train result

[Dingxiangtao]

作者你好，训练结果并不能达到论文中的结果，请问是有哪些可能的影响因素

[mrazhou]

你好，按照README里给出的训练命令应该是可以复现论文结果的。
是否在训练中对改命令有修改？比如Batch Size，Epochs、LR。
除了这个以外，可能一些库的版本也有影响。建议使用README里给出的环境按照命令。

[Dingxiangtao]

好的，感谢回复

[Dingxiangtao]

[Info] With message: _IncompatibleKeys(missing_keys=['fc.weight', 'fc.bias'], unexpected_keys=['', 'fc.0.weight', 'fc.0.bias', 'fc.2.weight', 'fc.2.bias']) 作者你好，SEN在训练中使用rn50.pth.tar时会报上面这个警告，请问这是什么原因？是这个导致了复现结果不佳吗

[mrazhou]

应该不是，你现在复现的指标是多少？可以把训练日志发一下。

[Dingxiangtao]

[Info] Load weight from pretrain_ckpt/rn50.pth.tar
[Info] With message: _IncompatibleKeys(missing_keys=['fc.weight', 'fc.bias'], unexpected_keys=['', 'fc.0.weight', 'fc.0.bias', 'fc.2.weight', 'fc.2.bias'])
git:
sha: N/A, status: clean, branch: N/A
[CMD] python train.py
--more_reproducibility
--savepath model_checkpoints/SEN
--model SEN
--batch_size 128
--proj_channel 512
--encoder_n_layers 2
--ft_layer 4
--model_stage 4
--weight_path pretrain_ckpt/rn50.pth.tar

[args]
Namespace(Split='VAL', batch_size=128, beam_size=1, checkpoint=None, data_folder='./data', data_name='LEVIR_CC_5_cap_per_img_5_min_word_freq', decay_interval=3, decay_lr=False, decay_rate=0.7, decoder='SEN_TransDecoder', decoder_lr=0.0002, decoder_n_layers=1, dropout=0.5, encoder_feat='SEN_DiffEncoder', encoder_feat_clip=False, encoder_feat_decay=False, encoder_image='SEN_Extractor', encoder_lr=0.0002, encoder_n_layers=2, epochs=40, feature_dim_de=1024, ft_layer=4, grad_clip=5.0, model='SEN', model_stage=4, more_reproducibility=True, n_heads=8, n_layers=3, print_freq=50, proj_channel=512, savepath='model_checkpoints/SEN', seed=42, stop_criteria=40, weight_path='pretrain_ckpt/rn50.pth.tar', workers=8)
03-06 09 : 36 : 12
--------Params--------
encoder_image params:24,547,584
encoder_feat params:10,582,016
decoder params:4,710,356
Total trainable params:39,839,956
Total params:39,858,900
--------Params--------
�[32m[Info]�[0m Checkpoint_savepath:model_checkpoints/SEN
Encoder_image_mode:SEN_Extractor Encoder_feat_mode:SEN_DiffEncoder Decoder_mode:SEN_TransDecoder
encoder_layers 3 decoder_layers 1 n_heads 8 dropout 0.5 encoder_lr 0.0002 decoder_lr 0.0002
03-06 09 : 36 : 17
Epoch: 0/40 step: 0/267 Loss: 8.0012 AVG_Loss: 8.0012 Top-5 Accuracy: 0.1764 Batch_time: 8.6444s
Epoch: 0/40 step: 50/267 Loss: 4.6206 AVG_Loss: 5.1309 Top-5 Accuracy: 12.8617 Batch_time: 0.7115s
Epoch: 0/40 step: 100/267 Loss: 3.9888 AVG_Loss: 4.6522 Top-5 Accuracy: 19.0111 Batch_time: 0.7083s
Epoch: 0/40 step: 150/267 Loss: 3.4944 AVG_Loss: 4.2906 Top-5 Accuracy: 31.0518 Batch_time: 1.0556s
Epoch: 0/40 step: 200/267 Loss: 2.6192 AVG_Loss: 3.9925 Top-5 Accuracy: 45.0581 Batch_time: 1.0737s
Epoch: 0/40 step: 250/267 Loss: 2.6036 AVG_Loss: 3.7315 Top-5 Accuracy: 44.2424 Batch_time: 1.1034s
eval_time: 1072.9100308418274
avg eval_time: 0.8048837440673874
fps: 1.2424154511390864
extractor_time: 0.013315576945402885
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.97898348702531, 'Bleu_2': 0.9754749485786772, 'Bleu_3': 0.9740411345348179, 'Bleu_4': 0.973024570160429, 'METEOR': 0.7554497257788143, 'ROUGE_L': 0.9792271549214834, 'CIDEr': 0.0}
nochange_acc: 0.9654654654654654
change_metric: {'Bleu_1': 0.5608660771781455, 'Bleu_2': 0.4054967771018916, 'Bleu_3': 0.2432777264959849, 'Bleu_4': 0.14824533717332086, 'METEOR': 0.15824483208399892, 'ROUGE_L': 0.3732533301172048, 'CIDEr': 0.1662799968683784}
change_acc: 0.8170914542728636
.......................................................
Bleu_1 0.7578428988332185
Bleu_2 0.6805712138116797
Bleu_3 0.6145317463407975
Bleu_4 0.5754030323754038
METEOR 0.33915889864883836
ROUGE_L 0.6760129455107904
CIDEr 1.1121800905774646

The best BLEU-4 score: 0.5754

03-06 09 : 58 : 51
Epoch: 1/40 step: 0/267 Loss: 2.3670 AVG_Loss: 2.3670 Top-5 Accuracy: 49.0248 Batch_time: 3.1967s
Epoch: 1/40 step: 50/267 Loss: 1.8408 AVG_Loss: 2.3577 Top-5 Accuracy: 55.5449 Batch_time: 1.0931s
Epoch: 1/40 step: 100/267 Loss: 2.0534 AVG_Loss: 2.2960 Top-5 Accuracy: 53.5363 Batch_time: 1.0868s
Epoch: 1/40 step: 150/267 Loss: 2.1565 AVG_Loss: 2.2239 Top-5 Accuracy: 49.4355 Batch_time: 1.0899s
Epoch: 1/40 step: 200/267 Loss: 1.7773 AVG_Loss: 2.1695 Top-5 Accuracy: 57.3171 Batch_time: 1.0567s
Epoch: 1/40 step: 250/267 Loss: 1.8032 AVG_Loss: 2.1224 Top-5 Accuracy: 55.0141 Batch_time: 1.0964s
eval_time: 1165.5740501880646
avg eval_time: 0.8743991374254048
fps: 1.1436424822471996
extractor_time: 0.014108708990487674
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.928899082568541, 'Bleu_2': 0.9116674591979416, 'Bleu_3': 0.9001504318535332, 'Bleu_4': 0.8879544412427911, 'METEOR': 0.6598685556671879, 'ROUGE_L': 0.9505972285204504, 'CIDEr': 0.0}
nochange_acc: 0.9309309309309309
change_metric: {'Bleu_1': 0.6077230015907715, 'Bleu_2': 0.4848303206410436, 'Bleu_3': 0.36034726847546594, 'Bleu_4': 0.25197163204847745, 'METEOR': 0.19207611203924554, 'ROUGE_L': 0.41206161249701706, 'CIDEr': 0.3373905481981422}
change_acc: 0.8590704647676162
.......................................................
Bleu_1 0.7336725377466506
Bleu_2 0.6493004119994912
Bleu_3 0.5664255888290922
Bleu_4 0.4898595154256354
METEOR 0.3202509895355005
ROUGE_L 0.6811274191523858
CIDEr 1.157146453403504

Epochs since last improvement: 1

The best BLEU-4 score: 0.5754

03-06 10 : 23 : 36
Epoch: 2/40 step: 0/267 Loss: 1.8382 AVG_Loss: 1.8382 Top-5 Accuracy: 55.7018 Batch_time: 4.0136s
Epoch: 2/40 step: 50/267 Loss: 1.6723 AVG_Loss: 1.8829 Top-5 Accuracy: 58.1132 Batch_time: 1.0801s
Epoch: 2/40 step: 100/267 Loss: 1.6791 AVG_Loss: 1.8554 Top-5 Accuracy: 58.9883 Batch_time: 1.0340s
Epoch: 2/40 step: 150/267 Loss: 1.7844 AVG_Loss: 1.8258 Top-5 Accuracy: 55.6818 Batch_time: 1.0531s
Epoch: 2/40 step: 200/267 Loss: 1.7021 AVG_Loss: 1.8096 Top-5 Accuracy: 58.4229 Batch_time: 1.0974s
Epoch: 2/40 step: 250/267 Loss: 1.5847 AVG_Loss: 1.7960 Top-5 Accuracy: 61.3825 Batch_time: 1.1050s
eval_time: 1072.5000522136688
avg eval_time: 0.8045761832060532
fps: 1.242890382381476
extractor_time: 0.013327445826491108
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9240223463684569, 'Bleu_2': 0.9067712078474487, 'Bleu_3': 0.8949904530453782, 'Bleu_4': 0.8827513325239496, 'METEOR': 0.6627907111040264, 'ROUGE_L': 0.9482992357838275, 'CIDEr': 0.0}
nochange_acc: 0.9309309309309309
change_metric: {'Bleu_1': 0.6368341048538405, 'Bleu_2': 0.5076299588266221, 'Bleu_3': 0.38727100988316016, 'Bleu_4': 0.29140141649242474, 'METEOR': 0.21156814093707904, 'ROUGE_L': 0.4494758064092313, 'CIDEr': 0.41269244528959215}
change_acc: 0.8575712143928036
.......................................................
Bleu_1 0.7506165565025187
Bleu_2 0.6622489620496043
Bleu_3 0.5808127153524137
Bleu_4 0.5094239291490528
METEOR 0.33538574903814294
ROUGE_L 0.6987004155341233
CIDEr 1.195789152401825

Epochs since last improvement: 2

The best BLEU-4 score: 0.5754

03-06 10 : 46 : 55
Epoch: 3/40 step: 0/267 Loss: 1.7929 AVG_Loss: 1.7929 Top-5 Accuracy: 57.8947 Batch_time: 3.0312s
Epoch: 3/40 step: 50/267 Loss: 1.7408 AVG_Loss: 1.6456 Top-5 Accuracy: 56.4081 Batch_time: 1.0832s
Epoch: 3/40 step: 100/267 Loss: 1.7381 AVG_Loss: 1.6597 Top-5 Accuracy: 56.8807 Batch_time: 1.0840s
Epoch: 3/40 step: 150/267 Loss: 1.6252 AVG_Loss: 1.6470 Top-5 Accuracy: 57.0312 Batch_time: 0.9919s
Epoch: 3/40 step: 200/267 Loss: 1.6225 AVG_Loss: 1.6460 Top-5 Accuracy: 58.0372 Batch_time: 1.0694s
Epoch: 3/40 step: 250/267 Loss: 1.6851 AVG_Loss: 1.6415 Top-5 Accuracy: 56.9295 Batch_time: 1.0357s
eval_time: 1178.2808005809784
avg eval_time: 0.8839315833315667
fps: 1.1313092764837835
extractor_time: 0.014404249835175315
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9272989889261127, 'Bleu_2': 0.9129361987280233, 'Bleu_3': 0.9046752863810968, 'Bleu_4': 0.8980537888252855, 'METEOR': 0.6672301398148054, 'ROUGE_L': 0.9400586876049545, 'CIDEr': 0.0}
nochange_acc: 0.9204204204204204
change_metric: {'Bleu_1': 0.6370276278182538, 'Bleu_2': 0.5080941450330193, 'Bleu_3': 0.3959925161193523, 'Bleu_4': 0.29624818449206436, 'METEOR': 0.21373695569860757, 'ROUGE_L': 0.44801568720999674, 'CIDEr': 0.43900940072142536}
change_acc: 0.8695652173913043
.......................................................
Bleu_1 0.7523918867201729
Bleu_2 0.6685407154939303
Bleu_3 0.5966964716215859
Bleu_4 0.531669968943917
METEOR 0.34765010940032715
ROUGE_L 0.6938526251417612
CIDEr 1.2055146572279416

Epochs since last improvement: 3

The best BLEU-4 score: 0.5754

[Decay LR] The new learning rate is 0.000140

03-06 11 : 11 : 54
Epoch: 4/40 step: 0/267 Loss: 1.6531 AVG_Loss: 1.6531 Top-5 Accuracy: 58.1104 Batch_time: 2.9274s
Epoch: 4/40 step: 50/267 Loss: 1.4457 AVG_Loss: 1.5772 Top-5 Accuracy: 62.3056 Batch_time: 1.1185s
Epoch: 4/40 step: 100/267 Loss: 1.6003 AVG_Loss: 1.5709 Top-5 Accuracy: 60.4478 Batch_time: 1.0941s
Epoch: 4/40 step: 150/267 Loss: 1.6750 AVG_Loss: 1.5703 Top-5 Accuracy: 55.8442 Batch_time: 1.0629s
Epoch: 4/40 step: 200/267 Loss: 1.5395 AVG_Loss: 1.5550 Top-5 Accuracy: 61.4630 Batch_time: 1.0940s
Epoch: 4/40 step: 250/267 Loss: 1.4902 AVG_Loss: 1.5483 Top-5 Accuracy: 61.5517 Batch_time: 1.1053s
eval_time: 1068.9542517662048
avg eval_time: 0.8019161678666202
fps: 1.2470131418603925
extractor_time: 0.013304265745582208
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9450675870000158, 'Bleu_2': 0.9342290238655554, 'Bleu_3': 0.9259315759971984, 'Bleu_4': 0.9167609369072411, 'METEOR': 0.7053480009118464, 'ROUGE_L': 0.9640842820891987, 'CIDEr': 0.0}
nochange_acc: 0.9534534534534534
change_metric: {'Bleu_1': 0.6493555870713245, 'Bleu_2': 0.5178218874139312, 'Bleu_3': 0.4054311488525256, 'Bleu_4': 0.30117111577097055, 'METEOR': 0.21525679730941474, 'ROUGE_L': 0.4469557708432035, 'CIDEr': 0.47253433979217635}
change_acc: 0.8455772113943029
.......................................................
Bleu_1 0.7611376702305912
Bleu_2 0.6713974731907257
Bleu_3 0.5918764707270889
Bleu_4 0.5129877438251411
METEOR 0.34071354803823284
ROUGE_L 0.7053260547815626
CIDEr 1.247473579247893

Epochs since last improvement: 4

The best BLEU-4 score: 0.5754

03-06 11 : 35 : 02
Epoch: 5/40 step: 0/267 Loss: 1.5933 AVG_Loss: 1.5933 Top-5 Accuracy: 58.4251 Batch_time: 2.6036s
Epoch: 5/40 step: 50/267 Loss: 1.7635 AVG_Loss: 1.5210 Top-5 Accuracy: 55.8347 Batch_time: 1.0807s
Epoch: 5/40 step: 100/267 Loss: 1.6150 AVG_Loss: 1.5020 Top-5 Accuracy: 55.5169 Batch_time: 1.0754s
Epoch: 5/40 step: 150/267 Loss: 1.5999 AVG_Loss: 1.5022 Top-5 Accuracy: 58.4775 Batch_time: 1.0444s
Epoch: 5/40 step: 200/267 Loss: 1.6449 AVG_Loss: 1.4964 Top-5 Accuracy: 55.7420 Batch_time: 1.0638s
Epoch: 5/40 step: 250/267 Loss: 1.4333 AVG_Loss: 1.4925 Top-5 Accuracy: 61.7864 Batch_time: 1.1070s
eval_time: 1167.7552030086517
avg eval_time: 0.8760354111092661
fps: 1.1415063675722488
extractor_time: 0.014010444585309383
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.939834616481055, 'Bleu_2': 0.9290244444054118, 'Bleu_3': 0.9204196008673176, 'Bleu_4': 0.9108421929796708, 'METEOR': 0.7045095061965451, 'ROUGE_L': 0.9613141544076321, 'CIDEr': 0.0}
nochange_acc: 0.9504504504504504
change_metric: {'Bleu_1': 0.6708300758887558, 'Bleu_2': 0.5482390185754903, 'Bleu_3': 0.43360478911100425, 'Bleu_4': 0.3312601883911612, 'METEOR': 0.22427937734642844, 'ROUGE_L': 0.46436442133121075, 'CIDEr': 0.49358515298210764}
change_acc: 0.8395802098950524
.......................................................
Bleu_1 0.7748267027923023
Bleu_2 0.6913501742531027
Bleu_3 0.6116329700802425
Bleu_4 0.5345770151556782
METEOR 0.3481228169707617
ROUGE_L 0.7126528851188301
CIDEr 1.2624049515376152

Epochs since last improvement: 5

The best BLEU-4 score: 0.5754

03-06 11 : 59 : 50
Epoch: 6/40 step: 0/267 Loss: 1.4014 AVG_Loss: 1.4014 Top-5 Accuracy: 61.3924 Batch_time: 2.9847s
Epoch: 6/40 step: 50/267 Loss: 1.3549 AVG_Loss: 1.4546 Top-5 Accuracy: 63.6771 Batch_time: 1.0712s
Epoch: 6/40 step: 100/267 Loss: 1.4733 AVG_Loss: 1.4650 Top-5 Accuracy: 59.2405 Batch_time: 1.0158s
Epoch: 6/40 step: 150/267 Loss: 1.5532 AVG_Loss: 1.4575 Top-5 Accuracy: 58.2245 Batch_time: 1.0201s
Epoch: 6/40 step: 200/267 Loss: 1.5254 AVG_Loss: 1.4523 Top-5 Accuracy: 59.0476 Batch_time: 1.0576s
Epoch: 6/40 step: 250/267 Loss: 1.4939 AVG_Loss: 1.4485 Top-5 Accuracy: 59.1979 Batch_time: 1.1187s
eval_time: 1072.0830118656158
avg eval_time: 0.8042633247303945
fps: 1.2433738668056515
extractor_time: 0.013309586969248263
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9135021097044005, 'Bleu_2': 0.8984049324139153, 'Bleu_3': 0.8858904805620758, 'Bleu_4': 0.8731768867744941, 'METEOR': 0.6732479409195641, 'ROUGE_L': 0.9454421638710908, 'CIDEr': 0.0}
nochange_acc: 0.9294294294294294
change_metric: {'Bleu_1': 0.6436202965738138, 'Bleu_2': 0.5135220715592143, 'Bleu_3': 0.4025618966666898, 'Bleu_4': 0.30578661412284786, 'METEOR': 0.21693097198402292, 'ROUGE_L': 0.45353008451051613, 'CIDEr': 0.5064984238751504}
change_acc: 0.8605697151424287
.......................................................
Bleu_1 0.7470956210901849
Bleu_2 0.6585241367641428
Bleu_3 0.58094443246241
Bleu_4 0.5088324992794745
METEOR 0.34037223320412974
ROUGE_L 0.6993016110327537
CIDEr 1.2405154549844346

Epochs since last improvement: 6

The best BLEU-4 score: 0.5754

[Decay LR] The new learning rate is 0.000098

03-06 12 : 23 : 08
Epoch: 7/40 step: 0/267 Loss: 1.5414 AVG_Loss: 1.5414 Top-5 Accuracy: 57.8162 Batch_time: 3.5031s
Epoch: 7/40 step: 50/267 Loss: 1.5055 AVG_Loss: 1.4356 Top-5 Accuracy: 59.6186 Batch_time: 1.0763s
Epoch: 7/40 step: 100/267 Loss: 1.3750 AVG_Loss: 1.4227 Top-5 Accuracy: 62.1779 Batch_time: 1.0552s
Epoch: 7/40 step: 150/267 Loss: 1.3450 AVG_Loss: 1.4162 Top-5 Accuracy: 63.5272 Batch_time: 1.0893s
Epoch: 7/40 step: 200/267 Loss: 1.5555 AVG_Loss: 1.4142 Top-5 Accuracy: 59.4783 Batch_time: 1.0481s
Epoch: 7/40 step: 250/267 Loss: 1.3074 AVG_Loss: 1.4118 Top-5 Accuracy: 64.1593 Batch_time: 1.2130s
eval_time: 1168.8288452625275
avg eval_time: 0.8768408441579351
fps: 1.140457822719629
extractor_time: 0.013977401523060666
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9464796794502157, 'Bleu_2': 0.935696089366197, 'Bleu_3': 0.9261990134678345, 'Bleu_4': 0.9157297706255819, 'METEOR': 0.7111832779489657, 'ROUGE_L': 0.967377530285568, 'CIDEr': 0.0}
nochange_acc: 0.9579579579579579
change_metric: {'Bleu_1': 0.6491439685353474, 'Bleu_2': 0.5282362608478195, 'Bleu_3': 0.42280656080090073, 'Bleu_4': 0.327788997604188, 'METEOR': 0.22177514592435632, 'ROUGE_L': 0.4549825465337622, 'CIDEr': 0.5210550318715084}
change_acc: 0.8380809595202399
.......................................................
Bleu_1 0.7616149966751247
Bleu_2 0.677277118059894
Bleu_3 0.6002853284423681
Bleu_4 0.5245186201328634
METEOR 0.3467435562213598
ROUGE_L 0.7109878422417163
CIDEr 1.2790731517192906

Epochs since last improvement: 7

The best BLEU-4 score: 0.5754

03-06 12 : 48 : 04
Epoch: 8/40 step: 0/267 Loss: 1.2096 AVG_Loss: 1.2096 Top-5 Accuracy: 64.8624 Batch_time: 3.5569s
Epoch: 8/40 step: 50/267 Loss: 1.4171 AVG_Loss: 1.3752 Top-5 Accuracy: 60.2094 Batch_time: 1.0231s
Epoch: 8/40 step: 100/267 Loss: 1.4073 AVG_Loss: 1.3790 Top-5 Accuracy: 62.5756 Batch_time: 1.0695s
Epoch: 8/40 step: 150/267 Loss: 1.2688 AVG_Loss: 1.3758 Top-5 Accuracy: 64.0719 Batch_time: 1.0500s
Epoch: 8/40 step: 200/267 Loss: 1.1009 AVG_Loss: 1.3791 Top-5 Accuracy: 68.1391 Batch_time: 1.1268s
Epoch: 8/40 step: 250/267 Loss: 1.2831 AVG_Loss: 1.3790 Top-5 Accuracy: 65.7194 Batch_time: 1.1111s
eval_time: 1070.691936969757
avg eval_time: 0.8032197576667345
fps: 1.2449892952146628
extractor_time: 0.013323748818216517
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.951360741169383, 'Bleu_2': 0.9428228339672978, 'Bleu_3': 0.9362311358196401, 'Bleu_4': 0.9287320844903101, 'METEOR': 0.7219207481467704, 'ROUGE_L': 0.968410090463815, 'CIDEr': 0.0}
nochange_acc: 0.9594594594594594
change_metric: {'Bleu_1': 0.6574069938312495, 'Bleu_2': 0.537449029230831, 'Bleu_3': 0.43256882850207223, 'Bleu_4': 0.3429312847166659, 'METEOR': 0.23099194196774, 'ROUGE_L': 0.4728300145031497, 'CIDEr': 0.5176775609498402}
change_acc: 0.8350824587706147
.......................................................
Bleu_1 0.766548716151082
Bleu_2 0.6842341223848425
Bleu_3 0.6091800530745962
Bleu_4 0.5376117982399563
METEOR 0.3540162454100901
ROUGE_L 0.7204341634827469
CIDEr 1.2854581393215705

Epochs since last improvement: 8

The best BLEU-4 score: 0.5754

03-06 13 : 11 : 20
Epoch: 9/40 step: 0/267 Loss: 1.5143 AVG_Loss: 1.5143 Top-5 Accuracy: 60.4866 Batch_time: 3.9691s
Epoch: 9/40 step: 50/267 Loss: 1.3603 AVG_Loss: 1.3724 Top-5 Accuracy: 62.6289 Batch_time: 1.0253s
Epoch: 9/40 step: 100/267 Loss: 1.1708 AVG_Loss: 1.3489 Top-5 Accuracy: 66.4623 Batch_time: 1.0975s
Epoch: 9/40 step: 150/267 Loss: 1.2950 AVG_Loss: 1.3576 Top-5 Accuracy: 65.8407 Batch_time: 1.0273s
Epoch: 9/40 step: 200/267 Loss: 1.2324 AVG_Loss: 1.3549 Top-5 Accuracy: 63.7947 Batch_time: 0.7156s
Epoch: 9/40 step: 250/267 Loss: 1.2642 AVG_Loss: 1.3565 Top-5 Accuracy: 64.8166 Batch_time: 1.0709s
eval_time: 1166.3101654052734
avg eval_time: 0.8749513618944287
fps: 1.1429206737101572
extractor_time: 0.013997827359633792
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9701110162251722, 'Bleu_2': 0.9644912857126247, 'Bleu_3': 0.9602613149919388, 'Bleu_4': 0.9558086765468866, 'METEOR': 0.7551763977324684, 'ROUGE_L': 0.9787978378520454, 'CIDEr': 0.0}
nochange_acc: 0.972972972972973
change_metric: {'Bleu_1': 0.6638790866862757, 'Bleu_2': 0.5380157385679746, 'Bleu_3': 0.428279135561949, 'Bleu_4': 0.3350316462988189, 'METEOR': 0.22580400869253345, 'ROUGE_L': 0.4639699781931928, 'CIDEr': 0.5204510795089898}
change_acc: 0.8140929535232384
.......................................................
Bleu_1 0.778833588415406
Bleu_2 0.6948353937070055
Bleu_3 0.6187104148958211
Bleu_4 0.547239184022357
METEOR 0.35669733937123466
ROUGE_L 0.721190799298066
CIDEr 1.295962278715301

Epochs since last improvement: 9

The best BLEU-4 score: 0.5754

[Decay LR] The new learning rate is 0.000069

03-06 13 : 36 : 07
Epoch: 10/40 step: 0/267 Loss: 1.3514 AVG_Loss: 1.3514 Top-5 Accuracy: 61.7486 Batch_time: 3.0054s
Epoch: 10/40 step: 50/267 Loss: 1.3904 AVG_Loss: 1.3462 Top-5 Accuracy: 62.1343 Batch_time: 1.0792s
Epoch: 10/40 step: 100/267 Loss: 1.4034 AVG_Loss: 1.3483 Top-5 Accuracy: 60.3448 Batch_time: 1.0694s
Epoch: 10/40 step: 150/267 Loss: 1.5350 AVG_Loss: 1.3451 Top-5 Accuracy: 59.0721 Batch_time: 1.0675s
Epoch: 10/40 step: 200/267 Loss: 1.3282 AVG_Loss: 1.3369 Top-5 Accuracy: 63.8352 Batch_time: 1.0922s
Epoch: 10/40 step: 250/267 Loss: 1.3787 AVG_Loss: 1.3364 Top-5 Accuracy: 61.6194 Batch_time: 1.1958s
eval_time: 1079.4011797904968
avg eval_time: 0.809753323173666
fps: 1.2349439902027204
extractor_time: 0.013361210404529844
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9241908006811687, 'Bleu_2': 0.9098126881400154, 'Bleu_3': 0.898174371571546, 'Bleu_4': 0.8856412198085187, 'METEOR': 0.6794342254744946, 'ROUGE_L': 0.9499837366941096, 'CIDEr': 0.0}
nochange_acc: 0.9354354354354354
change_metric: {'Bleu_1': 0.6626797902503371, 'Bleu_2': 0.537173156257234, 'Bleu_3': 0.43214772123388073, 'Bleu_4': 0.33882485538212853, 'METEOR': 0.23189352348304731, 'ROUGE_L': 0.4678463650067336, 'CIDEr': 0.5243186602770553}
change_acc: 0.8605697151424287
.......................................................
Bleu_1 0.7617169727813458
Bleu_2 0.6730936518236079
Bleu_3 0.5955256660377064
Bleu_4 0.5202439121579907
METEOR 0.34882785355391466
ROUGE_L 0.7087342041243573
CIDEr 1.2610908166092571

Epochs since last improvement: 10

The best BLEU-4 score: 0.5754

03-06 13 : 59 : 24
Epoch: 11/40 step: 0/267 Loss: 1.5008 AVG_Loss: 1.5008 Top-5 Accuracy: 57.3568 Batch_time: 2.7705s
Epoch: 11/40 step: 50/267 Loss: 1.3859 AVG_Loss: 1.3334 Top-5 Accuracy: 62.9116 Batch_time: 1.0712s
Epoch: 11/40 step: 100/267 Loss: 1.2730 AVG_Loss: 1.3281 Top-5 Accuracy: 64.5760 Batch_time: 1.0768s
Epoch: 11/40 step: 150/267 Loss: 1.2110 AVG_Loss: 1.3205 Top-5 Accuracy: 63.0709 Batch_time: 1.0751s
Epoch: 11/40 step: 200/267 Loss: 1.4104 AVG_Loss: 1.3192 Top-5 Accuracy: 61.5451 Batch_time: 1.0773s
Epoch: 11/40 step: 250/267 Loss: 1.3203 AVG_Loss: 1.3188 Top-5 Accuracy: 61.7316 Batch_time: 1.0930s
eval_time: 1170.104130268097
avg eval_time: 0.8777975470878446
fps: 1.1392148489336416
extractor_time: 0.014104846895680066
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9505320678743312, 'Bleu_2': 0.9414266197437029, 'Bleu_3': 0.9336926915044205, 'Bleu_4': 0.9249618832574328, 'METEOR': 0.7219346470470119, 'ROUGE_L': 0.9693240164111785, 'CIDEr': 0.0}
nochange_acc: 0.960960960960961
change_metric: {'Bleu_1': 0.6800907178377615, 'Bleu_2': 0.5537762682378828, 'Bleu_3': 0.44899728280484325, 'Bleu_4': 0.3559941729917583, 'METEOR': 0.23729120513755037, 'ROUGE_L': 0.46936414875927474, 'CIDEr': 0.5716734583216696}
change_acc: 0.8380809595202399
.......................................................
Bleu_1 0.779214153228885
Bleu_2 0.690941226785546
Bleu_3 0.6134441312449531
Bleu_4 0.5375744139082699
METEOR 0.356521678695246
ROUGE_L 0.7191565507518988
CIDEr 1.3085450901604263

Epochs since last improvement: 11

The best BLEU-4 score: 0.5754

03-06 14 : 24 : 07
Epoch: 12/40 step: 0/267 Loss: 1.1575 AVG_Loss: 1.1575 Top-5 Accuracy: 66.3919 Batch_time: 3.2084s
Epoch: 12/40 step: 50/267 Loss: 1.3152 AVG_Loss: 1.3250 Top-5 Accuracy: 61.5385 Batch_time: 1.0594s
Epoch: 12/40 step: 100/267 Loss: 1.3807 AVG_Loss: 1.3087 Top-5 Accuracy: 60.2243 Batch_time: 1.0874s
Epoch: 12/40 step: 150/267 Loss: 1.3103 AVG_Loss: 1.3063 Top-5 Accuracy: 63.1803 Batch_time: 1.0918s
Epoch: 12/40 step: 200/267 Loss: 1.3805 AVG_Loss: 1.3053 Top-5 Accuracy: 64.1476 Batch_time: 1.0880s
Epoch: 12/40 step: 250/267 Loss: 1.4778 AVG_Loss: 1.3014 Top-5 Accuracy: 59.5451 Batch_time: 1.0879s
eval_time: 1085.627501487732
avg eval_time: 0.8144242321738424
fps: 1.2278613043362216
extractor_time: 0.01335556544670912
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9780428300349748, 'Bleu_2': 0.9742937531057775, 'Bleu_3': 0.9715409143307268, 'Bleu_4': 0.9687381996594998, 'METEOR': 0.7735594534925164, 'ROUGE_L': 0.9845582464221573, 'CIDEr': 0.0}
nochange_acc: 0.9804804804804805
change_metric: {'Bleu_1': 0.699286176159966, 'Bleu_2': 0.5628597247156365, 'Bleu_3': 0.4487592982393824, 'Bleu_4': 0.35370584572037495, 'METEOR': 0.23852178684147227, 'ROUGE_L': 0.47743680761731505, 'CIDEr': 0.5423530866698286}
change_acc: 0.8200899550224887
.......................................................
Bleu_1 0.8040339020596371
Bleu_2 0.715056139783973
Bleu_3 0.6376582627553661
Bleu_4 0.567067786579089
METEOR 0.36774009503713223
ROUGE_L 0.7308073089256608
CIDEr 1.3189221682280912

Epochs since last improvement: 12

The best BLEU-4 score: 0.5754

[Decay LR] The new learning rate is 0.000048

03-06 14 : 47 : 32
Epoch: 13/40 step: 0/267 Loss: 1.3619 AVG_Loss: 1.3619 Top-5 Accuracy: 63.9662 Batch_time: 3.3678s
Epoch: 13/40 step: 50/267 Loss: 1.2177 AVG_Loss: 1.3147 Top-5 Accuracy: 63.8889 Batch_time: 0.8442s
Epoch: 13/40 step: 100/267 Loss: 1.3106 AVG_Loss: 1.2969 Top-5 Accuracy: 61.4385 Batch_time: 0.7134s
Epoch: 13/40 step: 150/267 Loss: 1.3669 AVG_Loss: 1.2859 Top-5 Accuracy: 62.4088 Batch_time: 1.0421s
Epoch: 13/40 step: 200/267 Loss: 1.3705 AVG_Loss: 1.2839 Top-5 Accuracy: 63.5088 Batch_time: 1.0782s
Epoch: 13/40 step: 250/267 Loss: 1.3113 AVG_Loss: 1.2845 Top-5 Accuracy: 64.5963 Batch_time: 1.0831s
eval_time: 1169.077342748642
avg eval_time: 0.8770272638774509
fps: 1.1402154085596732
extractor_time: 0.01431588776023962
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9300199942870808, 'Bleu_2': 0.9180189861981839, 'Bleu_3': 0.9088320460968466, 'Bleu_4': 0.8986876493438409, 'METEOR': 0.6914136852940798, 'ROUGE_L': 0.9522900095901322, 'CIDEr': 0.0}
nochange_acc: 0.93993993993994
change_metric: {'Bleu_1': 0.706896512629427, 'Bleu_2': 0.5837456113611149, 'Bleu_3': 0.477202825976287, 'Bleu_4': 0.38363005727106925, 'METEOR': 0.24837838497703144, 'ROUGE_L': 0.498151241536974, 'CIDEr': 0.5855960172044776}
change_acc: 0.8590704647676162
.......................................................
Bleu_1 0.7920715493009268
Bleu_2 0.7071145265332186
Bleu_3 0.6308139618508578
Bleu_4 0.5572004066156027
METEOR 0.36441527330742746
ROUGE_L 0.7250502809393771
CIDEr 1.3040498582871511

Epochs since last improvement: 13

The best BLEU-4 score: 0.5754

03-06 15 : 11 : 46
Epoch: 14/40 step: 0/267 Loss: 1.3856 AVG_Loss: 1.3856 Top-5 Accuracy: 63.4418 Batch_time: 2.9184s
Epoch: 14/40 step: 50/267 Loss: 1.3744 AVG_Loss: 1.2792 Top-5 Accuracy: 62.1982 Batch_time: 1.0546s
Epoch: 14/40 step: 100/267 Loss: 1.1831 AVG_Loss: 1.2689 Top-5 Accuracy: 64.6182 Batch_time: 0.9791s
Epoch: 14/40 step: 150/267 Loss: 1.2393 AVG_Loss: 1.2739 Top-5 Accuracy: 65.4639 Batch_time: 1.0439s
Epoch: 14/40 step: 200/267 Loss: 1.2288 AVG_Loss: 1.2722 Top-5 Accuracy: 64.8697 Batch_time: 1.0862s
Epoch: 14/40 step: 250/267 Loss: 1.2018 AVG_Loss: 1.2735 Top-5 Accuracy: 65.6584 Batch_time: 1.0569s
eval_time: 1103.3475596904755
avg eval_time: 0.8277175991676485
fps: 1.2081415219461304
extractor_time: 0.013627526640266023
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9623690399826892, 'Bleu_2': 0.9563589719735315, 'Bleu_3': 0.9531573579057552, 'Bleu_4': 0.9509559617479695, 'METEOR': 0.7368771392006923, 'ROUGE_L': 0.9657840141818881, 'CIDEr': 0.0}
nochange_acc: 0.9564564564564565
change_metric: {'Bleu_1': 0.658150028326297, 'Bleu_2': 0.5239854556945719, 'Bleu_3': 0.4156995757983757, 'Bleu_4': 0.3249535203285979, 'METEOR': 0.2225537270629628, 'ROUGE_L': 0.45892406527784363, 'CIDEr': 0.5280552541950404}
change_acc: 0.8380809595202399
.......................................................
Bleu_1 0.789001738493843
Bleu_2 0.7108699210709917
Bleu_3 0.6483296775443785
Bleu_4 0.5951758747800081
METEOR 0.3757274386293083
ROUGE_L 0.7121639197190242
CIDEr 1.2851787373984507

The best BLEU-4 score: 0.5952

03-06 15 : 35 : 30
Epoch: 15/40 step: 0/267 Loss: 1.3434 AVG_Loss: 1.3434 Top-5 Accuracy: 61.1378 Batch_time: 3.0189s
Epoch: 15/40 step: 50/267 Loss: 1.3218 AVG_Loss: 1.2641 Top-5 Accuracy: 63.6661 Batch_time: 0.7140s
Epoch: 15/40 step: 100/267 Loss: 1.1297 AVG_Loss: 1.2617 Top-5 Accuracy: 66.4835 Batch_time: 0.7102s
Epoch: 15/40 step: 150/267 Loss: 1.2386 AVG_Loss: 1.2645 Top-5 Accuracy: 66.0262 Batch_time: 0.7613s
Epoch: 15/40 step: 200/267 Loss: 1.3721 AVG_Loss: 1.2568 Top-5 Accuracy: 61.2792 Batch_time: 1.0066s
Epoch: 15/40 step: 250/267 Loss: 1.4846 AVG_Loss: 1.2602 Top-5 Accuracy: 60.9261 Batch_time: 1.1136s
eval_time: 1163.327538728714
avg eval_time: 0.8727138325046616
fps: 1.1458509797306995
extractor_time: 0.013953215183392558
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9701230228468161, 'Bleu_2': 0.9635692136946522, 'Bleu_3': 0.958393997131695, 'Bleu_4': 0.9524983819337642, 'METEOR': 0.7499280838779933, 'ROUGE_L': 0.9812469344547032, 'CIDEr': 0.0}
nochange_acc: 0.975975975975976
change_metric: {'Bleu_1': 0.6875529444707804, 'Bleu_2': 0.567478901158873, 'Bleu_3': 0.4641274245913259, 'Bleu_4': 0.37397667179218247, 'METEOR': 0.23579250136435997, 'ROUGE_L': 0.47514359705262477, 'CIDEr': 0.548137784391706}
change_acc: 0.8020989505247377
.......................................................
Bleu_1 0.7940023862754235
Bleu_2 0.7130105506444891
Bleu_3 0.6398969280875303
Bleu_4 0.5689851088599449
METEOR 0.3620744461502533
ROUGE_L 0.7280054295430856
CIDEr 1.3157488119087468

Epochs since last improvement: 1

The best BLEU-4 score: 0.5952

03-06 15 : 59 : 13
Epoch: 16/40 step: 0/267 Loss: 1.2633 AVG_Loss: 1.2633 Top-5 Accuracy: 63.7681 Batch_time: 3.7348s
Epoch: 16/40 step: 50/267 Loss: 1.1418 AVG_Loss: 1.2464 Top-5 Accuracy: 66.4835 Batch_time: 0.9904s
Epoch: 16/40 step: 100/267 Loss: 1.2283 AVG_Loss: 1.2459 Top-5 Accuracy: 63.6130 Batch_time: 1.0923s
Epoch: 16/40 step: 150/267 Loss: 1.1853 AVG_Loss: 1.2467 Top-5 Accuracy: 66.6667 Batch_time: 1.0443s
Epoch: 16/40 step: 200/267 Loss: 1.3211 AVG_Loss: 1.2460 Top-5 Accuracy: 65.5477 Batch_time: 1.0856s
Epoch: 16/40 step: 250/267 Loss: 1.1939 AVG_Loss: 1.2488 Top-5 Accuracy: 66.3082 Batch_time: 1.0946s
eval_time: 1135.8329319953918
avg eval_time: 0.8520877209267755
fps: 1.1735880889262753
extractor_time: 0.013732114950696598
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9150546677878006, 'Bleu_2': 0.8982555847604795, 'Bleu_3': 0.8862232061936138, 'Bleu_4': 0.8734439910685938, 'METEOR': 0.6652141826433587, 'ROUGE_L': 0.9414997499670564, 'CIDEr': 0.0}
nochange_acc: 0.924924924924925
change_metric: {'Bleu_1': 0.7242429102105502, 'Bleu_2': 0.597145330536907, 'Bleu_3': 0.4826161053567679, 'Bleu_4': 0.38418865966311555, 'METEOR': 0.2521143093381142, 'ROUGE_L': 0.5006773178344169, 'CIDEr': 0.6031249236805437}
change_acc: 0.8710644677661169
.......................................................
Bleu_1 0.7987549749973671
Bleu_2 0.709005761076704
Bleu_3 0.6264831931422654
Bleu_4 0.5488823134887597
METEOR 0.3613140836140679
ROUGE_L 0.7209231841512497
CIDEr 1.2993213760390745

Epochs since last improvement: 2

The best BLEU-4 score: 0.5952

03-06 16 : 23 : 27
Epoch: 17/40 step: 0/267 Loss: 1.1997 AVG_Loss: 1.1997 Top-5 Accuracy: 65.3356 Batch_time: 3.0178s
Epoch: 17/40 step: 50/267 Loss: 1.2914 AVG_Loss: 1.2323 Top-5 Accuracy: 63.3120 Batch_time: 0.7252s
Epoch: 17/40 step: 100/267 Loss: 1.2543 AVG_Loss: 1.2412 Top-5 Accuracy: 65.1920 Batch_time: 0.7157s
Epoch: 17/40 step: 150/267 Loss: 1.1981 AVG_Loss: 1.2339 Top-5 Accuracy: 66.6373 Batch_time: 0.8145s
Epoch: 17/40 step: 200/267 Loss: 1.2168 AVG_Loss: 1.2372 Top-5 Accuracy: 63.6802 Batch_time: 0.7129s
Epoch: 17/40 step: 250/267 Loss: 1.1269 AVG_Loss: 1.2388 Top-5 Accuracy: 67.9853 Batch_time: 0.7289s
eval_time: 1163.042585849762
avg eval_time: 0.8725000644034223
fps: 1.1461317205560972
extractor_time: 0.0138591974191172
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9614373356701836, 'Bleu_2': 0.9531164657326017, 'Bleu_3': 0.9463789250814706, 'Bleu_4': 0.9388168450217267, 'METEOR': 0.7372500314926644, 'ROUGE_L': 0.9752211351968653, 'CIDEr': 0.0}
nochange_acc: 0.9684684684684685
change_metric: {'Bleu_1': 0.685252946409553, 'Bleu_2': 0.5547993266253757, 'Bleu_3': 0.4464133248850584, 'Bleu_4': 0.35500982397711556, 'METEOR': 0.2366100241295963, 'ROUGE_L': 0.47141872473279883, 'CIDEr': 0.5429280066064701}
change_acc: 0.8320839580209896
.......................................................
Bleu_1 0.7873011567711901
Bleu_2 0.697462504414182
Bleu_3 0.6186934968637938
Bleu_4 0.5441824923752381
METEOR 0.3578930569293262
ROUGE_L 0.7231309568176212
CIDEr 1.304449988415121

Epochs since last improvement: 3

The best BLEU-4 score: 0.5952

[Decay LR] The new learning rate is 0.000034

03-06 16 : 46 : 47
Epoch: 18/40 step: 0/267 Loss: 1.0012 AVG_Loss: 1.0012 Top-5 Accuracy: 68.9952 Batch_time: 3.3240s
Epoch: 18/40 step: 50/267 Loss: 1.2677 AVG_Loss: 1.2215 Top-5 Accuracy: 62.9347 Batch_time: 1.0445s
Epoch: 18/40 step: 100/267 Loss: 1.2558 AVG_Loss: 1.2249 Top-5 Accuracy: 63.1399 Batch_time: 1.0827s
Epoch: 18/40 step: 150/267 Loss: 1.1758 AVG_Loss: 1.2240 Top-5 Accuracy: 66.1934 Batch_time: 1.0585s
Epoch: 18/40 step: 200/267 Loss: 1.2579 AVG_Loss: 1.2290 Top-5 Accuracy: 63.6364 Batch_time: 1.0454s
Epoch: 18/40 step: 250/267 Loss: 1.2273 AVG_Loss: 1.2275 Top-5 Accuracy: 63.1485 Batch_time: 1.0855s
eval_time: 1114.1798105239868
avg eval_time: 0.8358438188477021
fps: 1.196395758933295
extractor_time: 0.013479710161820087
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9607558139532091, 'Bleu_2': 0.9536755382426079, 'Bleu_3': 0.9480529484669222, 'Bleu_4': 0.9415557503344132, 'METEOR': 0.7433974799957, 'ROUGE_L': 0.9749178399646465, 'CIDEr': 0.0}
nochange_acc: 0.9684684684684685
change_metric: {'Bleu_1': 0.6855568495586264, 'Bleu_2': 0.557364887442422, 'Bleu_3': 0.4464908329755081, 'Bleu_4': 0.35016174378303583, 'METEOR': 0.2373191600383368, 'ROUGE_L': 0.46556427941969813, 'CIDEr': 0.5357024251111729}
change_acc: 0.8275862068965517
.......................................................
Bleu_1 0.7861741161879425
Bleu_2 0.6981507214830075
Bleu_3 0.6183425564327478
Bleu_4 0.541440606657672
METEOR 0.35971609193897147
ROUGE_L 0.7200500043431307
CIDEr 1.2963541717237654

Epochs since last improvement: 4

The best BLEU-4 score: 0.5952

03-06 17 : 10 : 39
Epoch: 19/40 step: 0/267 Loss: 1.1990 AVG_Loss: 1.1990 Top-5 Accuracy: 63.4699 Batch_time: 2.4542s
Epoch: 19/40 step: 50/267 Loss: 1.2928 AVG_Loss: 1.2306 Top-5 Accuracy: 62.3388 Batch_time: 0.7134s
Epoch: 19/40 step: 100/267 Loss: 1.2876 AVG_Loss: 1.2316 Top-5 Accuracy: 64.3172 Batch_time: 0.7100s
Epoch: 19/40 step: 150/267 Loss: 1.1432 AVG_Loss: 1.2207 Top-5 Accuracy: 66.4799 Batch_time: 0.7650s
Epoch: 19/40 step: 200/267 Loss: 1.2075 AVG_Loss: 1.2206 Top-5 Accuracy: 66.9929 Batch_time: 0.7107s
Epoch: 19/40 step: 250/267 Loss: 1.1582 AVG_Loss: 1.2203 Top-5 Accuracy: 66.8153 Batch_time: 0.7519s
eval_time: 1327.54358792305
avg eval_time: 0.9959066676091898
fps: 1.0041101566280672
extractor_time: 0.014737261566826032
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9506314580938718, 'Bleu_2': 0.9417397230318418, 'Bleu_3': 0.9355617546666853, 'Bleu_4': 0.928716335518781, 'METEOR': 0.7172390286602349, 'ROUGE_L': 0.965779051378461, 'CIDEr': 0.0}
nochange_acc: 0.9564564564564565
change_metric: {'Bleu_1': 0.6917596853016037, 'Bleu_2': 0.5695793907601647, 'Bleu_3': 0.464120281834676, 'Bleu_4': 0.3737934974397939, 'METEOR': 0.24104671743646533, 'ROUGE_L': 0.47949369764603267, 'CIDEr': 0.5771631403786737}
change_acc: 0.8470764617691154
.......................................................
Bleu_1 0.7882383944928201
Bleu_2 0.7046705707489465
Bleu_3 0.6295826727688439
Bleu_4 0.5581036135320055
METEOR 0.36234917865233474
ROUGE_L 0.7224539719039451
CIDEr 1.311188134988817

Epochs since last improvement: 5

The best BLEU-4 score: 0.5952

03-06 17 : 36 : 48
Epoch: 20/40 step: 0/267 Loss: 1.2921 AVG_Loss: 1.2921 Top-5 Accuracy: 62.6374 Batch_time: 3.0322s
Epoch: 20/40 step: 50/267 Loss: 1.1951 AVG_Loss: 1.1903 Top-5 Accuracy: 64.4406 Batch_time: 1.0388s
Epoch: 20/40 step: 100/267 Loss: 1.1278 AVG_Loss: 1.1984 Top-5 Accuracy: 66.6073 Batch_time: 1.1620s
Epoch: 20/40 step: 150/267 Loss: 1.3702 AVG_Loss: 1.2028 Top-5 Accuracy: 60.9325 Batch_time: 1.0419s
Epoch: 20/40 step: 200/267 Loss: 1.3300 AVG_Loss: 1.2081 Top-5 Accuracy: 62.8571 Batch_time: 1.1087s
Epoch: 20/40 step: 250/267 Loss: 1.2136 AVG_Loss: 1.2115 Top-5 Accuracy: 66.0228 Batch_time: 1.0482s
eval_time: 2218.1545855998993
avg eval_time: 1.664031947186721
fps: 0.6009500008041552
extractor_time: 0.01951772560325674
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9665085124194902, 'Bleu_2': 0.9599201795602987, 'Bleu_3': 0.9547419445177795, 'Bleu_4': 0.9491302036985926, 'METEOR': 0.745364081136892, 'ROUGE_L': 0.9782181147082488, 'CIDEr': 0.0}
nochange_acc: 0.9714714714714715
change_metric: {'Bleu_1': 0.7048892744638348, 'Bleu_2': 0.5822277760782931, 'Bleu_3': 0.4741427393265048, 'Bleu_4': 0.38252713894833, 'METEOR': 0.2445404278531116, 'ROUGE_L': 0.48891950494475567, 'CIDEr': 0.6058133465670821}
change_acc: 0.8290854572713643
.......................................................
Bleu_1 0.8041640950893468
Bleu_2 0.7219933351978479
Bleu_3 0.6467660009474329
Bleu_4 0.5763656018267631
METEOR 0.3687785177555799
ROUGE_L 0.7333852769646254
CIDEr 1.3407203163910182

Epochs since last improvement: 6

The best BLEU-4 score: 0.5952

[Decay LR] The new learning rate is 0.000024

03-06 18 : 19 : 09
Epoch: 21/40 step: 0/267 Loss: 1.1493 AVG_Loss: 1.1493 Top-5 Accuracy: 66.2478 Batch_time: 3.3614s
Epoch: 21/40 step: 50/267 Loss: 1.2201 AVG_Loss: 1.1791 Top-5 Accuracy: 66.0090 Batch_time: 1.0636s
Epoch: 21/40 step: 100/267 Loss: 1.1271 AVG_Loss: 1.1824 Top-5 Accuracy: 66.2350 Batch_time: 1.1363s
Epoch: 21/40 step: 150/267 Loss: 1.1298 AVG_Loss: 1.1900 Top-5 Accuracy: 65.7801 Batch_time: 1.0773s
Epoch: 21/40 step: 200/267 Loss: 1.0507 AVG_Loss: 1.1986 Top-5 Accuracy: 67.0300 Batch_time: 0.9884s
Epoch: 21/40 step: 250/267 Loss: 1.2107 AVG_Loss: 1.2015 Top-5 Accuracy: 64.1674 Batch_time: 1.1365s
eval_time: 2230.3996596336365
avg eval_time: 1.6732180492375368
fps: 0.5976507368275681
extractor_time: 0.02009271287596145
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9570323689484511, 'Bleu_2': 0.94892431632411, 'Bleu_3': 0.941497560785981, 'Bleu_4': 0.9330703085168909, 'METEOR': 0.7344508541266338, 'ROUGE_L': 0.9752485215576049, 'CIDEr': 0.0}
nochange_acc: 0.9684684684684685
change_metric: {'Bleu_1': 0.6853972757091193, 'Bleu_2': 0.5536440315671016, 'Bleu_3': 0.445165076199253, 'Bleu_4': 0.35330326289603814, 'METEOR': 0.23931315903976438, 'ROUGE_L': 0.4735295574286669, 'CIDEr': 0.5736025754213465}
change_acc: 0.8455772113943029
.......................................................
Bleu_1 0.7875831157341732
Bleu_2 0.6973255457776428
Bleu_3 0.6186506850981278
Bleu_4 0.5443611704381364
METEOR 0.36246813361943686
ROUGE_L 0.7242008478336727
CIDEr 1.3171283324403291

Epochs since last improvement: 7

The best BLEU-4 score: 0.5952

03-06 19 : 01 : 41
Epoch: 22/40 step: 0/267 Loss: 1.2033 AVG_Loss: 1.2033 Top-5 Accuracy: 65.1910 Batch_time: 3.1832s
Epoch: 22/40 step: 50/267 Loss: 1.1968 AVG_Loss: 1.1828 Top-5 Accuracy: 64.7687 Batch_time: 0.8723s
Epoch: 22/40 step: 100/267 Loss: 1.1708 AVG_Loss: 1.1874 Top-5 Accuracy: 65.3881 Batch_time: 0.7412s
Epoch: 22/40 step: 150/267 Loss: 1.3057 AVG_Loss: 1.1911 Top-5 Accuracy: 62.9446 Batch_time: 0.7619s
Epoch: 22/40 step: 200/267 Loss: 1.0793 AVG_Loss: 1.1940 Top-5 Accuracy: 69.2035 Batch_time: 0.7352s
Epoch: 22/40 step: 250/267 Loss: 1.3139 AVG_Loss: 1.1974 Top-5 Accuracy: 62.6412 Batch_time: 0.8107s
eval_time: 2324.6151514053345
avg eval_time: 1.743897337888473
fps: 0.573428250776969
extractor_time: 0.020374441003763667
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9704797047967725, 'Bleu_2': 0.9650989293799423, 'Bleu_3': 0.9607385075020625, 'Bleu_4': 0.9559277027106658, 'METEOR': 0.7575011520641118, 'ROUGE_L': 0.9802261525433785, 'CIDEr': 0.0}
nochange_acc: 0.9744744744744744
change_metric: {'Bleu_1': 0.7047479471084318, 'Bleu_2': 0.5781642324677145, 'Bleu_3': 0.47095549163077705, 'Bleu_4': 0.3806237673885782, 'METEOR': 0.24512767905281493, 'ROUGE_L': 0.4887985073457388, 'CIDEr': 0.5760265771926372}
change_acc: 0.8215892053973014
.......................................................
Bleu_1 0.8041457236119709
Bleu_2 0.7195665381157734
Bleu_3 0.6443594149872081
Bleu_4 0.5738454941154346
METEOR 0.37011416191402047
ROUGE_L 0.7343279984947472
CIDEr 1.3310609716600965

Epochs since last improvement: 8

The best BLEU-4 score: 0.5952

03-06 19 : 44 : 33
Epoch: 23/40 step: 0/267 Loss: 1.1457 AVG_Loss: 1.1457 Top-5 Accuracy: 66.1376 Batch_time: 3.0843s
Epoch: 23/40 step: 50/267 Loss: 1.1864 AVG_Loss: 1.1778 Top-5 Accuracy: 65.9649 Batch_time: 1.0307s
Epoch: 23/40 step: 100/267 Loss: 1.1325 AVG_Loss: 1.1766 Top-5 Accuracy: 67.3225 Batch_time: 1.0978s
Epoch: 23/40 step: 150/267 Loss: 1.2615 AVG_Loss: 1.1804 Top-5 Accuracy: 63.3721 Batch_time: 1.0817s
Epoch: 23/40 step: 200/267 Loss: 1.2942 AVG_Loss: 1.1881 Top-5 Accuracy: 63.8565 Batch_time: 1.0574s
Epoch: 23/40 step: 250/267 Loss: 1.1421 AVG_Loss: 1.1892 Top-5 Accuracy: 66.9369 Batch_time: 1.0677s
eval_time: 2082.1468443870544
avg eval_time: 1.562000633448653
fps: 0.6402046059303808
extractor_time: 0.019065309447269196
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9504662004659236, 'Bleu_2': 0.9401079666764269, 'Bleu_3': 0.9330347668079861, 'Bleu_4': 0.9256298099035087, 'METEOR': 0.7119136831144839, 'ROUGE_L': 0.9630792691361174, 'CIDEr': 0.0}
nochange_acc: 0.9519519519519519
change_metric: {'Bleu_1': 0.6893094693292973, 'Bleu_2': 0.5630079770850803, 'Bleu_3': 0.45383231616801056, 'Bleu_4': 0.3587913508832658, 'METEOR': 0.2368010200261784, 'ROUGE_L': 0.4701405771596538, 'CIDEr': 0.5560273431264701}
change_acc: 0.8260869565217391
.......................................................
Bleu_1 0.7856034164795279
Bleu_2 0.6979731734628369
Bleu_3 0.6191852295951283
Bleu_4 0.5433327146806309
METEOR 0.3561465233259572
ROUGE_L 0.7164250249138359
CIDEr 1.2950323576561185

Epochs since last improvement: 9

The best BLEU-4 score: 0.5952

[Decay LR] The new learning rate is 0.000016

03-06 20 : 24 : 37
Epoch: 24/40 step: 0/267 Loss: 1.3052 AVG_Loss: 1.3052 Top-5 Accuracy: 62.8478 Batch_time: 4.1706s
Epoch: 24/40 step: 50/267 Loss: 1.0925 AVG_Loss: 1.1646 Top-5 Accuracy: 67.1362 Batch_time: 0.9859s
Epoch: 24/40 step: 100/267 Loss: 1.0474 AVG_Loss: 1.1756 Top-5 Accuracy: 67.6923 Batch_time: 1.0812s
Epoch: 24/40 step: 150/267 Loss: 1.1148 AVG_Loss: 1.1784 Top-5 Accuracy: 67.0132 Batch_time: 1.0889s
Epoch: 24/40 step: 200/267 Loss: 1.1274 AVG_Loss: 1.1787 Top-5 Accuracy: 66.2963 Batch_time: 1.0632s
Epoch: 24/40 step: 250/267 Loss: 1.2975 AVG_Loss: 1.1804 Top-5 Accuracy: 63.7789 Batch_time: 1.0525s
eval_time: 2290.930376768112
avg eval_time: 1.718627439435943
fps: 0.5818596730471159
extractor_time: 0.019600871802270638
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9645888206025858, 'Bleu_2': 0.9571842012597536, 'Bleu_3': 0.9519228708548785, 'Bleu_4': 0.9460946239300296, 'METEOR': 0.7374986337236392, 'ROUGE_L': 0.97538877935539, 'CIDEr': 0.0}
nochange_acc: 0.9684684684684685
change_metric: {'Bleu_1': 0.7151755207610652, 'Bleu_2': 0.591149765898043, 'Bleu_3': 0.4846549820930561, 'Bleu_4': 0.3912719304194955, 'METEOR': 0.25266004369523415, 'ROUGE_L': 0.49576905646516156, 'CIDEr': 0.6079942896570532}
change_acc: 0.8260869565217391
.......................................................
Bleu_1 0.8064861206776697
Bleu_2 0.7212768864314626
Bleu_3 0.6445213583773817
Bleu_4 0.5697963853374762
METEOR 0.3719497631938027
ROUGE_L 0.735399015538599
CIDEr 1.3417596307096586

Epochs since last improvement: 10

The best BLEU-4 score: 0.5952

03-06 21 : 08 : 17
Epoch: 25/40 step: 0/267 Loss: 1.0598 AVG_Loss: 1.0598 Top-5 Accuracy: 67.7820 Batch_time: 4.3439s
Epoch: 25/40 step: 50/267 Loss: 1.1116 AVG_Loss: 1.1631 Top-5 Accuracy: 67.4699 Batch_time: 1.1024s
Epoch: 25/40 step: 100/267 Loss: 1.2198 AVG_Loss: 1.1735 Top-5 Accuracy: 63.1957 Batch_time: 1.0796s
Epoch: 25/40 step: 150/267 Loss: 1.1154 AVG_Loss: 1.1736 Top-5 Accuracy: 66.9326 Batch_time: 1.1756s
Epoch: 25/40 step: 200/267 Loss: 1.3178 AVG_Loss: 1.1775 Top-5 Accuracy: 64.4406 Batch_time: 1.1586s
Epoch: 25/40 step: 250/267 Loss: 1.2510 AVG_Loss: 1.1756 Top-5 Accuracy: 64.7255 Batch_time: 0.8406s
eval_time: 2865.4166135787964
avg eval_time: 2.1495998601491344
fps: 0.46520285869883815
extractor_time: 0.02300455564855188
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9647194910350015, 'Bleu_2': 0.9574397288733754, 'Bleu_3': 0.9510445101309791, 'Bleu_4': 0.9440280339187572, 'METEOR': 0.7433991963023647, 'ROUGE_L': 0.9781222906682886, 'CIDEr': 0.0}
nochange_acc: 0.9714714714714715
change_metric: {'Bleu_1': 0.693985907881374, 'Bleu_2': 0.5651148227057639, 'Bleu_3': 0.4558320500920878, 'Bleu_4': 0.36387302330245014, 'METEOR': 0.23645943212623277, 'ROUGE_L': 0.4741481706925187, 'CIDEr': 0.5473579515143976}
change_acc: 0.8140929535232384
.......................................................
Bleu_1 0.7968215581602555
Bleu_2 0.7099978709065928
Bleu_3 0.6325749083201793
Bleu_4 0.5597995098690305
METEOR 0.36167016442915423
ROUGE_L 0.7259461931260242
CIDEr 1.3095581846313333

Epochs since last improvement: 11

The best BLEU-4 score: 0.5952

03-06 22 : 01 : 25
Epoch: 26/40 step: 0/267 Loss: 0.9500 AVG_Loss: 0.9500 Top-5 Accuracy: 70.8934 Batch_time: 3.6933s
Epoch: 26/40 step: 50/267 Loss: 1.1777 AVG_Loss: 1.1596 Top-5 Accuracy: 65.6331 Batch_time: 0.7390s
Epoch: 26/40 step: 100/267 Loss: 1.0861 AVG_Loss: 1.1611 Top-5 Accuracy: 66.2921 Batch_time: 0.8153s
Epoch: 26/40 step: 150/267 Loss: 1.0545 AVG_Loss: 1.1658 Top-5 Accuracy: 68.2261 Batch_time: 0.7251s
Epoch: 26/40 step: 200/267 Loss: 1.2062 AVG_Loss: 1.1672 Top-5 Accuracy: 64.4794 Batch_time: 0.7420s
Epoch: 26/40 step: 250/267 Loss: 1.3055 AVG_Loss: 1.1714 Top-5 Accuracy: 63.1003 Batch_time: 0.8028s
eval_time: 3064.4547815322876
avg eval_time: 2.2989158151029914
fps: 0.4349876552374755
extractor_time: 0.023869241705892323
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9646792879341721, 'Bleu_2': 0.9581266479884006, 'Bleu_3': 0.9524159661801612, 'Bleu_4': 0.9462444606183078, 'METEOR': 0.7459447222107807, 'ROUGE_L': 0.9781412961390742, 'CIDEr': 0.0}
nochange_acc: 0.9714714714714715
change_metric: {'Bleu_1': 0.700564290879747, 'Bleu_2': 0.569108697757279, 'Bleu_3': 0.45794843172887834, 'Bleu_4': 0.3642900577633627, 'METEOR': 0.24082249403945857, 'ROUGE_L': 0.47464621487294556, 'CIDEr': 0.5546447012724093}
change_acc: 0.815592203898051
.......................................................
Bleu_1 0.8001101743716161
Bleu_2 0.7118216790622768
Bleu_3 0.6335033357072118
Bleu_4 0.5603838151857418
METEOR 0.36469089059506204
ROUGE_L 0.7262048976360675
CIDEr 1.3159439195604223

Epochs since last improvement: 12

The best BLEU-4 score: 0.5952

[Decay LR] The new learning rate is 0.000012

03-06 22 : 56 : 33
Epoch: 27/40 step: 0/267 Loss: 1.1025 AVG_Loss: 1.1025 Top-5 Accuracy: 67.5000 Batch_time: 2.8731s
Epoch: 27/40 step: 50/267 Loss: 1.1371 AVG_Loss: 1.1489 Top-5 Accuracy: 66.8930 Batch_time: 0.8970s
Epoch: 27/40 step: 100/267 Loss: 1.2683 AVG_Loss: 1.1533 Top-5 Accuracy: 63.5828 Batch_time: 0.7339s
Epoch: 27/40 step: 150/267 Loss: 1.3393 AVG_Loss: 1.1609 Top-5 Accuracy: 63.9901 Batch_time: 0.7304s
Epoch: 27/40 step: 200/267 Loss: 1.1780 AVG_Loss: 1.1615 Top-5 Accuracy: 65.1121 Batch_time: 0.8003s
Epoch: 27/40 step: 250/267 Loss: 1.1454 AVG_Loss: 1.1617 Top-5 Accuracy: 66.9442 Batch_time: 0.8047s
eval_time: 3171.548308134079
avg eval_time: 2.379256045111837
fps: 0.42029944698658733
extractor_time: 0.024878031314984356
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.947339398952041, 'Bleu_2': 0.9383347731718374, 'Bleu_3': 0.9315349545421572, 'Bleu_4': 0.9241626860478296, 'METEOR': 0.7198906381515723, 'ROUGE_L': 0.9653296105289871, 'CIDEr': 0.0}
nochange_acc: 0.9564564564564565
change_metric: {'Bleu_1': 0.6993674890574371, 'Bleu_2': 0.564222459618036, 'Bleu_3': 0.4508260466432898, 'Bleu_4': 0.35512767236683274, 'METEOR': 0.24062402743616704, 'ROUGE_L': 0.4810211778009142, 'CIDEr': 0.5619519479516588}
change_acc: 0.856071964017991
.......................................................
Bleu_1 0.7929354445797002
Bleu_2 0.7020330996104015
Bleu_3 0.6227863789312619
Bleu_4 0.5493641072745399
METEOR 0.36312483340251905
ROUGE_L 0.7229937330874083
CIDEr 1.3071449375720432

Epochs since last improvement: 13

The best BLEU-4 score: 0.5952

03-06 23 : 53 : 32
Epoch: 28/40 step: 0/267 Loss: 1.2264 AVG_Loss: 1.2264 Top-5 Accuracy: 65.3915 Batch_time: 3.6548s
Epoch: 28/40 step: 50/267 Loss: 1.2159 AVG_Loss: 1.1493 Top-5 Accuracy: 64.0381 Batch_time: 1.0772s
Epoch: 28/40 step: 100/267 Loss: 1.2574 AVG_Loss: 1.1495 Top-5 Accuracy: 67.0618 Batch_time: 0.7390s
Epoch: 28/40 step: 150/267 Loss: 1.1847 AVG_Loss: 1.1556 Top-5 Accuracy: 65.9649 Batch_time: 0.7306s
Epoch: 28/40 step: 200/267 Loss: 1.2001 AVG_Loss: 1.1578 Top-5 Accuracy: 65.2668 Batch_time: 0.7283s
Epoch: 28/40 step: 250/267 Loss: 1.0879 AVG_Loss: 1.1593 Top-5 Accuracy: 67.4054 Batch_time: 0.7545s
eval_time: 2990.6349081993103
avg eval_time: 2.2435370654158366
fps: 0.4457247510705384
extractor_time: 0.024044664837950973
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.963698093496745, 'Bleu_2': 0.9576441317594736, 'Bleu_3': 0.952916870408227, 'Bleu_4': 0.9481306875402533, 'METEOR': 0.7477941473468552, 'ROUGE_L': 0.9753860809204303, 'CIDEr': 0.0}
nochange_acc: 0.9684684684684685
change_metric: {'Bleu_1': 0.6862165978253277, 'Bleu_2': 0.550502384344542, 'Bleu_3': 0.43880889180747074, 'Bleu_4': 0.3483845278049977, 'METEOR': 0.2354565029665326, 'ROUGE_L': 0.4664113393724119, 'CIDEr': 0.5322374283188342}
change_acc: 0.8050974512743628
.......................................................
Bleu_1 0.7954713123145924
Bleu_2 0.7079472844781478
Bleu_3 0.6334077626781925
Bleu_4 0.5674049690118946
METEOR 0.3676679618495098
ROUGE_L 0.7207077968900264
CIDEr 1.299390210979416

Epochs since last improvement: 14

The best BLEU-4 score: 0.5952

03-07 00 : 47 : 52
Epoch: 29/40 step: 0/267 Loss: 1.1488 AVG_Loss: 1.1488 Top-5 Accuracy: 65.6420 Batch_time: 3.7438s
Epoch: 29/40 step: 50/267 Loss: 1.2429 AVG_Loss: 1.1279 Top-5 Accuracy: 64.4771 Batch_time: 0.9557s
Epoch: 29/40 step: 100/267 Loss: 1.1119 AVG_Loss: 1.1388 Top-5 Accuracy: 66.4200 Batch_time: 1.2114s
Epoch: 29/40 step: 150/267 Loss: 1.3162 AVG_Loss: 1.1494 Top-5 Accuracy: 62.1044 Batch_time: 1.0570s
Epoch: 29/40 step: 200/267 Loss: 1.0560 AVG_Loss: 1.1510 Top-5 Accuracy: 67.2727 Batch_time: 0.7452s
Epoch: 29/40 step: 250/267 Loss: 1.2308 AVG_Loss: 1.1537 Top-5 Accuracy: 64.4934 Batch_time: 1.1003s
eval_time: 2741.6408269405365
avg eval_time: 2.056744806407004
fps: 0.4862051903011406
extractor_time: 0.023050949256937035
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9514986376019205, 'Bleu_2': 0.9429210421840788, 'Bleu_3': 0.9359478153670157, 'Bleu_4': 0.9282418113479799, 'METEOR': 0.7291016689676884, 'ROUGE_L': 0.9698219575447234, 'CIDEr': 0.0}
nochange_acc: 0.960960960960961
change_metric: {'Bleu_1': 0.7095228316878919, 'Bleu_2': 0.5778289562481456, 'Bleu_3': 0.46516445755103797, 'Bleu_4': 0.3703849781022972, 'METEOR': 0.24628816742165285, 'ROUGE_L': 0.4887178303445705, 'CIDEr': 0.5801186773722681}
change_acc: 0.8305847076461769
.......................................................
Bleu_1 0.8046251392508234
Bleu_2 0.7165651882861429
Bleu_3 0.638451453175515
Bleu_4 0.566371736058489
METEOR 0.3692017038679146
ROUGE_L 0.7290894347821564
CIDEr 1.3197131294147497

Epochs since last improvement: 15

The best BLEU-4 score: 0.5952

[Decay LR] The new learning rate is 0.000008

03-07 01 : 38 : 58
Epoch: 30/40 step: 0/267 Loss: 1.2210 AVG_Loss: 1.2210 Top-5 Accuracy: 65.2406 Batch_time: 3.4199s
Epoch: 30/40 step: 50/267 Loss: 1.2407 AVG_Loss: 1.1407 Top-5 Accuracy: 63.5530 Batch_time: 1.1094s
Epoch: 30/40 step: 100/267 Loss: 1.1773 AVG_Loss: 1.1358 Top-5 Accuracy: 66.3591 Batch_time: 1.0415s
Epoch: 30/40 step: 150/267 Loss: 1.1699 AVG_Loss: 1.1414 Top-5 Accuracy: 65.0943 Batch_time: 0.7865s
Epoch: 30/40 step: 200/267 Loss: 1.1235 AVG_Loss: 1.1468 Top-5 Accuracy: 67.1329 Batch_time: 1.0469s
Epoch: 30/40 step: 250/267 Loss: 1.1177 AVG_Loss: 1.1503 Top-5 Accuracy: 68.1186 Batch_time: 1.0872s
eval_time: 2712.02992272377
avg eval_time: 2.0345310748115306
fps: 0.4915137509475668
extractor_time: 0.022497027061616937
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9588299024916146, 'Bleu_2': 0.9511289030527312, 'Bleu_3': 0.9450642974591031, 'Bleu_4': 0.9386626328060669, 'METEOR': 0.735602468254887, 'ROUGE_L': 0.974018356864583, 'CIDEr': 0.0}
nochange_acc: 0.9669669669669669
change_metric: {'Bleu_1': 0.7049522055693285, 'Bleu_2': 0.5670027001525827, 'Bleu_3': 0.450557820848164, 'Bleu_4': 0.35553901118432385, 'METEOR': 0.24023051144289811, 'ROUGE_L': 0.4727526553009607, 'CIDEr': 0.5366800050217869}
change_acc: 0.8185907046476761
.......................................................
Bleu_1 0.8023799126042396
Bleu_2 0.7102794281912409
Bleu_3 0.6296927887009862
Bleu_4 0.5572608519969482
METEOR 0.3646777161106535
ROUGE_L 0.7231974844392746
CIDEr 1.3028676416335796

Epochs since last improvement: 16

The best BLEU-4 score: 0.5952

03-07 02 : 29 : 38
Epoch: 31/40 step: 0/267 Loss: 1.1696 AVG_Loss: 1.1696 Top-5 Accuracy: 64.8241 Batch_time: 2.8762s
Epoch: 31/40 step: 50/267 Loss: 1.1266 AVG_Loss: 1.1330 Top-5 Accuracy: 66.2953 Batch_time: 1.0730s
Epoch: 31/40 step: 100/267 Loss: 1.0546 AVG_Loss: 1.1341 Top-5 Accuracy: 70.0661 Batch_time: 0.9834s
Epoch: 31/40 step: 150/267 Loss: 1.1034 AVG_Loss: 1.1361 Top-5 Accuracy: 66.9573 Batch_time: 1.0863s
Epoch: 31/40 step: 200/267 Loss: 1.2206 AVG_Loss: 1.1432 Top-5 Accuracy: 64.8390 Batch_time: 0.7384s
Epoch: 31/40 step: 250/267 Loss: 1.1637 AVG_Loss: 1.1455 Top-5 Accuracy: 66.1398 Batch_time: 1.0967s
eval_time: 2957.9417576789856
avg eval_time: 2.219011071026996
fps: 0.45065119911149565
extractor_time: 0.023819243022578155
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9481586402263603, 'Bleu_2': 0.9380618736592611, 'Bleu_3': 0.930284498201475, 'Bleu_4': 0.9218151256058333, 'METEOR': 0.7151156915247742, 'ROUGE_L': 0.9669743757838475, 'CIDEr': 0.0}
nochange_acc: 0.9564564564564565
change_metric: {'Bleu_1': 0.688224578769168, 'Bleu_2': 0.5570229473054389, 'Bleu_3': 0.4464599742143127, 'Bleu_4': 0.35276417596227694, 'METEOR': 0.23839212476359808, 'ROUGE_L': 0.47270693965063076, 'CIDEr': 0.5507698242221464}
change_acc: 0.8275862068965517
.......................................................
Bleu_1 0.7873374869078789
Bleu_2 0.6976259288994993
Bleu_3 0.6187557419408616
Bleu_4 0.5447695935277843
METEOR 0.3592627977626835
ROUGE_L 0.7196552610795297
CIDEr 1.29650190144478

Epochs since last improvement: 17

The best BLEU-4 score: 0.5952

03-07 03 : 24 : 21
Epoch: 32/40 step: 0/267 Loss: 1.2386 AVG_Loss: 1.2386 Top-5 Accuracy: 65.2641 Batch_time: 3.1191s
Epoch: 32/40 step: 50/267 Loss: 1.1533 AVG_Loss: 1.1347 Top-5 Accuracy: 65.9574 Batch_time: 1.0124s
Epoch: 32/40 step: 100/267 Loss: 1.1088 AVG_Loss: 1.1432 Top-5 Accuracy: 66.8971 Batch_time: 1.0471s
Epoch: 32/40 step: 150/267 Loss: 0.9800 AVG_Loss: 1.1421 Top-5 Accuracy: 71.0983 Batch_time: 1.0578s
Epoch: 32/40 step: 200/267 Loss: 1.2172 AVG_Loss: 1.1420 Top-5 Accuracy: 66.9894 Batch_time: 1.0382s
Epoch: 32/40 step: 250/267 Loss: 1.0239 AVG_Loss: 1.1403 Top-5 Accuracy: 67.7362 Batch_time: 1.1107s
eval_time: 2808.4016366004944
avg eval_time: 2.106827934433979
fps: 0.4746472095115163
extractor_time: 0.023110854384242728
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9761731096520845, 'Bleu_2': 0.9722678933308481, 'Bleu_3': 0.969276919179077, 'Bleu_4': 0.9664250329107448, 'METEOR': 0.7679633811719386, 'ROUGE_L': 0.9833192231274995, 'CIDEr': 0.0}
nochange_acc: 0.9774774774774775
change_metric: {'Bleu_1': 0.6870894865856068, 'Bleu_2': 0.5507516140327566, 'Bleu_3': 0.4388093352678549, 'Bleu_4': 0.3444188430410468, 'METEOR': 0.2336681934963362, 'ROUGE_L': 0.4707250618113645, 'CIDEr': 0.5235043307403846}
change_acc: 0.7916041979010495
.......................................................
Bleu_1 0.8053588175890689
Bleu_2 0.722028664034246
Bleu_3 0.6521698589875256
Bleu_4 0.5901379482530159
METEOR 0.3761649266036449
ROUGE_L 0.7268298715912189
CIDEr 1.3084103292366471

Epochs since last improvement: 18

The best BLEU-4 score: 0.5952

[Decay LR] The new learning rate is 0.000006

03-07 04 : 16 : 38
Epoch: 33/40 step: 0/267 Loss: 1.1229 AVG_Loss: 1.1229 Top-5 Accuracy: 66.5484 Batch_time: 3.6821s
Epoch: 33/40 step: 50/267 Loss: 1.0702 AVG_Loss: 1.1242 Top-5 Accuracy: 68.4400 Batch_time: 1.0853s
Epoch: 33/40 step: 100/267 Loss: 1.1209 AVG_Loss: 1.1251 Top-5 Accuracy: 67.2367 Batch_time: 1.0451s
Epoch: 33/40 step: 150/267 Loss: 1.2411 AVG_Loss: 1.1328 Top-5 Accuracy: 62.5325 Batch_time: 0.9732s
Epoch: 33/40 step: 200/267 Loss: 1.0328 AVG_Loss: 1.1348 Top-5 Accuracy: 69.1960 Batch_time: 1.0706s
Epoch: 33/40 step: 250/267 Loss: 1.0527 AVG_Loss: 1.1386 Top-5 Accuracy: 68.4677 Batch_time: 1.1366s
eval_time: 3008.394294500351
avg eval_time: 2.256859935859228
fps: 0.44309351418358256
extractor_time: 0.023570510380624025
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9616319913534284, 'Bleu_2': 0.9554252901390751, 'Bleu_3': 0.9501332612709519, 'Bleu_4': 0.9444935599276146, 'METEOR': 0.7472341459757094, 'ROUGE_L': 0.9761624452760814, 'CIDEr': 0.0}
nochange_acc: 0.96996996996997
change_metric: {'Bleu_1': 0.7066749313843602, 'Bleu_2': 0.5781672537708678, 'Bleu_3': 0.4693224725802135, 'Bleu_4': 0.376170293151587, 'METEOR': 0.24138305142006003, 'ROUGE_L': 0.48339466429609906, 'CIDEr': 0.578418711044469}
change_acc: 0.8275862068965517
.......................................................
Bleu_1 0.807288208108867
Bleu_2 0.7235066492891022
Bleu_3 0.6496703561615463
Bleu_4 0.5808439394220818
METEOR 0.3702222252720775
ROUGE_L 0.7295937206596912
CIDEr 1.3258490435434178

Epochs since last improvement: 19

The best BLEU-4 score: 0.5952

03-07 05 : 12 : 10
Epoch: 34/40 step: 0/267 Loss: 1.1474 AVG_Loss: 1.1474 Top-5 Accuracy: 66.9484 Batch_time: 3.0423s
Epoch: 34/40 step: 50/267 Loss: 1.0143 AVG_Loss: 1.1329 Top-5 Accuracy: 68.9189 Batch_time: 0.7325s
Epoch: 34/40 step: 100/267 Loss: 1.1447 AVG_Loss: 1.1321 Top-5 Accuracy: 66.0885 Batch_time: 0.7543s
Epoch: 34/40 step: 150/267 Loss: 1.1758 AVG_Loss: 1.1334 Top-5 Accuracy: 66.9787 Batch_time: 0.8716s
Epoch: 34/40 step: 200/267 Loss: 1.2021 AVG_Loss: 1.1340 Top-5 Accuracy: 66.0069 Batch_time: 0.7298s
Epoch: 34/40 step: 250/267 Loss: 1.1043 AVG_Loss: 1.1359 Top-5 Accuracy: 67.5918 Batch_time: 1.0085s
eval_time: 3223.954262971878
avg eval_time: 2.418570339813862
fps: 0.413467404085077
extractor_time: 0.02515565338478174
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9623406720738812, 'Bleu_2': 0.9547585574081164, 'Bleu_3': 0.9479350418990086, 'Bleu_4': 0.9402249882894719, 'METEOR': 0.7409690920093172, 'ROUGE_L': 0.9778968201261948, 'CIDEr': 0.0}
nochange_acc: 0.9714714714714715
change_metric: {'Bleu_1': 0.6790551084792331, 'Bleu_2': 0.5502009406992081, 'Bleu_3': 0.43988403703934364, 'Bleu_4': 0.34956370179221413, 'METEOR': 0.232539177421767, 'ROUGE_L': 0.45964477939480797, 'CIDEr': 0.5352926209493543}
change_acc: 0.7991004497751124
.......................................................
Bleu_1 0.7849706614783969
Bleu_2 0.6970473948521093
Bleu_3 0.6178985656521175
Bleu_4 0.5447461245297845
METEOR 0.3586061149470601
ROUGE_L 0.7185764066469488
CIDEr 1.299729015108129

Epochs since last improvement: 20

The best BLEU-4 score: 0.5952

03-07 06 : 10 : 15
Epoch: 35/40 step: 0/267 Loss: 1.1304 AVG_Loss: 1.1304 Top-5 Accuracy: 67.1832 Batch_time: 2.5003s
Epoch: 35/40 step: 50/267 Loss: 1.0071 AVG_Loss: 1.1181 Top-5 Accuracy: 68.9139 Batch_time: 0.7299s
Epoch: 35/40 step: 100/267 Loss: 1.1774 AVG_Loss: 1.1233 Top-5 Accuracy: 67.5258 Batch_time: 0.7370s
Epoch: 35/40 step: 150/267 Loss: 0.9442 AVG_Loss: 1.1220 Top-5 Accuracy: 71.6204 Batch_time: 0.7509s
Epoch: 35/40 step: 200/267 Loss: 1.1376 AVG_Loss: 1.1242 Top-5 Accuracy: 67.5699 Batch_time: 0.7682s
Epoch: 35/40 step: 250/267 Loss: 1.0820 AVG_Loss: 1.1301 Top-5 Accuracy: 67.8010 Batch_time: 0.7694s
eval_time: 3301.0146028995514
avg eval_time: 2.4763800471864603
fps: 0.4038152387539022
extractor_time: 0.025811045847704602
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9495677233426658, 'Bleu_2': 0.9394229659352441, 'Bleu_3': 0.9303734423889418, 'Bleu_4': 0.9200974987676765, 'METEOR': 0.7202239152541831, 'ROUGE_L': 0.9712154095415382, 'CIDEr': 0.0}
nochange_acc: 0.9624624624624625
change_metric: {'Bleu_1': 0.69475048916125, 'Bleu_2': 0.5651828612858211, 'Bleu_3': 0.451640721704936, 'Bleu_4': 0.35412721592615337, 'METEOR': 0.24258724079993227, 'ROUGE_L': 0.47407173111682227, 'CIDEr': 0.5708240731135038}
change_acc: 0.8320839580209896
.......................................................
Bleu_1 0.7906205241699003
Bleu_2 0.7002337805067826
Bleu_3 0.6173819844041009
Bleu_4 0.5385486549643781
METEOR 0.36066419760045804
ROUGE_L 0.7224570948308963
CIDEr 1.3122559656569037

Epochs since last improvement: 21

The best BLEU-4 score: 0.5952

[Decay LR] The new learning rate is 0.000004

03-07 07 : 09 : 21
Epoch: 36/40 step: 0/267 Loss: 1.1109 AVG_Loss: 1.1109 Top-5 Accuracy: 69.6339 Batch_time: 2.7148s
Epoch: 36/40 step: 50/267 Loss: 1.1468 AVG_Loss: 1.1138 Top-5 Accuracy: 66.9675 Batch_time: 0.7180s
Epoch: 36/40 step: 100/267 Loss: 1.2874 AVG_Loss: 1.1210 Top-5 Accuracy: 62.9260 Batch_time: 0.7805s
Epoch: 36/40 step: 150/267 Loss: 1.1379 AVG_Loss: 1.1200 Top-5 Accuracy: 67.1403 Batch_time: 0.7357s
Epoch: 36/40 step: 200/267 Loss: 1.1517 AVG_Loss: 1.1286 Top-5 Accuracy: 65.7989 Batch_time: 0.7284s
Epoch: 36/40 step: 250/267 Loss: 1.0365 AVG_Loss: 1.1303 Top-5 Accuracy: 69.4085 Batch_time: 0.7662s
eval_time: 3122.273145675659
avg eval_time: 2.3422904318647104
fps: 0.4269325385084268
extractor_time: 0.025017238074643934
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9527785775983436, 'Bleu_2': 0.9428498467591366, 'Bleu_3': 0.9347402002870313, 'Bleu_4': 0.9256697390674183, 'METEOR': 0.7197631847219382, 'ROUGE_L': 0.9700928679439399, 'CIDEr': 0.0}
nochange_acc: 0.960960960960961
change_metric: {'Bleu_1': 0.7123009352252534, 'Bleu_2': 0.5876769579854345, 'Bleu_3': 0.47568957900103487, 'Bleu_4': 0.37908439629590945, 'METEOR': 0.24813835737470832, 'ROUGE_L': 0.48938565811689144, 'CIDEr': 0.5917951233183938}
change_acc: 0.8305847076461769
.......................................................
Bleu_1 0.8030268567202432
Bleu_2 0.7166101925039206
Bleu_3 0.6359187031837249
Bleu_4 0.5588935094632688
METEOR 0.3658545657984792
ROUGE_L 0.7295589527491602
CIDEr 1.325443361318001

Epochs since last improvement: 22

The best BLEU-4 score: 0.5952

03-07 08 : 05 : 28
Epoch: 37/40 step: 0/267 Loss: 1.1360 AVG_Loss: 1.1360 Top-5 Accuracy: 66.4253 Batch_time: 3.2382s
Epoch: 37/40 step: 50/267 Loss: 1.1644 AVG_Loss: 1.1065 Top-5 Accuracy: 67.3027 Batch_time: 0.9917s
Epoch: 37/40 step: 100/267 Loss: 1.0842 AVG_Loss: 1.1164 Top-5 Accuracy: 68.2091 Batch_time: 1.0718s
Epoch: 37/40 step: 150/267 Loss: 1.0836 AVG_Loss: 1.1193 Top-5 Accuracy: 68.0997 Batch_time: 0.9505s
Epoch: 37/40 step: 200/267 Loss: 1.1391 AVG_Loss: 1.1215 Top-5 Accuracy: 64.7527 Batch_time: 1.0913s
Epoch: 37/40 step: 250/267 Loss: 1.2218 AVG_Loss: 1.1216 Top-5 Accuracy: 64.6194 Batch_time: 1.0979s
eval_time: 3052.727067232132
avg eval_time: 2.290117829881569
fps: 0.43665875482560385
extractor_time: 0.023958688737392785
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9828877005344966, 'Bleu_2': 0.9798908103605528, 'Bleu_3': 0.9774515941107417, 'Bleu_4': 0.9750430685940868, 'METEOR': 0.788709036150301, 'ROUGE_L': 0.9885398604759369, 'CIDEr': 0.0}
nochange_acc: 0.984984984984985
change_metric: {'Bleu_1': 0.6976386066257712, 'Bleu_2': 0.5640005576359985, 'Bleu_3': 0.45222006667249015, 'Bleu_4': 0.35949309912000943, 'METEOR': 0.23554079848393195, 'ROUGE_L': 0.47233840887219963, 'CIDEr': 0.5393575730736131}
change_acc: 0.7781109445277361
.......................................................
Bleu_1 0.8081461825980422
Bleu_2 0.7230528535865117
Bleu_3 0.6491945361770826
Bleu_4 0.5823872318799831
METEOR 0.37003317280825526
ROUGE_L 0.7302455107237293
CIDEr 1.3224326730755513

Epochs since last improvement: 23

The best BLEU-4 score: 0.5952

03-07 09 : 01 : 47
Epoch: 38/40 step: 0/267 Loss: 1.1993 AVG_Loss: 1.1993 Top-5 Accuracy: 63.4087 Batch_time: 3.1079s
Epoch: 38/40 step: 50/267 Loss: 1.1771 AVG_Loss: 1.1103 Top-5 Accuracy: 64.2036 Batch_time: 1.1409s
Epoch: 38/40 step: 100/267 Loss: 1.2023 AVG_Loss: 1.1216 Top-5 Accuracy: 65.5114 Batch_time: 1.0799s
Epoch: 38/40 step: 150/267 Loss: 1.1132 AVG_Loss: 1.1189 Top-5 Accuracy: 67.7074 Batch_time: 1.0234s
Epoch: 38/40 step: 200/267 Loss: 1.1133 AVG_Loss: 1.1190 Top-5 Accuracy: 66.6964 Batch_time: 1.0650s
Epoch: 38/40 step: 250/267 Loss: 1.0936 AVG_Loss: 1.1203 Top-5 Accuracy: 68.4066 Batch_time: 1.1280s
eval_time: 3061.741698741913
avg eval_time: 2.2968804941799794
fps: 0.4353731082369677
extractor_time: 0.02483063269985053
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.95775862068938, 'Bleu_2': 0.9486296916034686, 'Bleu_3': 0.9413334037394643, 'Bleu_4': 0.933293034931305, 'METEOR': 0.7245086203053683, 'ROUGE_L': 0.9732812811268096, 'CIDEr': 0.0}
nochange_acc: 0.9654654654654654
change_metric: {'Bleu_1': 0.7047327241968291, 'Bleu_2': 0.5806034665311334, 'Bleu_3': 0.4714397292563749, 'Bleu_4': 0.37958759673840087, 'METEOR': 0.24992623498709815, 'ROUGE_L': 0.4920072453903939, 'CIDEr': 0.5869915488763902}
change_acc: 0.8200899550224887
.......................................................
Bleu_1 0.7996366964781544
Bleu_2 0.7142229273049103
Bleu_3 0.63592822093239
Bleu_4 0.5624481651774699
METEOR 0.36955675499059487
ROUGE_L 0.7324637403644771
CIDEr 1.3284005185404613

Epochs since last improvement: 24

The best BLEU-4 score: 0.5952

[Decay LR] The new learning rate is 0.000003

03-07 09 : 58 : 17
Epoch: 39/40 step: 0/267 Loss: 1.2754 AVG_Loss: 1.2754 Top-5 Accuracy: 62.3946 Batch_time: 3.2258s
Epoch: 39/40 step: 50/267 Loss: 1.0578 AVG_Loss: 1.1008 Top-5 Accuracy: 69.0053 Batch_time: 1.0460s
Epoch: 39/40 step: 100/267 Loss: 1.0275 AVG_Loss: 1.1069 Top-5 Accuracy: 69.9454 Batch_time: 1.1037s
Epoch: 39/40 step: 150/267 Loss: 1.1859 AVG_Loss: 1.1123 Top-5 Accuracy: 64.8116 Batch_time: 1.1347s
Epoch: 39/40 step: 200/267 Loss: 1.0486 AVG_Loss: 1.1152 Top-5 Accuracy: 67.8108 Batch_time: 1.1272s
Epoch: 39/40 step: 250/267 Loss: 1.1326 AVG_Loss: 1.1195 Top-5 Accuracy: 65.9758 Batch_time: 1.1504s
eval_time: 2950.7485768795013
avg eval_time: 2.213614836368718
fps: 0.4517497730727315
extractor_time: 0.02337295885651253
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9787114845935634, 'Bleu_2': 0.9747125865612547, 'Bleu_3': 0.9715137220126924, 'Bleu_4': 0.9679606307669353, 'METEOR': 0.7703595509296288, 'ROUGE_L': 0.9860780792425783, 'CIDEr': 0.0}
nochange_acc: 0.9804804804804805
change_metric: {'Bleu_1': 0.7057016872429203, 'Bleu_2': 0.5777240440648889, 'Bleu_3': 0.46879539555370014, 'Bleu_4': 0.37727025207643783, 'METEOR': 0.24117718773606753, 'ROUGE_L': 0.4826623182830585, 'CIDEr': 0.561190816664411}
change_acc: 0.7886056971514243
.......................................................
Bleu_1 0.8097171841395627
Bleu_2 0.7263837190817682
Bleu_3 0.6523388040541583
Bleu_4 0.5835474459647642
METEOR 0.37143909159840216
ROUGE_L 0.7341813706454292
CIDEr 1.3303131915857669

Epochs since last improvement: 25

The best BLEU-4 score: 0.5952

[mrazhou]

参数这些都没问题，应该是库的版本问题，你用了torch2.0+的版本了。 建议conda create -n sen python=3.8; pip install -r requirements.txt

[Dingxiangtao]

好的，感谢答复，我去试试

[Dingxiangtao]

作者您好，我按照你的建议重新配置的环境，可是结果还是没有达到论文中的结果

[mrazhou]

麻烦再发一下训练日志。 一般来讲是不会有问题的，该代码经过多次多方验证可以复现结果。

…

---Original--- From: ***@***.***> Date: Tue, Mar 25, 2025 21:42 PM To: ***@***.***>; Cc: ***@***.******@***.***>; Subject: Re: [mrazhou/SEN] train result (Issue #1) 作者您好，我按照你的建议重新配置的环境，可是结果还是没有达到论文中的结果 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: ***@***.***> Dingxiangtao left a comment (mrazhou/SEN#1) 作者您好，我按照你的建议重新配置的环境，可是结果还是没有达到论文中的结果 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you commented.Message ID: ***@***.***>

[Dingxiangtao]

[Info] Load weight from pretrain_ckpt/rn50.pth.tar
[Info] With message: _IncompatibleKeys(missing_keys=['fc.weight', 'fc.bias'], unexpected_keys=['', 'fc.0.weight', 'fc.0.bias', 'fc.2.weight', 'fc.2.bias'])
git:
sha: N/A, status: clean, branch: N/A
[CMD] python train.py
--more_reproducibility
--savepath model_checkpoints/SEN
--model SEN
--batch_size 128
--proj_channel 512
--encoder_n_layers 2
--ft_layer 4
--model_stage 4
--weight_path pretrain_ckpt/rn50.pth.tar

[args]
Namespace(Split='VAL', batch_size=128, beam_size=1, checkpoint=None, data_folder='./data', data_name='LEVIR_CC_5_cap_per_img_5_min_word_freq', decay_interval=3, decay_lr=False, decay_rate=0.7, decoder='SEN_TransDecoder', decoder_lr=0.0002, decoder_n_layers=1, dropout=0.5, encoder_feat='SEN_DiffEncoder', encoder_feat_clip=False, encoder_feat_decay=False, encoder_image='SEN_Extractor', encoder_lr=0.0002, encoder_n_layers=2, epochs=40, feature_dim_de=1024, ft_layer=4, grad_clip=5.0, model='SEN', model_stage=4, more_reproducibility=True, n_heads=8, n_layers=3, print_freq=50, proj_channel=512, savepath='model_checkpoints/SEN', seed=42, stop_criteria=40, weight_path='pretrain_ckpt/rn50.pth.tar', workers=8)
03-25 18 : 23 : 42
--------Params--------
encoder_image params:24,547,584
encoder_feat params:10,582,016
decoder params:4,710,356
Total trainable params:39,839,956
Total params:39,858,900
--------Params--------
�[32m[Info]�[0m Checkpoint_savepath:model_checkpoints/SEN
Encoder_image_mode:SEN_Extractor Encoder_feat_mode:SEN_DiffEncoder Decoder_mode:SEN_TransDecoder
encoder_layers 3 decoder_layers 1 n_heads 8 dropout 0.5 encoder_lr 0.0002 decoder_lr 0.0002
03-25 18 : 23 : 46
Epoch: 0/40 step: 0/267 Loss: 8.0012 AVG_Loss: 8.0012 Top-5 Accuracy: 0.1764 Batch_time: 7.1568s
Epoch: 0/40 step: 50/267 Loss: 4.6206 AVG_Loss: 5.1309 Top-5 Accuracy: 12.8617 Batch_time: 0.6668s
Epoch: 0/40 step: 100/267 Loss: 3.9888 AVG_Loss: 4.6522 Top-5 Accuracy: 19.0111 Batch_time: 0.6593s
Epoch: 0/40 step: 150/267 Loss: 3.4944 AVG_Loss: 4.2906 Top-5 Accuracy: 31.0518 Batch_time: 0.6521s
Epoch: 0/40 step: 200/267 Loss: 2.6192 AVG_Loss: 3.9925 Top-5 Accuracy: 45.0581 Batch_time: 0.6593s
Epoch: 0/40 step: 250/267 Loss: 2.6036 AVG_Loss: 3.7315 Top-5 Accuracy: 44.2424 Batch_time: 0.6700s
eval_time: 51.77326941490173
avg eval_time: 0.038839661976670464
fps: 25.746877009399892
extractor_time: 0.011780673547159526
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.97898348702531, 'Bleu_2': 0.9754749485786772, 'Bleu_3': 0.9740411345348179, 'Bleu_4': 0.973024570160429, 'METEOR': 0.7554497257788143, 'ROUGE_L': 0.9792271549214834, 'CIDEr': 0.0}
nochange_acc: 0.9654654654654654
change_metric: {'Bleu_1': 0.5608660771781455, 'Bleu_2': 0.4054967771018916, 'Bleu_3': 0.2432777264959849, 'Bleu_4': 0.14824533717332086, 'METEOR': 0.15824483208399892, 'ROUGE_L': 0.3732533301172048, 'CIDEr': 0.1662799968683784}
change_acc: 0.8170914542728636
.......................................................
Bleu_1 0.7578428988332185
Bleu_2 0.6805712138116797
Bleu_3 0.6145317463407975
Bleu_4 0.5754030323754038
METEOR 0.33915889864883836
ROUGE_L 0.6760129455107904
CIDEr 1.1121800905774646

The best BLEU-4 score: 0.5754

03-25 18 : 28 : 16
Epoch: 1/40 step: 0/267 Loss: 2.3670 AVG_Loss: 2.3670 Top-5 Accuracy: 49.0248 Batch_time: 2.2936s
Epoch: 1/40 step: 50/267 Loss: 1.8408 AVG_Loss: 2.3577 Top-5 Accuracy: 55.5449 Batch_time: 0.6595s
Epoch: 1/40 step: 100/267 Loss: 2.0534 AVG_Loss: 2.2960 Top-5 Accuracy: 53.5363 Batch_time: 0.6609s
Epoch: 1/40 step: 150/267 Loss: 2.1565 AVG_Loss: 2.2239 Top-5 Accuracy: 49.4355 Batch_time: 0.6623s
Epoch: 1/40 step: 200/267 Loss: 1.7773 AVG_Loss: 2.1695 Top-5 Accuracy: 57.3171 Batch_time: 0.6600s
Epoch: 1/40 step: 250/267 Loss: 1.8032 AVG_Loss: 2.1224 Top-5 Accuracy: 55.0141 Batch_time: 0.6731s
eval_time: 51.326653718948364
avg eval_time: 0.03850461644332211
fps: 25.97091186382746
extractor_time: 0.011763788038684476
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.928899082568541, 'Bleu_2': 0.9116674591979416, 'Bleu_3': 0.9001504318535332, 'Bleu_4': 0.8879544412427911, 'METEOR': 0.6598685556671879, 'ROUGE_L': 0.9505972285204504, 'CIDEr': 0.0}
nochange_acc: 0.9309309309309309
change_metric: {'Bleu_1': 0.6077230015907715, 'Bleu_2': 0.4848303206410436, 'Bleu_3': 0.36034726847546594, 'Bleu_4': 0.25197163204847745, 'METEOR': 0.19207611203924554, 'ROUGE_L': 0.41206161249701706, 'CIDEr': 0.3373905481981422}
change_acc: 0.8590704647676162
.......................................................
Bleu_1 0.7336725377466506
Bleu_2 0.6493004119994912
Bleu_3 0.5664255888290922
Bleu_4 0.4898595154256354
METEOR 0.3202509895355005
ROUGE_L 0.6811274191523858
CIDEr 1.157146453403504

Epochs since last improvement: 1

The best BLEU-4 score: 0.5754

03-25 18 : 32 : 39
Epoch: 2/40 step: 0/267 Loss: 1.8382 AVG_Loss: 1.8382 Top-5 Accuracy: 55.7018 Batch_time: 2.9702s
Epoch: 2/40 step: 50/267 Loss: 1.6723 AVG_Loss: 1.8829 Top-5 Accuracy: 58.1132 Batch_time: 0.6647s
Epoch: 2/40 step: 100/267 Loss: 1.6791 AVG_Loss: 1.8554 Top-5 Accuracy: 58.9883 Batch_time: 0.6644s
Epoch: 2/40 step: 150/267 Loss: 1.7844 AVG_Loss: 1.8258 Top-5 Accuracy: 55.6818 Batch_time: 0.6540s
Epoch: 2/40 step: 200/267 Loss: 1.7021 AVG_Loss: 1.8096 Top-5 Accuracy: 58.4229 Batch_time: 0.6554s
Epoch: 2/40 step: 250/267 Loss: 1.5847 AVG_Loss: 1.7960 Top-5 Accuracy: 61.3825 Batch_time: 0.6722s
eval_time: 62.4789457321167
avg eval_time: 0.046870927030845236
fps: 21.335187147928846
extractor_time: 0.011714831385859313
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9240223463684569, 'Bleu_2': 0.9067712078474487, 'Bleu_3': 0.8949904530453782, 'Bleu_4': 0.8827513325239496, 'METEOR': 0.6627907111040264, 'ROUGE_L': 0.9482992357838275, 'CIDEr': 0.0}
nochange_acc: 0.9309309309309309
change_metric: {'Bleu_1': 0.6368341048538405, 'Bleu_2': 0.5076299588266221, 'Bleu_3': 0.38727100988316016, 'Bleu_4': 0.29140141649242474, 'METEOR': 0.21156814093707904, 'ROUGE_L': 0.4494758064092313, 'CIDEr': 0.41269244528959215}
change_acc: 0.8575712143928036
.......................................................
Bleu_1 0.7506165565025187
Bleu_2 0.6622489620496043
Bleu_3 0.5808127153524137
Bleu_4 0.5094239291490528
METEOR 0.33538574903814294
ROUGE_L 0.6987004155341233
CIDEr 1.195789152401825

Epochs since last improvement: 2

The best BLEU-4 score: 0.5754

03-25 18 : 37 : 14
Epoch: 3/40 step: 0/267 Loss: 1.7929 AVG_Loss: 1.7929 Top-5 Accuracy: 57.8947 Batch_time: 2.6951s
Epoch: 3/40 step: 50/267 Loss: 1.7408 AVG_Loss: 1.6456 Top-5 Accuracy: 56.4081 Batch_time: 0.6712s
Epoch: 3/40 step: 100/267 Loss: 1.7381 AVG_Loss: 1.6597 Top-5 Accuracy: 56.8807 Batch_time: 0.6599s
Epoch: 3/40 step: 150/267 Loss: 1.6252 AVG_Loss: 1.6470 Top-5 Accuracy: 57.0312 Batch_time: 0.6608s
Epoch: 3/40 step: 200/267 Loss: 1.6225 AVG_Loss: 1.6460 Top-5 Accuracy: 58.0372 Batch_time: 0.6606s
Epoch: 3/40 step: 250/267 Loss: 1.6851 AVG_Loss: 1.6415 Top-5 Accuracy: 56.9295 Batch_time: 0.6800s
eval_time: 59.223421573638916
avg eval_time: 0.04442867334856633
fps: 22.507986951455283
extractor_time: 0.011721771280298712
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9272989889261127, 'Bleu_2': 0.9129361987280233, 'Bleu_3': 0.9046752863810968, 'Bleu_4': 0.8980537888252855, 'METEOR': 0.6672301398148054, 'ROUGE_L': 0.9400586876049545, 'CIDEr': 0.0}
nochange_acc: 0.9204204204204204
change_metric: {'Bleu_1': 0.6370276278182538, 'Bleu_2': 0.5080941450330193, 'Bleu_3': 0.3959925161193523, 'Bleu_4': 0.29624818449206436, 'METEOR': 0.21373695569860757, 'ROUGE_L': 0.44801568720999674, 'CIDEr': 0.43900940072142536}
change_acc: 0.8695652173913043
.......................................................
Bleu_1 0.7523918867201729
Bleu_2 0.6685407154939303
Bleu_3 0.5966964716215859
Bleu_4 0.531669968943917
METEOR 0.34765010940032715
ROUGE_L 0.6938526251417612
CIDEr 1.2055146572279416

Epochs since last improvement: 3

The best BLEU-4 score: 0.5754

[Decay LR] The new learning rate is 0.000140

03-25 18 : 41 : 45
Epoch: 4/40 step: 0/267 Loss: 1.6531 AVG_Loss: 1.6531 Top-5 Accuracy: 58.1104 Batch_time: 2.7437s
Epoch: 4/40 step: 50/267 Loss: 1.4457 AVG_Loss: 1.5772 Top-5 Accuracy: 62.3056 Batch_time: 0.6615s
Epoch: 4/40 step: 100/267 Loss: 1.6003 AVG_Loss: 1.5709 Top-5 Accuracy: 60.4478 Batch_time: 0.6754s
Epoch: 4/40 step: 150/267 Loss: 1.6750 AVG_Loss: 1.5703 Top-5 Accuracy: 55.8442 Batch_time: 0.6599s
Epoch: 4/40 step: 200/267 Loss: 1.5395 AVG_Loss: 1.5550 Top-5 Accuracy: 61.4630 Batch_time: 0.6564s
Epoch: 4/40 step: 250/267 Loss: 1.4902 AVG_Loss: 1.5483 Top-5 Accuracy: 61.5517 Batch_time: 0.6745s
eval_time: 61.49014067649841
avg eval_time: 0.04612913779182177
fps: 21.678272082884884
extractor_time: 0.011691331684544672
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9450675870000158, 'Bleu_2': 0.9342290238655554, 'Bleu_3': 0.9259315759971984, 'Bleu_4': 0.9167609369072411, 'METEOR': 0.7053480009118464, 'ROUGE_L': 0.9640842820891987, 'CIDEr': 0.0}
nochange_acc: 0.9534534534534534
change_metric: {'Bleu_1': 0.6493555870713245, 'Bleu_2': 0.5178218874139312, 'Bleu_3': 0.4054311488525256, 'Bleu_4': 0.30117111577097055, 'METEOR': 0.21525679730941474, 'ROUGE_L': 0.4469557708432035, 'CIDEr': 0.47253433979217635}
change_acc: 0.8455772113943029
.......................................................
Bleu_1 0.7611376702305912
Bleu_2 0.6713974731907257
Bleu_3 0.5918764707270889
Bleu_4 0.5129877438251411
METEOR 0.34071354803823284
ROUGE_L 0.7053260547815626
CIDEr 1.247473579247893

Epochs since last improvement: 4

The best BLEU-4 score: 0.5754

03-25 18 : 46 : 19
Epoch: 5/40 step: 0/267 Loss: 1.5933 AVG_Loss: 1.5933 Top-5 Accuracy: 58.4251 Batch_time: 2.1194s
Epoch: 5/40 step: 50/267 Loss: 1.7635 AVG_Loss: 1.5210 Top-5 Accuracy: 55.8347 Batch_time: 0.6603s
Epoch: 5/40 step: 100/267 Loss: 1.6150 AVG_Loss: 1.5020 Top-5 Accuracy: 55.5169 Batch_time: 0.6594s
Epoch: 5/40 step: 150/267 Loss: 1.5999 AVG_Loss: 1.5022 Top-5 Accuracy: 58.4775 Batch_time: 0.6616s
Epoch: 5/40 step: 200/267 Loss: 1.6449 AVG_Loss: 1.4964 Top-5 Accuracy: 55.7420 Batch_time: 0.6602s
Epoch: 5/40 step: 250/267 Loss: 1.4333 AVG_Loss: 1.4925 Top-5 Accuracy: 61.7864 Batch_time: 0.6746s
eval_time: 56.73727512359619
avg eval_time: 0.042563597242007645
fps: 23.494254827997988
extractor_time: 0.01174788160245399
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.939834616481055, 'Bleu_2': 0.9290244444054118, 'Bleu_3': 0.9204196008673176, 'Bleu_4': 0.9108421929796708, 'METEOR': 0.7045095061965451, 'ROUGE_L': 0.9613141544076321, 'CIDEr': 0.0}
nochange_acc: 0.9504504504504504
change_metric: {'Bleu_1': 0.6708300758887558, 'Bleu_2': 0.5482390185754903, 'Bleu_3': 0.43360478911100425, 'Bleu_4': 0.3312601883911612, 'METEOR': 0.22427937734642844, 'ROUGE_L': 0.46436442133121075, 'CIDEr': 0.49358515298210764}
change_acc: 0.8395802098950524
.......................................................
Bleu_1 0.7748267027923023
Bleu_2 0.6913501742531027
Bleu_3 0.6116329700802425
Bleu_4 0.5345770151556782
METEOR 0.3481228169707617
ROUGE_L 0.7126528851188301
CIDEr 1.2624049515376152

Epochs since last improvement: 5

The best BLEU-4 score: 0.5754

03-25 18 : 50 : 48
Epoch: 6/40 step: 0/267 Loss: 1.4014 AVG_Loss: 1.4014 Top-5 Accuracy: 61.3924 Batch_time: 2.3261s
Epoch: 6/40 step: 50/267 Loss: 1.3549 AVG_Loss: 1.4546 Top-5 Accuracy: 63.6771 Batch_time: 0.6627s
Epoch: 6/40 step: 100/267 Loss: 1.4733 AVG_Loss: 1.4650 Top-5 Accuracy: 59.2405 Batch_time: 0.6761s
Epoch: 6/40 step: 150/267 Loss: 1.5532 AVG_Loss: 1.4575 Top-5 Accuracy: 58.2245 Batch_time: 0.6616s
Epoch: 6/40 step: 200/267 Loss: 1.5254 AVG_Loss: 1.4523 Top-5 Accuracy: 59.0476 Batch_time: 0.6776s
Epoch: 6/40 step: 250/267 Loss: 1.4939 AVG_Loss: 1.4485 Top-5 Accuracy: 59.1979 Batch_time: 0.6972s
eval_time: 73.44490671157837
avg eval_time: 0.0550974543972831
fps: 18.14965883522399
extractor_time: 0.011974895438661454
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9135021097044005, 'Bleu_2': 0.8984049324139153, 'Bleu_3': 0.8858904805620758, 'Bleu_4': 0.8731768867744941, 'METEOR': 0.6732479409195641, 'ROUGE_L': 0.9454421638710908, 'CIDEr': 0.0}
nochange_acc: 0.9294294294294294
change_metric: {'Bleu_1': 0.6436202965738138, 'Bleu_2': 0.5135220715592143, 'Bleu_3': 0.4025618966666898, 'Bleu_4': 0.30578661412284786, 'METEOR': 0.21693097198402292, 'ROUGE_L': 0.45353008451051613, 'CIDEr': 0.5064984238751504}
change_acc: 0.8605697151424287
.......................................................
Bleu_1 0.7470956210901849
Bleu_2 0.6585241367641428
Bleu_3 0.58094443246241
Bleu_4 0.5088324992794745
METEOR 0.34037223320412974
ROUGE_L 0.6993016110327537
CIDEr 1.2405154549844346

Epochs since last improvement: 6

The best BLEU-4 score: 0.5754

[Decay LR] The new learning rate is 0.000098

03-25 18 : 55 : 36
Epoch: 7/40 step: 0/267 Loss: 1.5414 AVG_Loss: 1.5414 Top-5 Accuracy: 57.8162 Batch_time: 3.0269s
Epoch: 7/40 step: 50/267 Loss: 1.5055 AVG_Loss: 1.4356 Top-5 Accuracy: 59.6186 Batch_time: 0.6637s
Epoch: 7/40 step: 100/267 Loss: 1.3750 AVG_Loss: 1.4227 Top-5 Accuracy: 62.1779 Batch_time: 0.6663s
Epoch: 7/40 step: 150/267 Loss: 1.3450 AVG_Loss: 1.4162 Top-5 Accuracy: 63.5272 Batch_time: 0.6610s
Epoch: 7/40 step: 200/267 Loss: 1.5555 AVG_Loss: 1.4142 Top-5 Accuracy: 59.4783 Batch_time: 0.6526s
Epoch: 7/40 step: 250/267 Loss: 1.3074 AVG_Loss: 1.4118 Top-5 Accuracy: 64.1593 Batch_time: 0.6735s
eval_time: 58.24077844619751
avg eval_time: 0.04369150671132596
fps: 22.887743528899044
extractor_time: 0.011747234313003299
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9464796794502157, 'Bleu_2': 0.935696089366197, 'Bleu_3': 0.9261990134678345, 'Bleu_4': 0.9157297706255819, 'METEOR': 0.7111832779489657, 'ROUGE_L': 0.967377530285568, 'CIDEr': 0.0}
nochange_acc: 0.9579579579579579
change_metric: {'Bleu_1': 0.6491439685353474, 'Bleu_2': 0.5282362608478195, 'Bleu_3': 0.42280656080090073, 'Bleu_4': 0.327788997604188, 'METEOR': 0.22177514592435632, 'ROUGE_L': 0.4549825465337622, 'CIDEr': 0.5210550318715084}
change_acc: 0.8380809595202399
.......................................................
Bleu_1 0.7616149966751247
Bleu_2 0.677277118059894
Bleu_3 0.6002853284423681
Bleu_4 0.5245186201328634
METEOR 0.3467435562213598
ROUGE_L 0.7109878422417163
CIDEr 1.2790731517192906

Epochs since last improvement: 7

The best BLEU-4 score: 0.5754

03-25 19 : 00 : 07
Epoch: 8/40 step: 0/267 Loss: 1.2096 AVG_Loss: 1.2096 Top-5 Accuracy: 64.8624 Batch_time: 2.0381s
Epoch: 8/40 step: 50/267 Loss: 1.4171 AVG_Loss: 1.3752 Top-5 Accuracy: 60.2094 Batch_time: 0.6582s
Epoch: 8/40 step: 100/267 Loss: 1.4073 AVG_Loss: 1.3790 Top-5 Accuracy: 62.5756 Batch_time: 0.6533s
Epoch: 8/40 step: 150/267 Loss: 1.2688 AVG_Loss: 1.3758 Top-5 Accuracy: 64.0719 Batch_time: 0.6687s
Epoch: 8/40 step: 200/267 Loss: 1.1009 AVG_Loss: 1.3791 Top-5 Accuracy: 68.1391 Batch_time: 0.6615s
Epoch: 8/40 step: 250/267 Loss: 1.2831 AVG_Loss: 1.3790 Top-5 Accuracy: 65.7194 Batch_time: 0.6811s
eval_time: 57.6157169342041
avg eval_time: 0.043222593348990326
fps: 23.136048129406372
extractor_time: 0.01171158742654261
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.951360741169383, 'Bleu_2': 0.9428228339672978, 'Bleu_3': 0.9362311358196401, 'Bleu_4': 0.9287320844903101, 'METEOR': 0.7219207481467704, 'ROUGE_L': 0.968410090463815, 'CIDEr': 0.0}
nochange_acc: 0.9594594594594594
change_metric: {'Bleu_1': 0.6574069938312495, 'Bleu_2': 0.537449029230831, 'Bleu_3': 0.43256882850207223, 'Bleu_4': 0.3429312847166659, 'METEOR': 0.23099194196774, 'ROUGE_L': 0.4728300145031497, 'CIDEr': 0.5176775609498402}
change_acc: 0.8350824587706147
.......................................................
Bleu_1 0.766548716151082
Bleu_2 0.6842341223848425
Bleu_3 0.6091800530745962
Bleu_4 0.5376117982399563
METEOR 0.3540162454100901
ROUGE_L 0.7204341634827469
CIDEr 1.2854581393215705

Epochs since last improvement: 8

The best BLEU-4 score: 0.5754

03-25 19 : 04 : 37
Epoch: 9/40 step: 0/267 Loss: 1.5143 AVG_Loss: 1.5143 Top-5 Accuracy: 60.4866 Batch_time: 3.1882s
Epoch: 9/40 step: 50/267 Loss: 1.3603 AVG_Loss: 1.3724 Top-5 Accuracy: 62.6289 Batch_time: 0.6369s
Epoch: 9/40 step: 100/267 Loss: 1.1708 AVG_Loss: 1.3489 Top-5 Accuracy: 66.4623 Batch_time: 0.6625s
Epoch: 9/40 step: 150/267 Loss: 1.2950 AVG_Loss: 1.3576 Top-5 Accuracy: 65.8407 Batch_time: 0.6590s
Epoch: 9/40 step: 200/267 Loss: 1.2324 AVG_Loss: 1.3549 Top-5 Accuracy: 63.7947 Batch_time: 0.6613s
Epoch: 9/40 step: 250/267 Loss: 1.2642 AVG_Loss: 1.3565 Top-5 Accuracy: 64.8166 Batch_time: 0.6724s
eval_time: 55.414637327194214
avg eval_time: 0.041571370838105186
fps: 24.055016224852974
extractor_time: 0.011668205261230469
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9701110162251722, 'Bleu_2': 0.9644912857126247, 'Bleu_3': 0.9602613149919388, 'Bleu_4': 0.9558086765468866, 'METEOR': 0.7551763977324684, 'ROUGE_L': 0.9787978378520454, 'CIDEr': 0.0}
nochange_acc: 0.972972972972973
change_metric: {'Bleu_1': 0.6638790866862757, 'Bleu_2': 0.5380157385679746, 'Bleu_3': 0.428279135561949, 'Bleu_4': 0.3350316462988189, 'METEOR': 0.22580400869253345, 'ROUGE_L': 0.4639699781931928, 'CIDEr': 0.5204510795089898}
change_acc: 0.8140929535232384
.......................................................
Bleu_1 0.778833588415406
Bleu_2 0.6948353937070055
Bleu_3 0.6187104148958211
Bleu_4 0.547239184022357
METEOR 0.35669733937123466
ROUGE_L 0.721190799298066
CIDEr 1.295962278715301

Epochs since last improvement: 9

The best BLEU-4 score: 0.5754

[Decay LR] The new learning rate is 0.000069

03-25 19 : 09 : 05
Epoch: 10/40 step: 0/267 Loss: 1.3514 AVG_Loss: 1.3514 Top-5 Accuracy: 61.7486 Batch_time: 2.0755s
Epoch: 10/40 step: 50/267 Loss: 1.3904 AVG_Loss: 1.3462 Top-5 Accuracy: 62.1343 Batch_time: 0.6605s
Epoch: 10/40 step: 100/267 Loss: 1.4034 AVG_Loss: 1.3483 Top-5 Accuracy: 60.3448 Batch_time: 0.6625s
Epoch: 10/40 step: 150/267 Loss: 1.5350 AVG_Loss: 1.3451 Top-5 Accuracy: 59.0721 Batch_time: 0.6590s
Epoch: 10/40 step: 200/267 Loss: 1.3282 AVG_Loss: 1.3369 Top-5 Accuracy: 63.8352 Batch_time: 0.6589s
Epoch: 10/40 step: 250/267 Loss: 1.3787 AVG_Loss: 1.3364 Top-5 Accuracy: 61.6194 Batch_time: 0.6859s
eval_time: 54.7930634021759
avg eval_time: 0.04110507382008695
fps: 24.32789694958112
extractor_time: 0.011834214347158261
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9241908006811687, 'Bleu_2': 0.9098126881400154, 'Bleu_3': 0.898174371571546, 'Bleu_4': 0.8856412198085187, 'METEOR': 0.6794342254744946, 'ROUGE_L': 0.9499837366941096, 'CIDEr': 0.0}
nochange_acc: 0.9354354354354354
change_metric: {'Bleu_1': 0.6626797902503371, 'Bleu_2': 0.537173156257234, 'Bleu_3': 0.43214772123388073, 'Bleu_4': 0.33882485538212853, 'METEOR': 0.23189352348304731, 'ROUGE_L': 0.4678463650067336, 'CIDEr': 0.5243186602770553}
change_acc: 0.8605697151424287
.......................................................
Bleu_1 0.7617169727813458
Bleu_2 0.6730936518236079
Bleu_3 0.5955256660377064
Bleu_4 0.5202439121579907
METEOR 0.34882785355391466
ROUGE_L 0.7087342041243573
CIDEr 1.2610908166092571

Epochs since last improvement: 10

The best BLEU-4 score: 0.5754

03-25 19 : 13 : 33
Epoch: 11/40 step: 0/267 Loss: 1.5008 AVG_Loss: 1.5008 Top-5 Accuracy: 57.3568 Batch_time: 3.0734s
Epoch: 11/40 step: 50/267 Loss: 1.3859 AVG_Loss: 1.3334 Top-5 Accuracy: 62.9116 Batch_time: 0.6607s
Epoch: 11/40 step: 100/267 Loss: 1.2730 AVG_Loss: 1.3281 Top-5 Accuracy: 64.5760 Batch_time: 0.6579s
Epoch: 11/40 step: 150/267 Loss: 1.2110 AVG_Loss: 1.3205 Top-5 Accuracy: 63.0709 Batch_time: 0.6568s
Epoch: 11/40 step: 200/267 Loss: 1.4104 AVG_Loss: 1.3192 Top-5 Accuracy: 61.5451 Batch_time: 0.6638s
Epoch: 11/40 step: 250/267 Loss: 1.3203 AVG_Loss: 1.3188 Top-5 Accuracy: 61.7316 Batch_time: 0.6857s
eval_time: 53.73110890388489
avg eval_time: 0.04030840878010869
fps: 24.80871932839676
extractor_time: 0.011731626212999325
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9505320678743312, 'Bleu_2': 0.9414266197437029, 'Bleu_3': 0.9336926915044205, 'Bleu_4': 0.9249618832574328, 'METEOR': 0.7219346470470119, 'ROUGE_L': 0.9693240164111785, 'CIDEr': 0.0}
nochange_acc: 0.960960960960961
change_metric: {'Bleu_1': 0.6800907178377615, 'Bleu_2': 0.5537762682378828, 'Bleu_3': 0.44899728280484325, 'Bleu_4': 0.3559941729917583, 'METEOR': 0.23729120513755037, 'ROUGE_L': 0.46936414875927474, 'CIDEr': 0.5716734583216696}
change_acc: 0.8380809595202399
.......................................................
Bleu_1 0.779214153228885
Bleu_2 0.690941226785546
Bleu_3 0.6134441312449531
Bleu_4 0.5375744139082699
METEOR 0.356521678695246
ROUGE_L 0.7191565507518988
CIDEr 1.3085450901604263

Epochs since last improvement: 11

The best BLEU-4 score: 0.5754

03-25 19 : 17 : 59
Epoch: 12/40 step: 0/267 Loss: 1.1575 AVG_Loss: 1.1575 Top-5 Accuracy: 66.3919 Batch_time: 3.3594s
Epoch: 12/40 step: 50/267 Loss: 1.3152 AVG_Loss: 1.3250 Top-5 Accuracy: 61.5385 Batch_time: 0.6595s
Epoch: 12/40 step: 100/267 Loss: 1.3807 AVG_Loss: 1.3087 Top-5 Accuracy: 60.2243 Batch_time: 0.6541s
Epoch: 12/40 step: 150/267 Loss: 1.3103 AVG_Loss: 1.3063 Top-5 Accuracy: 63.1803 Batch_time: 0.6525s
Epoch: 12/40 step: 200/267 Loss: 1.3805 AVG_Loss: 1.3053 Top-5 Accuracy: 64.1476 Batch_time: 0.6546s
Epoch: 12/40 step: 250/267 Loss: 1.4778 AVG_Loss: 1.3014 Top-5 Accuracy: 59.5451 Batch_time: 0.6678s
eval_time: 54.85703253746033
avg eval_time: 0.04115306266876244
fps: 24.299528033889395
extractor_time: 0.011770641544396413
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9780428300349748, 'Bleu_2': 0.9742937531057775, 'Bleu_3': 0.9715409143307268, 'Bleu_4': 0.9687381996594998, 'METEOR': 0.7735594534925164, 'ROUGE_L': 0.9845582464221573, 'CIDEr': 0.0}
nochange_acc: 0.9804804804804805
change_metric: {'Bleu_1': 0.699286176159966, 'Bleu_2': 0.5628597247156365, 'Bleu_3': 0.4487592982393824, 'Bleu_4': 0.35370584572037495, 'METEOR': 0.23852178684147227, 'ROUGE_L': 0.47743680761731505, 'CIDEr': 0.5423530866698286}
change_acc: 0.8200899550224887
.......................................................
Bleu_1 0.8040339020596371
Bleu_2 0.715056139783973
Bleu_3 0.6376582627553661
Bleu_4 0.567067786579089
METEOR 0.36774009503713223
ROUGE_L 0.7308073089256608
CIDEr 1.3189221682280912

Epochs since last improvement: 12

The best BLEU-4 score: 0.5754

[Decay LR] The new learning rate is 0.000048

03-25 19 : 22 : 27
Epoch: 13/40 step: 0/267 Loss: 1.3619 AVG_Loss: 1.3619 Top-5 Accuracy: 63.9662 Batch_time: 2.1052s
Epoch: 13/40 step: 50/267 Loss: 1.2177 AVG_Loss: 1.3147 Top-5 Accuracy: 63.8889 Batch_time: 0.6595s
Epoch: 13/40 step: 100/267 Loss: 1.3106 AVG_Loss: 1.2969 Top-5 Accuracy: 61.4385 Batch_time: 0.6541s
Epoch: 13/40 step: 150/267 Loss: 1.3669 AVG_Loss: 1.2859 Top-5 Accuracy: 62.4088 Batch_time: 0.6608s
Epoch: 13/40 step: 200/267 Loss: 1.3705 AVG_Loss: 1.2839 Top-5 Accuracy: 63.5088 Batch_time: 0.6591s
Epoch: 13/40 step: 250/267 Loss: 1.3113 AVG_Loss: 1.2845 Top-5 Accuracy: 64.5963 Batch_time: 0.6979s
eval_time: 51.89466142654419
avg eval_time: 0.03893072875209617
fps: 25.6866499049586
extractor_time: 0.011643282560504476
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9300199942870808, 'Bleu_2': 0.9180189861981839, 'Bleu_3': 0.9088320460968466, 'Bleu_4': 0.8986876493438409, 'METEOR': 0.6914136852940798, 'ROUGE_L': 0.9522900095901322, 'CIDEr': 0.0}
nochange_acc: 0.93993993993994
change_metric: {'Bleu_1': 0.706896512629427, 'Bleu_2': 0.5837456113611149, 'Bleu_3': 0.477202825976287, 'Bleu_4': 0.38363005727106925, 'METEOR': 0.24837838497703144, 'ROUGE_L': 0.498151241536974, 'CIDEr': 0.5855960172044776}
change_acc: 0.8590704647676162
.......................................................
Bleu_1 0.7920715493009268
Bleu_2 0.7071145265332186
Bleu_3 0.6308139618508578
Bleu_4 0.5572004066156027
METEOR 0.36441527330742746
ROUGE_L 0.7250502809393771
CIDEr 1.3040498582871511

Epochs since last improvement: 13

The best BLEU-4 score: 0.5754

03-25 19 : 26 : 52
Epoch: 14/40 step: 0/267 Loss: 1.3856 AVG_Loss: 1.3856 Top-5 Accuracy: 63.4418 Batch_time: 2.2918s
Epoch: 14/40 step: 50/267 Loss: 1.3744 AVG_Loss: 1.2792 Top-5 Accuracy: 62.1982 Batch_time: 0.6603s
Epoch: 14/40 step: 100/267 Loss: 1.1831 AVG_Loss: 1.2689 Top-5 Accuracy: 64.6182 Batch_time: 0.6624s
Epoch: 14/40 step: 150/267 Loss: 1.2393 AVG_Loss: 1.2739 Top-5 Accuracy: 65.4639 Batch_time: 0.6589s
Epoch: 14/40 step: 200/267 Loss: 1.2288 AVG_Loss: 1.2722 Top-5 Accuracy: 64.8697 Batch_time: 0.6635s
Epoch: 14/40 step: 250/267 Loss: 1.2018 AVG_Loss: 1.2735 Top-5 Accuracy: 65.6584 Batch_time: 0.6744s
eval_time: 53.810052156448364
avg eval_time: 0.04036763102509255
fps: 24.772323136287074
extractor_time: 0.011618288674036185
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9623690399826892, 'Bleu_2': 0.9563589719735315, 'Bleu_3': 0.9531573579057552, 'Bleu_4': 0.9509559617479695, 'METEOR': 0.7368771392006923, 'ROUGE_L': 0.9657840141818881, 'CIDEr': 0.0}
nochange_acc: 0.9564564564564565
change_metric: {'Bleu_1': 0.658150028326297, 'Bleu_2': 0.5239854556945719, 'Bleu_3': 0.4156995757983757, 'Bleu_4': 0.3249535203285979, 'METEOR': 0.2225537270629628, 'ROUGE_L': 0.45892406527784363, 'CIDEr': 0.5280552541950404}
change_acc: 0.8380809595202399
.......................................................
Bleu_1 0.789001738493843
Bleu_2 0.7108699210709917
Bleu_3 0.6483296775443785
Bleu_4 0.5951758747800081
METEOR 0.3757274386293083
ROUGE_L 0.7121639197190242
CIDEr 1.2851787373984507

The best BLEU-4 score: 0.5952

03-25 19 : 31 : 21
Epoch: 15/40 step: 0/267 Loss: 1.3434 AVG_Loss: 1.3434 Top-5 Accuracy: 61.1378 Batch_time: 2.0916s
Epoch: 15/40 step: 50/267 Loss: 1.3218 AVG_Loss: 1.2641 Top-5 Accuracy: 63.6661 Batch_time: 0.6598s
Epoch: 15/40 step: 100/267 Loss: 1.1297 AVG_Loss: 1.2617 Top-5 Accuracy: 66.4835 Batch_time: 0.6619s
Epoch: 15/40 step: 150/267 Loss: 1.2386 AVG_Loss: 1.2645 Top-5 Accuracy: 66.0262 Batch_time: 0.6395s
Epoch: 15/40 step: 200/267 Loss: 1.3721 AVG_Loss: 1.2568 Top-5 Accuracy: 61.2792 Batch_time: 0.6640s
Epoch: 15/40 step: 250/267 Loss: 1.4846 AVG_Loss: 1.2602 Top-5 Accuracy: 60.9261 Batch_time: 0.6881s
eval_time: 62.99709367752075
avg eval_time: 0.047259635166932296
fps: 21.15970630047739
extractor_time: 0.011670906026830194
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9701230228468161, 'Bleu_2': 0.9635692136946522, 'Bleu_3': 0.958393997131695, 'Bleu_4': 0.9524983819337642, 'METEOR': 0.7499280838779933, 'ROUGE_L': 0.9812469344547032, 'CIDEr': 0.0}
nochange_acc: 0.975975975975976
change_metric: {'Bleu_1': 0.6875529444707804, 'Bleu_2': 0.567478901158873, 'Bleu_3': 0.4641274245913259, 'Bleu_4': 0.37397667179218247, 'METEOR': 0.23579250136435997, 'ROUGE_L': 0.47514359705262477, 'CIDEr': 0.548137784391706}
change_acc: 0.8020989505247377
.......................................................
Bleu_1 0.7940023862754235
Bleu_2 0.7130105506444891
Bleu_3 0.6398969280875303
Bleu_4 0.5689851088599449
METEOR 0.3620744461502533
ROUGE_L 0.7280054295430856
CIDEr 1.3157488119087468

Epochs since last improvement: 1

The best BLEU-4 score: 0.5952

03-25 19 : 35 : 56
Epoch: 16/40 step: 0/267 Loss: 1.2633 AVG_Loss: 1.2633 Top-5 Accuracy: 63.7681 Batch_time: 2.1347s
Epoch: 16/40 step: 50/267 Loss: 1.1418 AVG_Loss: 1.2464 Top-5 Accuracy: 66.4835 Batch_time: 0.6577s
Epoch: 16/40 step: 100/267 Loss: 1.2283 AVG_Loss: 1.2459 Top-5 Accuracy: 63.6130 Batch_time: 0.6608s
Epoch: 16/40 step: 150/267 Loss: 1.1853 AVG_Loss: 1.2467 Top-5 Accuracy: 66.6667 Batch_time: 0.6605s
Epoch: 16/40 step: 200/267 Loss: 1.3211 AVG_Loss: 1.2460 Top-5 Accuracy: 65.5477 Batch_time: 0.6532s
Epoch: 16/40 step: 250/267 Loss: 1.1939 AVG_Loss: 1.2488 Top-5 Accuracy: 66.3082 Batch_time: 0.6790s
eval_time: 54.37233352661133
avg eval_time: 0.040789447506835205
fps: 24.516144765933078
extractor_time: 0.011743416575140641
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9150546677878006, 'Bleu_2': 0.8982555847604795, 'Bleu_3': 0.8862232061936138, 'Bleu_4': 0.8734439910685938, 'METEOR': 0.6652141826433587, 'ROUGE_L': 0.9414997499670564, 'CIDEr': 0.0}
nochange_acc: 0.924924924924925
change_metric: {'Bleu_1': 0.7242429102105502, 'Bleu_2': 0.597145330536907, 'Bleu_3': 0.4826161053567679, 'Bleu_4': 0.38418865966311555, 'METEOR': 0.2521143093381142, 'ROUGE_L': 0.5006773178344169, 'CIDEr': 0.6031249236805437}
change_acc: 0.8710644677661169
.......................................................
Bleu_1 0.7987549749973671
Bleu_2 0.709005761076704
Bleu_3 0.6264831931422654
Bleu_4 0.5488823134887597
METEOR 0.3613140836140679
ROUGE_L 0.7209231841512497
CIDEr 1.2993213760390745

Epochs since last improvement: 2

The best BLEU-4 score: 0.5952

03-25 19 : 40 : 24
Epoch: 17/40 step: 0/267 Loss: 1.1997 AVG_Loss: 1.1997 Top-5 Accuracy: 65.3356 Batch_time: 2.0530s
Epoch: 17/40 step: 50/267 Loss: 1.2914 AVG_Loss: 1.2323 Top-5 Accuracy: 63.3120 Batch_time: 0.6617s
Epoch: 17/40 step: 100/267 Loss: 1.2543 AVG_Loss: 1.2412 Top-5 Accuracy: 65.1920 Batch_time: 0.6603s
Epoch: 17/40 step: 150/267 Loss: 1.1981 AVG_Loss: 1.2339 Top-5 Accuracy: 66.6373 Batch_time: 0.6603s
Epoch: 17/40 step: 200/267 Loss: 1.2168 AVG_Loss: 1.2372 Top-5 Accuracy: 63.6802 Batch_time: 0.6687s
Epoch: 17/40 step: 250/267 Loss: 1.1269 AVG_Loss: 1.2388 Top-5 Accuracy: 67.9853 Batch_time: 0.7044s
eval_time: 64.15506362915039
avg eval_time: 0.04812832980431387
fps: 20.777783149050133
extractor_time: 0.012389847504314586
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9614373356701836, 'Bleu_2': 0.9531164657326017, 'Bleu_3': 0.9463789250814706, 'Bleu_4': 0.9388168450217267, 'METEOR': 0.7372500314926644, 'ROUGE_L': 0.9752211351968653, 'CIDEr': 0.0}
nochange_acc: 0.9684684684684685
change_metric: {'Bleu_1': 0.685252946409553, 'Bleu_2': 0.5547993266253757, 'Bleu_3': 0.4464133248850584, 'Bleu_4': 0.35500982397711556, 'METEOR': 0.2366100241295963, 'ROUGE_L': 0.47141872473279883, 'CIDEr': 0.5429280066064701}
change_acc: 0.8320839580209896
.......................................................
Bleu_1 0.7873011567711901
Bleu_2 0.697462504414182
Bleu_3 0.6186934968637938
Bleu_4 0.5441824923752381
METEOR 0.3578930569293262
ROUGE_L 0.7231309568176212
CIDEr 1.304449988415121

Epochs since last improvement: 3

The best BLEU-4 score: 0.5952

[Decay LR] The new learning rate is 0.000034

03-25 19 : 45 : 01
Epoch: 18/40 step: 0/267 Loss: 1.0012 AVG_Loss: 1.0012 Top-5 Accuracy: 68.9952 Batch_time: 3.4503s
Epoch: 18/40 step: 50/267 Loss: 1.2677 AVG_Loss: 1.2215 Top-5 Accuracy: 62.9347 Batch_time: 0.6602s
Epoch: 18/40 step: 100/267 Loss: 1.2558 AVG_Loss: 1.2249 Top-5 Accuracy: 63.1399 Batch_time: 0.6617s
Epoch: 18/40 step: 150/267 Loss: 1.1758 AVG_Loss: 1.2240 Top-5 Accuracy: 66.1934 Batch_time: 0.6597s
Epoch: 18/40 step: 200/267 Loss: 1.2579 AVG_Loss: 1.2290 Top-5 Accuracy: 63.6364 Batch_time: 0.6620s
Epoch: 18/40 step: 250/267 Loss: 1.2273 AVG_Loss: 1.2275 Top-5 Accuracy: 63.1485 Batch_time: 0.6596s
eval_time: 57.29163098335266
avg eval_time: 0.04297946810454063
fps: 23.266923582387317
extractor_time: 0.011799931973330704
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9607558139532091, 'Bleu_2': 0.9536755382426079, 'Bleu_3': 0.9480529484669222, 'Bleu_4': 0.9415557503344132, 'METEOR': 0.7433974799957, 'ROUGE_L': 0.9749178399646465, 'CIDEr': 0.0}
nochange_acc: 0.9684684684684685
change_metric: {'Bleu_1': 0.6855568495586264, 'Bleu_2': 0.557364887442422, 'Bleu_3': 0.4464908329755081, 'Bleu_4': 0.35016174378303583, 'METEOR': 0.2373191600383368, 'ROUGE_L': 0.46556427941969813, 'CIDEr': 0.5357024251111729}
change_acc: 0.8275862068965517
.......................................................
Bleu_1 0.7861741161879425
Bleu_2 0.6981507214830075
Bleu_3 0.6183425564327478
Bleu_4 0.541440606657672
METEOR 0.35971609193897147
ROUGE_L 0.7200500043431307
CIDEr 1.2963541717237654

Epochs since last improvement: 4

The best BLEU-4 score: 0.5952

03-25 19 : 49 : 31
Epoch: 19/40 step: 0/267 Loss: 1.1990 AVG_Loss: 1.1990 Top-5 Accuracy: 63.4699 Batch_time: 3.4468s
Epoch: 19/40 step: 50/267 Loss: 1.2928 AVG_Loss: 1.2306 Top-5 Accuracy: 62.3388 Batch_time: 0.6601s
Epoch: 19/40 step: 100/267 Loss: 1.2876 AVG_Loss: 1.2316 Top-5 Accuracy: 64.3172 Batch_time: 0.6566s
Epoch: 19/40 step: 150/267 Loss: 1.1432 AVG_Loss: 1.2207 Top-5 Accuracy: 66.4799 Batch_time: 0.6599s
Epoch: 19/40 step: 200/267 Loss: 1.2075 AVG_Loss: 1.2206 Top-5 Accuracy: 66.9929 Batch_time: 0.6630s
Epoch: 19/40 step: 250/267 Loss: 1.1582 AVG_Loss: 1.2203 Top-5 Accuracy: 66.8153 Batch_time: 0.6611s
eval_time: 53.20582890510559
avg eval_time: 0.039914350266395796
fps: 25.05364595254123
extractor_time: 0.011826485328359525
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9506314580938718, 'Bleu_2': 0.9417397230318418, 'Bleu_3': 0.9355617546666853, 'Bleu_4': 0.928716335518781, 'METEOR': 0.7172390286602349, 'ROUGE_L': 0.965779051378461, 'CIDEr': 0.0}
nochange_acc: 0.9564564564564565
change_metric: {'Bleu_1': 0.6917596853016037, 'Bleu_2': 0.5695793907601647, 'Bleu_3': 0.464120281834676, 'Bleu_4': 0.3737934974397939, 'METEOR': 0.24104671743646533, 'ROUGE_L': 0.47949369764603267, 'CIDEr': 0.5771631403786737}
change_acc: 0.8470764617691154
.......................................................
Bleu_1 0.7882383944928201
Bleu_2 0.7046705707489465
Bleu_3 0.6295826727688439
Bleu_4 0.5581036135320055
METEOR 0.36234917865233474
ROUGE_L 0.7224539719039451
CIDEr 1.311188134988817

Epochs since last improvement: 5

The best BLEU-4 score: 0.5952

03-25 19 : 53 : 57
Epoch: 20/40 step: 0/267 Loss: 1.2921 AVG_Loss: 1.2921 Top-5 Accuracy: 62.6374 Batch_time: 2.0970s
Epoch: 20/40 step: 50/267 Loss: 1.1951 AVG_Loss: 1.1903 Top-5 Accuracy: 64.4406 Batch_time: 0.6617s
Epoch: 20/40 step: 100/267 Loss: 1.1278 AVG_Loss: 1.1984 Top-5 Accuracy: 66.6073 Batch_time: 0.6603s
Epoch: 20/40 step: 150/267 Loss: 1.3702 AVG_Loss: 1.2028 Top-5 Accuracy: 60.9325 Batch_time: 0.6599s
Epoch: 20/40 step: 200/267 Loss: 1.3300 AVG_Loss: 1.2081 Top-5 Accuracy: 62.8571 Batch_time: 0.6615s
Epoch: 20/40 step: 250/267 Loss: 1.2136 AVG_Loss: 1.2115 Top-5 Accuracy: 66.0228 Batch_time: 0.6662s
eval_time: 55.8706169128418
avg eval_time: 0.04191344104489257
fps: 23.858694849915135
extractor_time: 0.011740268841777094
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9665085124194902, 'Bleu_2': 0.9599201795602987, 'Bleu_3': 0.9547419445177795, 'Bleu_4': 0.9491302036985926, 'METEOR': 0.745364081136892, 'ROUGE_L': 0.9782181147082488, 'CIDEr': 0.0}
nochange_acc: 0.9714714714714715
change_metric: {'Bleu_1': 0.7048892744638348, 'Bleu_2': 0.5822277760782931, 'Bleu_3': 0.4741427393265048, 'Bleu_4': 0.38252713894833, 'METEOR': 0.2445404278531116, 'ROUGE_L': 0.48891950494475567, 'CIDEr': 0.6058133465670821}
change_acc: 0.8290854572713643
.......................................................
Bleu_1 0.8041640950893468
Bleu_2 0.7219933351978479
Bleu_3 0.6467660009474329
Bleu_4 0.5763656018267631
METEOR 0.3687785177555799
ROUGE_L 0.7333852769646254
CIDEr 1.3407203163910182

Epochs since last improvement: 6

The best BLEU-4 score: 0.5952

[Decay LR] The new learning rate is 0.000024

03-25 19 : 58 : 25
Epoch: 21/40 step: 0/267 Loss: 1.1493 AVG_Loss: 1.1493 Top-5 Accuracy: 66.2478 Batch_time: 3.3568s
Epoch: 21/40 step: 50/267 Loss: 1.2201 AVG_Loss: 1.1791 Top-5 Accuracy: 66.0090 Batch_time: 0.6583s
Epoch: 21/40 step: 100/267 Loss: 1.1271 AVG_Loss: 1.1824 Top-5 Accuracy: 66.2350 Batch_time: 0.6652s
Epoch: 21/40 step: 150/267 Loss: 1.1298 AVG_Loss: 1.1900 Top-5 Accuracy: 65.7801 Batch_time: 0.6617s
Epoch: 21/40 step: 200/267 Loss: 1.0507 AVG_Loss: 1.1986 Top-5 Accuracy: 67.0300 Batch_time: 0.6602s
Epoch: 21/40 step: 250/267 Loss: 1.2107 AVG_Loss: 1.2015 Top-5 Accuracy: 64.1674 Batch_time: 0.6715s
eval_time: 55.85975122451782
avg eval_time: 0.041905289740823574
fps: 23.86333577896285
extractor_time: 0.011815186648644277
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9570323689484511, 'Bleu_2': 0.94892431632411, 'Bleu_3': 0.941497560785981, 'Bleu_4': 0.9330703085168909, 'METEOR': 0.7344508541266338, 'ROUGE_L': 0.9752485215576049, 'CIDEr': 0.0}
nochange_acc: 0.9684684684684685
change_metric: {'Bleu_1': 0.6853972757091193, 'Bleu_2': 0.5536440315671016, 'Bleu_3': 0.445165076199253, 'Bleu_4': 0.35330326289603814, 'METEOR': 0.23931315903976438, 'ROUGE_L': 0.4735295574286669, 'CIDEr': 0.5736025754213465}
change_acc: 0.8455772113943029
.......................................................
Bleu_1 0.7875831157341732
Bleu_2 0.6973255457776428
Bleu_3 0.6186506850981278
Bleu_4 0.5443611704381364
METEOR 0.36246813361943686
ROUGE_L 0.7242008478336727
CIDEr 1.3171283324403291

Epochs since last improvement: 7

The best BLEU-4 score: 0.5952

03-25 20 : 02 : 54
Epoch: 22/40 step: 0/267 Loss: 1.2033 AVG_Loss: 1.2033 Top-5 Accuracy: 65.1910 Batch_time: 3.1980s
Epoch: 22/40 step: 50/267 Loss: 1.1968 AVG_Loss: 1.1828 Top-5 Accuracy: 64.7687 Batch_time: 0.6700s
Epoch: 22/40 step: 100/267 Loss: 1.1708 AVG_Loss: 1.1874 Top-5 Accuracy: 65.3881 Batch_time: 0.6638s
Epoch: 22/40 step: 150/267 Loss: 1.3057 AVG_Loss: 1.1911 Top-5 Accuracy: 62.9446 Batch_time: 0.6879s
Epoch: 22/40 step: 200/267 Loss: 1.0793 AVG_Loss: 1.1940 Top-5 Accuracy: 69.2035 Batch_time: 0.6748s
Epoch: 22/40 step: 250/267 Loss: 1.3139 AVG_Loss: 1.1974 Top-5 Accuracy: 62.6412 Batch_time: 0.6906s
eval_time: 65.19101738929749
avg eval_time: 0.048905489414326694
fps: 20.4476023443506
extractor_time: 0.012287360604150976
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9704797047967725, 'Bleu_2': 0.9650989293799423, 'Bleu_3': 0.9607385075020625, 'Bleu_4': 0.9559277027106658, 'METEOR': 0.7575011520641118, 'ROUGE_L': 0.9802261525433785, 'CIDEr': 0.0}
nochange_acc: 0.9744744744744744
change_metric: {'Bleu_1': 0.7047479471084318, 'Bleu_2': 0.5781642324677145, 'Bleu_3': 0.47095549163077705, 'Bleu_4': 0.3806237673885782, 'METEOR': 0.24512767905281493, 'ROUGE_L': 0.4887985073457388, 'CIDEr': 0.5760265771926372}
change_acc: 0.8215892053973014
.......................................................
Bleu_1 0.8041457236119709
Bleu_2 0.7195665381157734
Bleu_3 0.6443594149872081
Bleu_4 0.5738454941154346
METEOR 0.37011416191402047
ROUGE_L 0.7343279984947472
CIDEr 1.3310609716600965

Epochs since last improvement: 8

The best BLEU-4 score: 0.5952

03-25 20 : 07 : 34
Epoch: 23/40 step: 0/267 Loss: 1.1457 AVG_Loss: 1.1457 Top-5 Accuracy: 66.1376 Batch_time: 2.1881s
Epoch: 23/40 step: 50/267 Loss: 1.1864 AVG_Loss: 1.1778 Top-5 Accuracy: 65.9649 Batch_time: 0.6593s
Epoch: 23/40 step: 100/267 Loss: 1.1325 AVG_Loss: 1.1766 Top-5 Accuracy: 67.3225 Batch_time: 0.6509s
Epoch: 23/40 step: 150/267 Loss: 1.2615 AVG_Loss: 1.1804 Top-5 Accuracy: 63.3721 Batch_time: 0.6551s
Epoch: 23/40 step: 200/267 Loss: 1.2942 AVG_Loss: 1.1881 Top-5 Accuracy: 63.8565 Batch_time: 0.6726s
Epoch: 23/40 step: 250/267 Loss: 1.1421 AVG_Loss: 1.1892 Top-5 Accuracy: 66.9369 Batch_time: 0.6699s
eval_time: 59.35249662399292
avg eval_time: 0.04452550384395568
fps: 22.459038386283183
extractor_time: 0.012084553855214902
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9504662004659236, 'Bleu_2': 0.9401079666764269, 'Bleu_3': 0.9330347668079861, 'Bleu_4': 0.9256298099035087, 'METEOR': 0.7119136831144839, 'ROUGE_L': 0.9630792691361174, 'CIDEr': 0.0}
nochange_acc: 0.9519519519519519
change_metric: {'Bleu_1': 0.6893094693292973, 'Bleu_2': 0.5630079770850803, 'Bleu_3': 0.45383231616801056, 'Bleu_4': 0.3587913508832658, 'METEOR': 0.2368010200261784, 'ROUGE_L': 0.4701405771596538, 'CIDEr': 0.5560273431264701}
change_acc: 0.8260869565217391
.......................................................
Bleu_1 0.7856034164795279
Bleu_2 0.6979731734628369
Bleu_3 0.6191852295951283
Bleu_4 0.5433327146806309
METEOR 0.3561465233259572
ROUGE_L 0.7164250249138359
CIDEr 1.2950323576561185

Epochs since last improvement: 9

The best BLEU-4 score: 0.5952

[Decay LR] The new learning rate is 0.000016

03-25 20 : 12 : 05
Epoch: 24/40 step: 0/267 Loss: 1.3052 AVG_Loss: 1.3052 Top-5 Accuracy: 62.8478 Batch_time: 2.3160s
Epoch: 24/40 step: 50/267 Loss: 1.0925 AVG_Loss: 1.1646 Top-5 Accuracy: 67.1362 Batch_time: 0.6626s
Epoch: 24/40 step: 100/267 Loss: 1.0474 AVG_Loss: 1.1756 Top-5 Accuracy: 67.6923 Batch_time: 0.6631s
Epoch: 24/40 step: 150/267 Loss: 1.1148 AVG_Loss: 1.1784 Top-5 Accuracy: 67.0132 Batch_time: 0.6603s
Epoch: 24/40 step: 200/267 Loss: 1.1274 AVG_Loss: 1.1787 Top-5 Accuracy: 66.2963 Batch_time: 0.6599s
Epoch: 24/40 step: 250/267 Loss: 1.2975 AVG_Loss: 1.1804 Top-5 Accuracy: 63.7789 Batch_time: 0.6721s
eval_time: 56.851722240448
avg eval_time: 0.04264945404384696
fps: 23.446958992064054
extractor_time: 0.011774992281033058
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9645888206025858, 'Bleu_2': 0.9571842012597536, 'Bleu_3': 0.9519228708548785, 'Bleu_4': 0.9460946239300296, 'METEOR': 0.7374986337236392, 'ROUGE_L': 0.97538877935539, 'CIDEr': 0.0}
nochange_acc: 0.9684684684684685
change_metric: {'Bleu_1': 0.7151755207610652, 'Bleu_2': 0.591149765898043, 'Bleu_3': 0.4846549820930561, 'Bleu_4': 0.3912719304194955, 'METEOR': 0.25266004369523415, 'ROUGE_L': 0.49576905646516156, 'CIDEr': 0.6079942896570532}
change_acc: 0.8260869565217391
.......................................................
Bleu_1 0.8064861206776697
Bleu_2 0.7212768864314626
Bleu_3 0.6445213583773817
Bleu_4 0.5697963853374762
METEOR 0.3719497631938027
ROUGE_L 0.735399015538599
CIDEr 1.3417596307096586

Epochs since last improvement: 10

The best BLEU-4 score: 0.5952

03-25 20 : 16 : 35
Epoch: 25/40 step: 0/267 Loss: 1.0598 AVG_Loss: 1.0598 Top-5 Accuracy: 67.7820 Batch_time: 2.0988s
Epoch: 25/40 step: 50/267 Loss: 1.1116 AVG_Loss: 1.1631 Top-5 Accuracy: 67.4699 Batch_time: 0.6612s
Epoch: 25/40 step: 100/267 Loss: 1.2198 AVG_Loss: 1.1735 Top-5 Accuracy: 63.1957 Batch_time: 0.6614s
Epoch: 25/40 step: 150/267 Loss: 1.1154 AVG_Loss: 1.1736 Top-5 Accuracy: 66.9326 Batch_time: 0.6583s
Epoch: 25/40 step: 200/267 Loss: 1.3178 AVG_Loss: 1.1775 Top-5 Accuracy: 64.4406 Batch_time: 0.6504s
Epoch: 25/40 step: 250/267 Loss: 1.2510 AVG_Loss: 1.1756 Top-5 Accuracy: 64.7255 Batch_time: 0.6877s
eval_time: 55.073089599609375
avg eval_time: 0.04131514598620358
fps: 24.204198632964562
extractor_time: 0.011682577269111285
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9647194910350015, 'Bleu_2': 0.9574397288733754, 'Bleu_3': 0.9510445101309791, 'Bleu_4': 0.9440280339187572, 'METEOR': 0.7433991963023647, 'ROUGE_L': 0.9781222906682886, 'CIDEr': 0.0}
nochange_acc: 0.9714714714714715
change_metric: {'Bleu_1': 0.693985907881374, 'Bleu_2': 0.5651148227057639, 'Bleu_3': 0.4558320500920878, 'Bleu_4': 0.36387302330245014, 'METEOR': 0.23645943212623277, 'ROUGE_L': 0.4741481706925187, 'CIDEr': 0.5473579515143976}
change_acc: 0.8140929535232384
.......................................................
Bleu_1 0.7968215581602555
Bleu_2 0.7099978709065928
Bleu_3 0.6325749083201793
Bleu_4 0.5597995098690305
METEOR 0.36167016442915423
ROUGE_L 0.7259461931260242
CIDEr 1.3095581846313333

Epochs since last improvement: 11

The best BLEU-4 score: 0.5952

03-25 20 : 21 : 02
Epoch: 26/40 step: 0/267 Loss: 0.9500 AVG_Loss: 0.9500 Top-5 Accuracy: 70.8934 Batch_time: 3.2181s
Epoch: 26/40 step: 50/267 Loss: 1.1777 AVG_Loss: 1.1596 Top-5 Accuracy: 65.6331 Batch_time: 0.6595s
Epoch: 26/40 step: 100/267 Loss: 1.0861 AVG_Loss: 1.1611 Top-5 Accuracy: 66.2921 Batch_time: 0.6533s
Epoch: 26/40 step: 150/267 Loss: 1.0545 AVG_Loss: 1.1658 Top-5 Accuracy: 68.2261 Batch_time: 0.6567s
Epoch: 26/40 step: 200/267 Loss: 1.2062 AVG_Loss: 1.1672 Top-5 Accuracy: 64.4794 Batch_time: 0.6589s
Epoch: 26/40 step: 250/267 Loss: 1.3055 AVG_Loss: 1.1714 Top-5 Accuracy: 63.1003 Batch_time: 0.6752s
eval_time: 55.17216634750366
avg eval_time: 0.04138947212865991
fps: 24.160733359717227
extractor_time: 0.01169826120518243
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9646792879341721, 'Bleu_2': 0.9581266479884006, 'Bleu_3': 0.9524159661801612, 'Bleu_4': 0.9462444606183078, 'METEOR': 0.7459447222107807, 'ROUGE_L': 0.9781412961390742, 'CIDEr': 0.0}
nochange_acc: 0.9714714714714715
change_metric: {'Bleu_1': 0.700564290879747, 'Bleu_2': 0.569108697757279, 'Bleu_3': 0.45794843172887834, 'Bleu_4': 0.3642900577633627, 'METEOR': 0.24082249403945857, 'ROUGE_L': 0.47464621487294556, 'CIDEr': 0.5546447012724093}
change_acc: 0.815592203898051
.......................................................
Bleu_1 0.8001101743716161
Bleu_2 0.7118216790622768
Bleu_3 0.6335033357072118
Bleu_4 0.5603838151857418
METEOR 0.36469089059506204
ROUGE_L 0.7262048976360675
CIDEr 1.3159439195604223

Epochs since last improvement: 12

The best BLEU-4 score: 0.5952

[Decay LR] The new learning rate is 0.000012

03-25 20 : 25 : 30
Epoch: 27/40 step: 0/267 Loss: 1.1025 AVG_Loss: 1.1025 Top-5 Accuracy: 67.5000 Batch_time: 2.1295s
Epoch: 27/40 step: 50/267 Loss: 1.1371 AVG_Loss: 1.1489 Top-5 Accuracy: 66.8930 Batch_time: 0.6631s
Epoch: 27/40 step: 100/267 Loss: 1.2683 AVG_Loss: 1.1533 Top-5 Accuracy: 63.5828 Batch_time: 0.6573s
Epoch: 27/40 step: 150/267 Loss: 1.3393 AVG_Loss: 1.1609 Top-5 Accuracy: 63.9901 Batch_time: 0.6592s
Epoch: 27/40 step: 200/267 Loss: 1.1780 AVG_Loss: 1.1615 Top-5 Accuracy: 65.1121 Batch_time: 0.6618s
Epoch: 27/40 step: 250/267 Loss: 1.1454 AVG_Loss: 1.1617 Top-5 Accuracy: 66.9442 Batch_time: 0.6669s
eval_time: 57.63129472732544
avg eval_time: 0.04323427961539793
fps: 23.129794433855885
extractor_time: 0.011781509711343546
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.947339398952041, 'Bleu_2': 0.9383347731718374, 'Bleu_3': 0.9315349545421572, 'Bleu_4': 0.9241626860478296, 'METEOR': 0.7198906381515723, 'ROUGE_L': 0.9653296105289871, 'CIDEr': 0.0}
nochange_acc: 0.9564564564564565
change_metric: {'Bleu_1': 0.6993674890574371, 'Bleu_2': 0.564222459618036, 'Bleu_3': 0.4508260466432898, 'Bleu_4': 0.35512767236683274, 'METEOR': 0.24062402743616704, 'ROUGE_L': 0.4810211778009142, 'CIDEr': 0.5619519479516588}
change_acc: 0.856071964017991
.......................................................
Bleu_1 0.7929354445797002
Bleu_2 0.7020330996104015
Bleu_3 0.6227863789312619
Bleu_4 0.5493641072745399
METEOR 0.36312483340251905
ROUGE_L 0.7229937330874083
CIDEr 1.3071449375720432

Epochs since last improvement: 13

The best BLEU-4 score: 0.5952

03-25 20 : 30 : 00
Epoch: 28/40 step: 0/267 Loss: 1.2264 AVG_Loss: 1.2264 Top-5 Accuracy: 65.3915 Batch_time: 3.1249s
Epoch: 28/40 step: 50/267 Loss: 1.2159 AVG_Loss: 1.1493 Top-5 Accuracy: 64.0381 Batch_time: 0.6486s
Epoch: 28/40 step: 100/267 Loss: 1.2574 AVG_Loss: 1.1495 Top-5 Accuracy: 67.0618 Batch_time: 0.6633s
Epoch: 28/40 step: 150/267 Loss: 1.1847 AVG_Loss: 1.1556 Top-5 Accuracy: 65.9649 Batch_time: 0.6723s
Epoch: 28/40 step: 200/267 Loss: 1.2001 AVG_Loss: 1.1578 Top-5 Accuracy: 65.2668 Batch_time: 0.6476s
Epoch: 28/40 step: 250/267 Loss: 1.0879 AVG_Loss: 1.1593 Top-5 Accuracy: 67.4054 Batch_time: 0.6788s
eval_time: 54.762986183166504
avg eval_time: 0.04108251026494111
fps: 24.34125844674534
extractor_time: 0.01174302272839557
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.963698093496745, 'Bleu_2': 0.9576441317594736, 'Bleu_3': 0.952916870408227, 'Bleu_4': 0.9481306875402533, 'METEOR': 0.7477941473468552, 'ROUGE_L': 0.9753860809204303, 'CIDEr': 0.0}
nochange_acc: 0.9684684684684685
change_metric: {'Bleu_1': 0.6862165978253277, 'Bleu_2': 0.550502384344542, 'Bleu_3': 0.43880889180747074, 'Bleu_4': 0.3483845278049977, 'METEOR': 0.2354565029665326, 'ROUGE_L': 0.4664113393724119, 'CIDEr': 0.5322374283188342}
change_acc: 0.8050974512743628
.......................................................
Bleu_1 0.7954713123145924
Bleu_2 0.7079472844781478
Bleu_3 0.6334077626781925
Bleu_4 0.5674049690118946
METEOR 0.3676679618495098
ROUGE_L 0.7207077968900264
CIDEr 1.299390210979416

Epochs since last improvement: 14

The best BLEU-4 score: 0.5952

03-25 20 : 34 : 28
Epoch: 29/40 step: 0/267 Loss: 1.1488 AVG_Loss: 1.1488 Top-5 Accuracy: 65.6420 Batch_time: 2.1353s
Epoch: 29/40 step: 50/267 Loss: 1.2429 AVG_Loss: 1.1279 Top-5 Accuracy: 64.4771 Batch_time: 0.6630s
Epoch: 29/40 step: 100/267 Loss: 1.1119 AVG_Loss: 1.1388 Top-5 Accuracy: 66.4200 Batch_time: 0.6687s
Epoch: 29/40 step: 150/267 Loss: 1.3162 AVG_Loss: 1.1494 Top-5 Accuracy: 62.1044 Batch_time: 0.6624s
Epoch: 29/40 step: 200/267 Loss: 1.0560 AVG_Loss: 1.1510 Top-5 Accuracy: 67.2727 Batch_time: 0.6618s
Epoch: 29/40 step: 250/267 Loss: 1.2308 AVG_Loss: 1.1537 Top-5 Accuracy: 64.4934 Batch_time: 0.6785s
eval_time: 57.26002025604248
avg eval_time: 0.0429557541305645
fps: 23.279768222913482
extractor_time: 0.011774223367701056
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9514986376019205, 'Bleu_2': 0.9429210421840788, 'Bleu_3': 0.9359478153670157, 'Bleu_4': 0.9282418113479799, 'METEOR': 0.7291016689676884, 'ROUGE_L': 0.9698219575447234, 'CIDEr': 0.0}
nochange_acc: 0.960960960960961
change_metric: {'Bleu_1': 0.7095228316878919, 'Bleu_2': 0.5778289562481456, 'Bleu_3': 0.46516445755103797, 'Bleu_4': 0.3703849781022972, 'METEOR': 0.24628816742165285, 'ROUGE_L': 0.4887178303445705, 'CIDEr': 0.5801186773722681}
change_acc: 0.8305847076461769
.......................................................
Bleu_1 0.8046251392508234
Bleu_2 0.7165651882861429
Bleu_3 0.638451453175515
Bleu_4 0.566371736058489
METEOR 0.3692017038679146
ROUGE_L 0.7290894347821564
CIDEr 1.3197131294147497

Epochs since last improvement: 15

The best BLEU-4 score: 0.5952

[Decay LR] The new learning rate is 0.000008

03-25 20 : 38 : 58
Epoch: 30/40 step: 0/267 Loss: 1.2210 AVG_Loss: 1.2210 Top-5 Accuracy: 65.2406 Batch_time: 3.0896s
Epoch: 30/40 step: 50/267 Loss: 1.2407 AVG_Loss: 1.1407 Top-5 Accuracy: 63.5530 Batch_time: 0.6610s
Epoch: 30/40 step: 100/267 Loss: 1.1773 AVG_Loss: 1.1358 Top-5 Accuracy: 66.3591 Batch_time: 0.6586s
Epoch: 30/40 step: 150/267 Loss: 1.1699 AVG_Loss: 1.1414 Top-5 Accuracy: 65.0943 Batch_time: 0.6602s
Epoch: 30/40 step: 200/267 Loss: 1.1235 AVG_Loss: 1.1468 Top-5 Accuracy: 67.1329 Batch_time: 0.6645s
Epoch: 30/40 step: 250/267 Loss: 1.1177 AVG_Loss: 1.1503 Top-5 Accuracy: 68.1186 Batch_time: 0.6954s
eval_time: 55.76529383659363
avg eval_time: 0.04183442898469139
fps: 23.903756409963986
extractor_time: 0.011670419531304946
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9588299024916146, 'Bleu_2': 0.9511289030527312, 'Bleu_3': 0.9450642974591031, 'Bleu_4': 0.9386626328060669, 'METEOR': 0.735602468254887, 'ROUGE_L': 0.974018356864583, 'CIDEr': 0.0}
nochange_acc: 0.9669669669669669
change_metric: {'Bleu_1': 0.7049522055693285, 'Bleu_2': 0.5670027001525827, 'Bleu_3': 0.450557820848164, 'Bleu_4': 0.35553901118432385, 'METEOR': 0.24023051144289811, 'ROUGE_L': 0.4727526553009607, 'CIDEr': 0.5366800050217869}
change_acc: 0.8185907046476761
.......................................................
Bleu_1 0.8023799126042396
Bleu_2 0.7102794281912409
Bleu_3 0.6296927887009862
Bleu_4 0.5572608519969482
METEOR 0.3646777161106535
ROUGE_L 0.7231974844392746
CIDEr 1.3028676416335796

Epochs since last improvement: 16

The best BLEU-4 score: 0.5952

03-25 20 : 43 : 27
Epoch: 31/40 step: 0/267 Loss: 1.1696 AVG_Loss: 1.1696 Top-5 Accuracy: 64.8241 Batch_time: 3.0530s
Epoch: 31/40 step: 50/267 Loss: 1.1266 AVG_Loss: 1.1330 Top-5 Accuracy: 66.2953 Batch_time: 0.6642s
Epoch: 31/40 step: 100/267 Loss: 1.0546 AVG_Loss: 1.1341 Top-5 Accuracy: 70.0661 Batch_time: 0.6601s
Epoch: 31/40 step: 150/267 Loss: 1.1034 AVG_Loss: 1.1361 Top-5 Accuracy: 66.9573 Batch_time: 0.6587s
Epoch: 31/40 step: 200/267 Loss: 1.2206 AVG_Loss: 1.1432 Top-5 Accuracy: 64.8390 Batch_time: 0.6590s
Epoch: 31/40 step: 250/267 Loss: 1.1637 AVG_Loss: 1.1455 Top-5 Accuracy: 66.1398 Batch_time: 0.6714s
eval_time: 54.23042869567871
avg eval_time: 0.04068299226982649
fps: 24.58029619275753
extractor_time: 0.011598257041925429
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9481586402263603, 'Bleu_2': 0.9380618736592611, 'Bleu_3': 0.930284498201475, 'Bleu_4': 0.9218151256058333, 'METEOR': 0.7151156915247742, 'ROUGE_L': 0.9669743757838475, 'CIDEr': 0.0}
nochange_acc: 0.9564564564564565
change_metric: {'Bleu_1': 0.688224578769168, 'Bleu_2': 0.5570229473054389, 'Bleu_3': 0.4464599742143127, 'Bleu_4': 0.35276417596227694, 'METEOR': 0.23839212476359808, 'ROUGE_L': 0.47270693965063076, 'CIDEr': 0.5507698242221464}
change_acc: 0.8275862068965517
.......................................................
Bleu_1 0.7873374869078789
Bleu_2 0.6976259288994993
Bleu_3 0.6187557419408616
Bleu_4 0.5447695935277843
METEOR 0.3592627977626835
ROUGE_L 0.7196552610795297
CIDEr 1.29650190144478

Epochs since last improvement: 17

The best BLEU-4 score: 0.5952

03-25 20 : 47 : 54
Epoch: 32/40 step: 0/267 Loss: 1.2386 AVG_Loss: 1.2386 Top-5 Accuracy: 65.2641 Batch_time: 3.2536s
Epoch: 32/40 step: 50/267 Loss: 1.1533 AVG_Loss: 1.1347 Top-5 Accuracy: 65.9574 Batch_time: 0.6625s
Epoch: 32/40 step: 100/267 Loss: 1.1088 AVG_Loss: 1.1432 Top-5 Accuracy: 66.8971 Batch_time: 0.6596s
Epoch: 32/40 step: 150/267 Loss: 0.9800 AVG_Loss: 1.1421 Top-5 Accuracy: 71.0983 Batch_time: 0.6595s
Epoch: 32/40 step: 200/267 Loss: 1.2172 AVG_Loss: 1.1420 Top-5 Accuracy: 66.9894 Batch_time: 0.6591s
Epoch: 32/40 step: 250/267 Loss: 1.0239 AVG_Loss: 1.1403 Top-5 Accuracy: 67.7362 Batch_time: 0.6612s
eval_time: 56.88075661659241
avg eval_time: 0.04267123527126212
fps: 23.434990659233904
extractor_time: 0.011787570515523168
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9761731096520845, 'Bleu_2': 0.9722678933308481, 'Bleu_3': 0.969276919179077, 'Bleu_4': 0.9664250329107448, 'METEOR': 0.7679633811719386, 'ROUGE_L': 0.9833192231274995, 'CIDEr': 0.0}
nochange_acc: 0.9774774774774775
change_metric: {'Bleu_1': 0.6870894865856068, 'Bleu_2': 0.5507516140327566, 'Bleu_3': 0.4388093352678549, 'Bleu_4': 0.3444188430410468, 'METEOR': 0.2336681934963362, 'ROUGE_L': 0.4707250618113645, 'CIDEr': 0.5235043307403846}
change_acc: 0.7916041979010495
.......................................................
Bleu_1 0.8053588175890689
Bleu_2 0.722028664034246
Bleu_3 0.6521698589875256
Bleu_4 0.5901379482530159
METEOR 0.3761649266036449
ROUGE_L 0.7268298715912189
CIDEr 1.3084103292366471

Epochs since last improvement: 18

The best BLEU-4 score: 0.5952

[Decay LR] The new learning rate is 0.000006

03-25 20 : 52 : 25
Epoch: 33/40 step: 0/267 Loss: 1.1229 AVG_Loss: 1.1229 Top-5 Accuracy: 66.5484 Batch_time: 2.1678s
Epoch: 33/40 step: 50/267 Loss: 1.0702 AVG_Loss: 1.1242 Top-5 Accuracy: 68.4400 Batch_time: 0.6635s
Epoch: 33/40 step: 100/267 Loss: 1.1209 AVG_Loss: 1.1251 Top-5 Accuracy: 67.2367 Batch_time: 0.6572s
Epoch: 33/40 step: 150/267 Loss: 1.2411 AVG_Loss: 1.1328 Top-5 Accuracy: 62.5325 Batch_time: 0.6609s
Epoch: 33/40 step: 200/267 Loss: 1.0328 AVG_Loss: 1.1348 Top-5 Accuracy: 69.1960 Batch_time: 0.6670s
Epoch: 33/40 step: 250/267 Loss: 1.0527 AVG_Loss: 1.1386 Top-5 Accuracy: 68.4677 Batch_time: 0.6723s
eval_time: 55.67078733444214
avg eval_time: 0.04176353138367753
fps: 23.944335329622792
extractor_time: 0.011773424048398727
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9616319913534284, 'Bleu_2': 0.9554252901390751, 'Bleu_3': 0.9501332612709519, 'Bleu_4': 0.9444935599276146, 'METEOR': 0.7472341459757094, 'ROUGE_L': 0.9761624452760814, 'CIDEr': 0.0}
nochange_acc: 0.96996996996997
change_metric: {'Bleu_1': 0.7066749313843602, 'Bleu_2': 0.5781672537708678, 'Bleu_3': 0.4693224725802135, 'Bleu_4': 0.376170293151587, 'METEOR': 0.24138305142006003, 'ROUGE_L': 0.48339466429609906, 'CIDEr': 0.578418711044469}
change_acc: 0.8275862068965517
.......................................................
Bleu_1 0.807288208108867
Bleu_2 0.7235066492891022
Bleu_3 0.6496703561615463
Bleu_4 0.5808439394220818
METEOR 0.3702222252720775
ROUGE_L 0.7295937206596912
CIDEr 1.3258490435434178

Epochs since last improvement: 19

The best BLEU-4 score: 0.5952

03-25 20 : 56 : 52
Epoch: 34/40 step: 0/267 Loss: 1.1474 AVG_Loss: 1.1474 Top-5 Accuracy: 66.9484 Batch_time: 2.1001s
Epoch: 34/40 step: 50/267 Loss: 1.0143 AVG_Loss: 1.1329 Top-5 Accuracy: 68.9189 Batch_time: 0.6686s
Epoch: 34/40 step: 100/267 Loss: 1.1447 AVG_Loss: 1.1321 Top-5 Accuracy: 66.0885 Batch_time: 0.6659s
Epoch: 34/40 step: 150/267 Loss: 1.1758 AVG_Loss: 1.1334 Top-5 Accuracy: 66.9787 Batch_time: 0.6580s
Epoch: 34/40 step: 200/267 Loss: 1.2021 AVG_Loss: 1.1340 Top-5 Accuracy: 66.0069 Batch_time: 0.6517s
Epoch: 34/40 step: 250/267 Loss: 1.1043 AVG_Loss: 1.1359 Top-5 Accuracy: 67.5918 Batch_time: 0.6787s
eval_time: 59.24438190460205
avg eval_time: 0.044444397527833496
fps: 22.500023751559365
extractor_time: 0.011728695077459703
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9623406720738812, 'Bleu_2': 0.9547585574081164, 'Bleu_3': 0.9479350418990086, 'Bleu_4': 0.9402249882894719, 'METEOR': 0.7409690920093172, 'ROUGE_L': 0.9778968201261948, 'CIDEr': 0.0}
nochange_acc: 0.9714714714714715
change_metric: {'Bleu_1': 0.6790551084792331, 'Bleu_2': 0.5502009406992081, 'Bleu_3': 0.43988403703934364, 'Bleu_4': 0.34956370179221413, 'METEOR': 0.232539177421767, 'ROUGE_L': 0.45964477939480797, 'CIDEr': 0.5352926209493543}
change_acc: 0.7991004497751124
.......................................................
Bleu_1 0.7849706614783969
Bleu_2 0.6970473948521093
Bleu_3 0.6178985656521175
Bleu_4 0.5447461245297845
METEOR 0.3586061149470601
ROUGE_L 0.7185764066469488
CIDEr 1.299729015108129

Epochs since last improvement: 20

The best BLEU-4 score: 0.5952

03-25 21 : 01 : 24
Epoch: 35/40 step: 0/267 Loss: 1.1304 AVG_Loss: 1.1304 Top-5 Accuracy: 67.1832 Batch_time: 3.4781s
Epoch: 35/40 step: 50/267 Loss: 1.0071 AVG_Loss: 1.1181 Top-5 Accuracy: 68.9139 Batch_time: 0.6595s
Epoch: 35/40 step: 100/267 Loss: 1.1774 AVG_Loss: 1.1233 Top-5 Accuracy: 67.5258 Batch_time: 0.6584s
Epoch: 35/40 step: 150/267 Loss: 0.9442 AVG_Loss: 1.1220 Top-5 Accuracy: 71.6204 Batch_time: 0.6703s
Epoch: 35/40 step: 200/267 Loss: 1.1376 AVG_Loss: 1.1242 Top-5 Accuracy: 67.5699 Batch_time: 0.6528s
Epoch: 35/40 step: 250/267 Loss: 1.0820 AVG_Loss: 1.1301 Top-5 Accuracy: 67.8010 Batch_time: 0.6807s
eval_time: 62.78432750701904
avg eval_time: 0.04710002063542314
fps: 21.231413203414114
extractor_time: 0.011654743465014356
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9495677233426658, 'Bleu_2': 0.9394229659352441, 'Bleu_3': 0.9303734423889418, 'Bleu_4': 0.9200974987676765, 'METEOR': 0.7202239152541831, 'ROUGE_L': 0.9712154095415382, 'CIDEr': 0.0}
nochange_acc: 0.9624624624624625
change_metric: {'Bleu_1': 0.69475048916125, 'Bleu_2': 0.5651828612858211, 'Bleu_3': 0.451640721704936, 'Bleu_4': 0.35412721592615337, 'METEOR': 0.24258724079993227, 'ROUGE_L': 0.47407173111682227, 'CIDEr': 0.5708240731135038}
change_acc: 0.8320839580209896
.......................................................
Bleu_1 0.7906205241699003
Bleu_2 0.7002337805067826
Bleu_3 0.6173819844041009
Bleu_4 0.5385486549643781
METEOR 0.36066419760045804
ROUGE_L 0.7224570948308963
CIDEr 1.3122559656569037

Epochs since last improvement: 21

The best BLEU-4 score: 0.5952

[Decay LR] The new learning rate is 0.000004

03-25 21 : 06 : 00
Epoch: 36/40 step: 0/267 Loss: 1.1109 AVG_Loss: 1.1109 Top-5 Accuracy: 69.6339 Batch_time: 2.3023s
Epoch: 36/40 step: 50/267 Loss: 1.1468 AVG_Loss: 1.1138 Top-5 Accuracy: 66.9675 Batch_time: 0.6639s
Epoch: 36/40 step: 100/267 Loss: 1.2874 AVG_Loss: 1.1210 Top-5 Accuracy: 62.9260 Batch_time: 0.6600s
Epoch: 36/40 step: 150/267 Loss: 1.1379 AVG_Loss: 1.1200 Top-5 Accuracy: 67.1403 Batch_time: 0.6608s
Epoch: 36/40 step: 200/267 Loss: 1.1517 AVG_Loss: 1.1286 Top-5 Accuracy: 65.7989 Batch_time: 0.6453s
Epoch: 36/40 step: 250/267 Loss: 1.0365 AVG_Loss: 1.1303 Top-5 Accuracy: 69.4085 Batch_time: 0.6714s
eval_time: 57.1152126789093
avg eval_time: 0.04284712128950435
fps: 23.33879079631321
extractor_time: 0.011771541739976773
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9527785775983436, 'Bleu_2': 0.9428498467591366, 'Bleu_3': 0.9347402002870313, 'Bleu_4': 0.9256697390674183, 'METEOR': 0.7197631847219382, 'ROUGE_L': 0.9700928679439399, 'CIDEr': 0.0}
nochange_acc: 0.960960960960961
change_metric: {'Bleu_1': 0.7123009352252534, 'Bleu_2': 0.5876769579854345, 'Bleu_3': 0.47568957900103487, 'Bleu_4': 0.37908439629590945, 'METEOR': 0.24813835737470832, 'ROUGE_L': 0.48938565811689144, 'CIDEr': 0.5917951233183938}
change_acc: 0.8305847076461769
.......................................................
Bleu_1 0.8030268567202432
Bleu_2 0.7166101925039206
Bleu_3 0.6359187031837249
Bleu_4 0.5588935094632688
METEOR 0.3658545657984792
ROUGE_L 0.7295589527491602
CIDEr 1.325443361318001

Epochs since last improvement: 22

The best BLEU-4 score: 0.5952

03-25 21 : 10 : 29
Epoch: 37/40 step: 0/267 Loss: 1.1360 AVG_Loss: 1.1360 Top-5 Accuracy: 66.4253 Batch_time: 3.1475s
Epoch: 37/40 step: 50/267 Loss: 1.1644 AVG_Loss: 1.1065 Top-5 Accuracy: 67.3027 Batch_time: 0.6688s
Epoch: 37/40 step: 100/267 Loss: 1.0842 AVG_Loss: 1.1164 Top-5 Accuracy: 68.2091 Batch_time: 0.6563s
Epoch: 37/40 step: 150/267 Loss: 1.0836 AVG_Loss: 1.1193 Top-5 Accuracy: 68.0997 Batch_time: 0.6592s
Epoch: 37/40 step: 200/267 Loss: 1.1391 AVG_Loss: 1.1215 Top-5 Accuracy: 64.7527 Batch_time: 0.6633s
Epoch: 37/40 step: 250/267 Loss: 1.2218 AVG_Loss: 1.1216 Top-5 Accuracy: 64.6194 Batch_time: 0.6724s
eval_time: 71.41445922851562
avg eval_time: 0.05357423798088194
fps: 18.665687794884768
extractor_time: 0.011767074745218287
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9828877005344966, 'Bleu_2': 0.9798908103605528, 'Bleu_3': 0.9774515941107417, 'Bleu_4': 0.9750430685940868, 'METEOR': 0.788709036150301, 'ROUGE_L': 0.9885398604759369, 'CIDEr': 0.0}
nochange_acc: 0.984984984984985
change_metric: {'Bleu_1': 0.6976386066257712, 'Bleu_2': 0.5640005576359985, 'Bleu_3': 0.45222006667249015, 'Bleu_4': 0.35949309912000943, 'METEOR': 0.23554079848393195, 'ROUGE_L': 0.47233840887219963, 'CIDEr': 0.5393575730736131}
change_acc: 0.7781109445277361
.......................................................
Bleu_1 0.8081461825980422
Bleu_2 0.7230528535865117
Bleu_3 0.6491945361770826
Bleu_4 0.5823872318799831
METEOR 0.37003317280825526
ROUGE_L 0.7302455107237293
CIDEr 1.3224326730755513

Epochs since last improvement: 23

The best BLEU-4 score: 0.5952

03-25 21 : 15 : 14
Epoch: 38/40 step: 0/267 Loss: 1.1993 AVG_Loss: 1.1993 Top-5 Accuracy: 63.4087 Batch_time: 3.1775s
Epoch: 38/40 step: 50/267 Loss: 1.1771 AVG_Loss: 1.1103 Top-5 Accuracy: 64.2036 Batch_time: 0.6566s
Epoch: 38/40 step: 100/267 Loss: 1.2023 AVG_Loss: 1.1216 Top-5 Accuracy: 65.5114 Batch_time: 0.6819s
Epoch: 38/40 step: 150/267 Loss: 1.1132 AVG_Loss: 1.1189 Top-5 Accuracy: 67.7074 Batch_time: 0.6611s
Epoch: 38/40 step: 200/267 Loss: 1.1133 AVG_Loss: 1.1190 Top-5 Accuracy: 66.6964 Batch_time: 0.6622s
Epoch: 38/40 step: 250/267 Loss: 1.0936 AVG_Loss: 1.1203 Top-5 Accuracy: 68.4066 Batch_time: 0.6746s
eval_time: 53.6063551902771
avg eval_time: 0.04021482009773226
fps: 24.86645464457494
extractor_time: 0.011668768487116133
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.95775862068938, 'Bleu_2': 0.9486296916034686, 'Bleu_3': 0.9413334037394643, 'Bleu_4': 0.933293034931305, 'METEOR': 0.7245086203053683, 'ROUGE_L': 0.9732812811268096, 'CIDEr': 0.0}
nochange_acc: 0.9654654654654654
change_metric: {'Bleu_1': 0.7047327241968291, 'Bleu_2': 0.5806034665311334, 'Bleu_3': 0.4714397292563749, 'Bleu_4': 0.37958759673840087, 'METEOR': 0.24992623498709815, 'ROUGE_L': 0.4920072453903939, 'CIDEr': 0.5869915488763902}
change_acc: 0.8200899550224887
.......................................................
Bleu_1 0.7996366964781544
Bleu_2 0.7142229273049103
Bleu_3 0.63592822093239
Bleu_4 0.5624481651774699
METEOR 0.36955675499059487
ROUGE_L 0.7324637403644771
CIDEr 1.3284005185404613

Epochs since last improvement: 24

The best BLEU-4 score: 0.5952

[Decay LR] The new learning rate is 0.000003

03-25 21 : 19 : 40
Epoch: 39/40 step: 0/267 Loss: 1.2754 AVG_Loss: 1.2754 Top-5 Accuracy: 62.3946 Batch_time: 2.5327s
Epoch: 39/40 step: 50/267 Loss: 1.0578 AVG_Loss: 1.1008 Top-5 Accuracy: 69.0053 Batch_time: 0.6597s
Epoch: 39/40 step: 100/267 Loss: 1.0275 AVG_Loss: 1.1069 Top-5 Accuracy: 69.9454 Batch_time: 0.6601s
Epoch: 39/40 step: 150/267 Loss: 1.1859 AVG_Loss: 1.1123 Top-5 Accuracy: 64.8116 Batch_time: 0.6625s
Epoch: 39/40 step: 200/267 Loss: 1.0486 AVG_Loss: 1.1152 Top-5 Accuracy: 67.8108 Batch_time: 0.6624s
Epoch: 39/40 step: 250/267 Loss: 1.1326 AVG_Loss: 1.1195 Top-5 Accuracy: 65.9758 Batch_time: 0.6870s
eval_time: 63.513439416885376
avg eval_time: 0.04764699131049165
fps: 20.987684059282973
extractor_time: 0.012323453742940893
len(nochange_references): 666
len(change_references): 667
nochange_metric: {'Bleu_1': 0.9787114845935634, 'Bleu_2': 0.9747125865612547, 'Bleu_3': 0.9715137220126924, 'Bleu_4': 0.9679606307669353, 'METEOR': 0.7703595509296288, 'ROUGE_L': 0.9860780792425783, 'CIDEr': 0.0}
nochange_acc: 0.9804804804804805
change_metric: {'Bleu_1': 0.7057016872429203, 'Bleu_2': 0.5777240440648889, 'Bleu_3': 0.46879539555370014, 'Bleu_4': 0.37727025207643783, 'METEOR': 0.24117718773606753, 'ROUGE_L': 0.4826623182830585, 'CIDEr': 0.561190816664411}
change_acc: 0.7886056971514243
.......................................................
Bleu_1 0.8097171841395627
Bleu_2 0.7263837190817682
Bleu_3 0.6523388040541583
Bleu_4 0.5835474459647642
METEOR 0.37143909159840216
ROUGE_L 0.7341813706454292
CIDEr 1.3303131915857669

Epochs since last improvement: 25

The best BLEU-4 score: 0.5952

[mrazhou]

你加我微信吧。

[Dingxiangtao]

怎么加啊？

[mrazhou]

你发个消息到我邮箱，我回你微信号。github好像传不了二维码。 ***@***.***

[mrazhou]

mrazhou@foxmail.com