2.2 之后损失没有改善

2019-10-09 14:01:09,550 -- INFO -- Iter 981000: loss 2.2273828502893447, time 0.08978772163391113 2019-10-09 14:03:29,982 -- INFO -- Iter 982000: loss 2.251806115269661, time 0.1407618522644043 2019-10-09 14:05:51,542 -- INFO -- Iter 983000: loss 2.2446057442426683, time 0.2159106731414795 2019-10-09 14:08:11,021 -- INFO -- Iter 984000: loss 2.2406707513332367, time 0.09187722206115723 2019-10-09 14:10:33,569 -- INFO -- Iter 985000: loss 2.243191688776016, time 0.28462576866149 2019-10-09 14:12:52,848 -- INFO -- Iter 986000: loss 2.229316849589348, time 0.09166932106018066 2019-10-09 14:15:12,966 -- INFO -- Iter 987000: loss 2.238621209859848, time 0.0910036563873291 2019-10-09 14:17:32,610 -- INFO -- Iter 988000: loss 2.2455421295166014, time 0.09028363227844238 2019-10-09 14:19:50,839 -- INFO -- Iter 989000: loss 2.2432832870483397, time 0.34220433235168457 2019-10-09 14:22:06,687 -- INFO -- Iter 990000: loss 2.2466694812774657, time 0.09603047370910645 2019-10-09 14:22:15,398 -- INFO -- Test bucket 0: loss 3.1786046028137207, time 0.05455660820007324 2019-10-09 14:22:15,458 -- INFO -- Test bucket 1: loss 3.5405633449554443, time 0.05947542190551758 2019-10-09 14:22:15,532 -- INFO -- Test bucket 2: loss 3.5922553539276123, time 0.07441473007202148 2019-10-09 14:22:15,633 -- INFO -- Test bucket 3: loss 3.529684543609619, time 0.10061907768249512 2019-10-09 14:22:15,755 -- INFO -- Test bucket 4: loss 3.7335658073425293, time 0.12195014953613281 2019-10-09 14:22:15,903 -- INFO -- Test bucket 5: loss 3.8613126277923584, time 0.14754486083984375 2019-10-09 14:24:21,939 -- INFO -- Iter 991000: loss 2.2421313049793246, time 0.08835697174072266 2019-10-09 14:26:39,390 -- INFO -- Iter 992000: loss 2.2535737413167953, time 0.16527605056762695 2019-10-09 14:29:01,609 -- INFO -- Iter 993000: loss 2.263053869485855, time 0.1388380527496338 2019-10-09 14:31:20,481 -- INFO -- Iter 994000: loss 2.2561977257728576, time 0.08809399604797363 2019-10-09 14:33:43,944 -- INFO -- Iter 995000: loss 2.2676013087034224, time 0.09026694297790527 2019-10-09 14:36:02,296 -- INFO -- Iter 996000: loss 2.2528336789608003, time 0.09229564666748047 2019-10-09 14:38:20,023 -- INFO -- Iter 997000: loss 2.2519494262933732, time 0.08894729614257812 2019-10-09 14:40:37,962 -- INFO -- Iter 998000: loss 2.266415533065796, time 0.08812499046325684 2019-10-09 14:42:58,623 -- INFO -- Iter 999000: loss 2.2743323941230775, time 0.09042882919311523 2019-10-09 14:45:17,886 -- INFO -- Iter 1000000: loss 2.2651904397010805, time 0.08829283714294434