matmul(h_new, ones_matrix))).squeeze(2)
z_t_ex = decoder.w_h_t(decoder.tanh(decoder.w_s(s_t) + decoder.w_g(h)))
alpha_t = decoder.softmax(z_t)
alpha_t_prime = decoder.softmax(torch.cat([z_t, z_t_ex],dim = 1))
beta_t = alpha_t_prime[:,-1:]
one_minus_beta = torch.ones(k, 1).to(device) - beta_t
context_vector = (encoder_out_small * alpha_t.unsqueeze(2)).sum(dim=1)
context_vector_prime = (s_t * beta_t) + (context_vector * one_minus_beta)
# scores = decoder.fc(torch.cat([h, context_vector_prime], dim =1))
scores = decoder.fc(h + context_vector_prime)
scores = F.log_softmax(scores, dim=1)
scores = top_k_scores.expand_as(scores) + scores
if step == 1:
top_k_scores, top_k_words = scores[0].topk(k, 0, True, True)
else:
top_k_scores, top_k_words = scores.view(-1).topk(k, 0, True, True)
prev_word_inds = top_k_words / vocab_size
next_word_inds = top_k_words % vocab_size
seqs = torch.cat([seqs[prev_word_inds], next_word_inds.unsqueeze(1)], dim=1)
incomplete_inds = [ind for ind, next_word in enumerate(next_word_inds) if next_word != word_map['<end>']]
complete_inds = list(set(range(len(next_word_inds))) - set(incomplete_inds))
if len(complete_inds) > 0:
complete_seqs.extend(seqs[complete_inds].tolist())
complete_seqs_scores.extend(top_k_scores[complete_inds])
k -= len(complete_inds)
if k == 0:
break
seqs = seqs[incomplete_inds]
h = h[prev_word_inds[incomplete_inds]]
c = c[prev_word_inds[incomplete_inds]]
encoder_out_small = encoder_out_small[prev_word_inds[incomplete_inds]]
global_img = global_img[prev_word_inds[incomplete_inds]]
encoder_out = encoder_out[prev_word_inds[incomplete_inds]]
top_k_scores = top_k_scores[incomplete_inds].unsqueeze(1)
k_prev_words = next_word_inds[incomplete_inds].unsqueeze(1)
if step > 50:
break
step += 1
try:
i = complete_seqs_scores.index(max(complete_seqs_scores))
seq = complete_seqs[i]
except:
seq = []
empty_hypo += 1
img_caps = allcaps[0].tolist()
img_captions = list(
map(lambda c: [w for w in c if w not in {word_map['<start>'], word_map['<end>'], word_map['<pad>']}],
img_caps)) # remove <start> and pads
references.append(img_captions)
hypotheses.append([w for w in seq if w not in {word_map['<start>'], word_map['<end>'], word_map['<pad>']}])
image_names.append(image_name)
assert len(references) == len(hypotheses) == len(image_names)
# Print the number of hypotheses which remain empty
print('The number of empty hypotheses is {}.\n'.format(empty_hypo))
captions_dict['references']=references
captions_dict['hypotheses']=hypotheses
captions_dict['image_names'] = image_names
if captions_dump==True:
with open('generated_captions_f8k.json', 'w') as gencap:
json.dump(captions_dict, gencap)
save_captions_mscoco_format(word_map_file,references,hypotheses,image_names,str(beam_size)+'_f8ktest')
bleu4 = corpus_bleu(references, hypotheses)
bleu3 = corpus_bleu(references, hypotheses, (1.0/3.0,1.0/3.0,1.0/3.0,))
bleu2 = corpus_bleu(references, hypotheses, (1.0/2.0,1.0/2.0,))
bleu1 = corpus_bleu(references, hypotheses, (1.0/1.0,))
print("The BLEU scores for model are {}.\n".format([bleu1,bleu2,bleu3,bleu4]))
with open('eval_run_logs.txt', 'a') as eval_run:
eval_run.write("For beam-size {} the BLEU scores for model are {}.\n".format(beam_size, [bleu1,bleu2,bleu3,bleu4]))
return bleu1,bleu2,bleu3,bleu4
def main():
beam_size = args.beam_size
was_fine_tuned=False
scores=evaluate(args.beam_size)
print("\nBLEU scores @ beam size of %d is %.4f, %.4f, %.4f, %.4f." % (beam_size, scores[0],scores[1],scores[2],scores[3]))
with open('eval_run_logs.txt', 'a') as eval_run:
eval_run.write('The model is trained on {dataname} and {was} fine tuned.\n'
'The BLEU scores are {bleu_1}, {bleu_2}, {bleu_3}, {bleu_4}.\n'
'The beam_size was {beam}.'
'The model was trained for {epochs} epochs.\n\n\n'.format(dataname=data_name,
was ='was' if was_fine_tuned==True else 'was not',
bleu_1=scores[0], bleu_2=scores[1], bleu_3=scores[2],
bleu_4=scores[3], beam=beam_size, epochs=checkpoint['epoch']))
if __name__ == '__main__':
main()