-
Notifications
You must be signed in to change notification settings - Fork 6
/
eval.py
213 lines (186 loc) · 8.36 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
205
206
207
208
209
210
211
212
213
from __future__ import absolute_import
from __future__ import division
from __future__ import print_function
# import os
# os.environ["CUDA_VISIBLE_DEVICES"] = "3"
import json
import numpy as np
import time
from six.moves import cPickle
import ipdb
import models
import opts
# from dataloader_relative import *
from dataloader import *
import misc.utils as utils
# import misc.utils as utils
import eval_utils_h as eval_utils
# import eval_utils
from eval_online import eval_online
from dataloaderraw import *
import argparse
import torch
try:
import tensorboardX as tb
except ImportError:
print("tensorboardX is not installed")
tb = None
def add_summary_value(writer, key, value, iteration):
if writer:
writer.add_scalar(key, value, iteration)
# Input arguments and options
parser = argparse.ArgumentParser()
# Input paths
# parser.add_argument('--model', type=str, default='log/tmp/train_ours/log_refine_aoa_{}_aoa3_new/model_22.pth',
# help='path to model to evaluate')
# parser.add_argument('--cnn_model', type=str, default='resnet101',
# help='resnet101, resnet152')
# parser.add_argument('--infos_path', type=str, default='log/tmp/train_ours/log_refine_aoa_{}_aoa3_new/infos_22.pkl',
# help='path to infos to evaluate')
parser.add_argument('--model', type=str, default='log/tmp/train_ours/log_refine_aoa_{}_aoa{}{}/model_{}.pth',
help='path to model to evaluate')
parser.add_argument('--cnn_model', type=str, default='resnet101',
help='resnet101, resnet152')
parser.add_argument('--infos_path', type=str, default='log/tmp/train_ours/log_refine_aoa_{}_aoa{}{}/infos_{}.pkl',
help='path to infos to evaluate')
# parser.add_argument('--model_id', type=str, default=None, help='which specifi chech point to load')
# parser.add_argument('--append_info', type=str, default=None, help="such as old/new")
opts.add_eval_options(parser)
opt = parser.parse_args()
# opt.dump_images = 1
# opt.dump_json = 1
# opt.num_images = 1
opt.language_eval = 1
opt.beam_size = 3
opt.batch_size = 500
opt.split = 'test'
opt.test_online = 0
opt.use_val = getattr(opt, 'use_val', 0)
opt.use_test = getattr(opt, 'use_test', 0)
save_results = 0
val_and_test = 0
# if opt.use_test:
# opt.split = 'val'
aoa_id = '3d1'
aoa_num = 3
append_info = '_new10_new1_37_rl'
opt.caption_model = 'aoa' + aoa_id
opt.id = 'h_v' + aoa_id
opt.input_flag_dir = 'data/tmp/cocobu_flag_h_v1'
model_ids = ['best_720000']#+list(range(70, 81))
best_cider = -1
best_epoch = -1
write_summary = False
print("============================================")
print("=========beam search size:{}=================".format(opt.beam_size))
print("=========test on {} set================".format(opt.split))
if write_summary:
# summary_path = 'log/tmp/train_ours/log_refine_aoa_{}_aoa{}{}'.format(opt.id, aoa_num, append_info)
summary_path = 'log/tmp/train_ours/log_refine_{}'.format(opt.id)
print("write test summary to {}".format(summary_path))
tb_summary_writer = tb and tb.SummaryWriter(summary_path)
else:
tb_summary_writer = None
for model_id in model_ids:
opt.model = 'log/tmp/train_ours/log_refine_aoa_{}_aoa{}{}/model_{}.pth'.format(opt.id, aoa_num, append_info, model_id)
opt.infos_path = 'log/tmp/train_ours/log_refine_aoa_{}_aoa{}{}/infos_{}.pkl'.format(opt.id, aoa_num, append_info, model_id)
test_result = {}
# Load infos
print("Evluation using infor_path:{}".format(opt.infos_path))
with open(opt.infos_path, 'rb') as f:
infos = utils.pickle_load(f)
iteration = infos.get('iter', 0)
epoch = infos.get('epoch', 0)
print("=====start from {} epoch-- {} iterations=============".format(epoch, iteration))
print("=====refine aoa: {} ==========".format(infos['opt'].refine_aoa))
# print("=====aoa num: {} ==========".format(infos['opt'].aoa_num))
print("=====learning rate decay every: {} ==========".format(infos['opt'].learning_rate_decay_every))
# caption_model = getattr(infos['opt'], 'caption_model','')
# print("caption model: {}".format(caption_model))
# override and collect parameters
# replace = ['input_fc_dir', 'input_att_dir', 'input_box_dir',
# 'input_label_h5', 'input_json', 'batch_size', 'id']
ignore = ['start_from']
if not opt.test_online:
replace = ['input_fc_dir', 'input_att_dir', 'input_box_dir', 'input_flag_dir',
'input_label_h5', 'input_json', 'batch_size', 'id']
else:
replace = ['input_json', 'batch_size', 'id']
for k in vars(infos['opt']).keys():
if k in replace:
setattr(opt, k, getattr(opt, k) or getattr(infos['opt'], k, ''))
elif k not in ignore:
if k not in vars(opt):
# copy over options from model
vars(opt).update({k: vars(infos['opt'])[k]})
# infos['opt'].use_val=0
# infos['opt'].use_test=0
vocab = infos['vocab'] # ix -> word mapping
# Setup the model
opt.vocab = vocab
# opt.caption_model = 'transformer'
model = models.setup(opt)
del opt.vocab
print("Evluation using model:{}".format(opt.model))
model.load_state_dict(torch.load(opt.model))
model.cuda()
model.eval()
crit = utils.LanguageModelCriterion()
# ipdb.set_trace()
# Create the Data Loader instance
if len(opt.image_folder) == 0:
loader = DataLoader(opt)
else:
loader = DataLoaderRaw({'folder_path': opt.image_folder,
'coco_json': opt.coco_json,
'batch_size': opt.batch_size,
'cnn_model': opt.cnn_model})
# When eval using provided pretrained model, the vocab may be different from what you have in your cocotalk.json
# So make sure to use the vocab in infos file.
loader.ix_to_word = infos['vocab']
if val_and_test:
loader.split_ix['val'] = loader.split_ix['test'] + loader.split_ix['val']
loader.split_ix['test'] = loader.split_ix['val']
# loader.split_ix['val'] = loader.split_ix['val'][:1000]
# ipdb.set_trace()
# Set sample options
opt.datset = opt.input_json
print(opt.language_eval)
try:
loss, split_predictions, lang_stats = eval_utils.eval_split(model, crit, loader, vars(opt))
except OSError:
loss, split_predictions, lang_stats = eval_utils.eval_split(model, crit, loader, vars(opt))
current_score = lang_stats['CIDEr']
if best_cider < current_score:
best_cider = current_score
best_epoch = model_id
if tb_summary_writer is not None:
add_summary_value(tb_summary_writer, 'loss/test loss', loss, iteration)
if lang_stats is not None:
bleu_dict = {}
for k, v in lang_stats.items():
if 'Bleu' in k:
bleu_dict[k] = v
if len(bleu_dict) > 0:
tb_summary_writer.add_scalars('test/Bleu', bleu_dict, epoch)
for k, v in lang_stats.items():
if 'Bleu' not in k:
add_summary_value(tb_summary_writer, 'test/' + k, v, iteration)
if save_results:
model_name = opt.model.split('/')[-2].split('h_')[-1] + '_' + 'bs{}_{}'.format(opt.beam_size, model_id)
cache_path = os.path.join('eval_results', opt.split, model_name + '.json')
with open(cache_path, 'w') as f:
json.dump(split_predictions, f)
print("Write the results file to {}".format(cache_path))
# if opt.dump_json == 1:
# # dump the json
# model_name = os.path.basename(opt.model).split('.')[0]
# result_file = '_'.join(['bs'+str(opt.beam_size), model_name])
# json.dump(split_predictions, open('vis/h/vis-{}.json'.format(result_file), 'w'))
# print("Dump the json to {}".format('vis/h/vis-{}.json'.format(result_file)))
print("=================================================================")
print("===================Evalluation {} DONE!======================".format(opt.model))
print("===================Best {} Cider = {:.5f} in epoch {}!======================".format(opt.split, best_cider, best_epoch))
print("======Best Val Cider = {:.4f} in epoch {}: iter {}!======".format(infos.get('best_val_score', -1), infos.get('best_epoch', None), infos.get('best_itr', None)))
print("=================================================================")
del model