怎么预测出来的概率全是0 啊

[paohaijiao]

===============Eval a batch=======================
the step 6801.0 test accuracy: 0.0
===============Eval a batch=======================
the step 6802.0 takes 5.253191947937012 loss 0.7621740102767944
the step 6803.0 takes 5.179174900054932 loss 0.836201548576355
the step 6804.0 takes 5.21918511390686 loss 0.7611073851585388
the step 6805.0 takes 5.158170461654663 loss 0.7541408538818359
the step 6806.0 takes 5.20117974281311 loss 0.7468845844268799
the step 6807.0 takes 5.162172317504883 loss 0.827800989151001
the step 6808.0 takes 5.18017578125 loss 0.7844552397727966
the step 6809.0 takes 5.195178985595703 loss 0.7899702191352844
the step 6810.0 takes 5.188177824020386 loss 0.8137649297714233
the step 6811.0 takes 5.229186773300171 loss 0.7540019750595093
the step 6812.0 takes 5.173173904418945 loss 0.7108616828918457
the step 6813.0 takes 5.21818470954895 loss 0.785955548286438
the step 6814.0 takes 5.249190807342529 loss 0.7812217473983765
the step 6815.0 takes 5.238188028335571 loss 0.8018498420715332
the step 6816.0 takes 5.2111828327178955 loss 0.8020637631416321
the step 6817.0 takes 5.189177513122559 loss 0.8187956809997559
the step 6818.0 takes 5.226185083389282 loss 0.6946130990982056
the step 6819.0 takes 5.19417929649353 loss 0.8060418367385864
the step 6820.0 takes 5.136166334152222 loss 0.7476954460144043
the step 6821.0 takes 5.186176776885986 loss 0.7232500910758972
the step 6822.0 takes 5.16417121887207 loss 0.7656339406967163
the step 6823.0 takes 5.187177419662476 loss 0.7978050112724304
the step 6824.0 takes 5.1751744747161865 loss 0.7152007818222046
the step 6825.0 takes 5.178175210952759 loss 0.822441041469574
the step 6826.0 takes 5.1611714363098145 loss 0.851509690284729
the step 6827.0 takes 5.2411887645721436 loss 0.7786080241203308
the step 6828.0 takes 5.161171913146973 loss 0.8658627271652222
the step 6829.0 takes 5.20918345451355 loss 0.8724192380905151
the step 6830.0 takes 5.199178695678711 loss 0.7804743051528931
the step 6831.0 takes 5.165170669555664 loss 0.7231662273406982
the step 6832.0 takes 5.1601715087890625 loss 0.7696665525436401
the step 6833.0 takes 5.186176776885986 loss 0.7821593284606934
the step 6834.0 takes 5.163172721862793 loss 0.8162339329719543
the step 6835.0 takes 5.160170316696167 loss 0.8535160422325134
the step 6836.0 takes 5.14216685295105 loss 0.8528579473495483
the step 6837.0 takes 5.1971800327301025 loss 0.717275857925415
the step 6838.0 takes 5.241189479827881 loss 0.6899185180664062
the step 6839.0 takes 5.163172245025635 loss 0.7475823163986206
the step 6840.0 takes 5.212183237075806 loss 0.7100281715393066
the step 6841.0 takes 5.185176372528076 loss 0.7437551021575928
the step 6842.0 takes 5.214183568954468 loss 0.8610426187515259
the step 6843.0 takes 5.196179628372192 loss 0.8723220825195312
the step 6844.0 takes 5.184176206588745 loss 0.7162365913391113
the step 6845.0 takes 5.203181028366089 loss 0.7437829971313477
the step 6846.0 takes 5.197179079055786 loss 0.7429450154304504
the step 6847.0 takes 5.218184232711792 loss 0.7336486577987671
the step 6848.0 takes 5.2151830196380615 loss 0.7843906283378601
the step 6849.0 takes 5.240188837051392 loss 0.7857959270477295
the step 6850.0 takes 5.143167972564697 loss 0.7797924280166626
the step 6851.0 takes 5.222185134887695 loss 0.7834088802337646

[hellobug-fei]

相同问题

[zhaojc001]

===============Eval a batch=======================
I0809 01:34:06.247253 139943277836096 Chinese_OCR.py:228] ===============Eval a batch=======================
the step 15901.0 test accuracy: 0.0
I0809 01:34:06.247759 139943277836096 Chinese_OCR.py:230] the step 15901.0 test accuracy: 0.0
===============Eval a batch=======================
I0809 01:34:06.247858 139943277836096 Chinese_OCR.py:231] ===============Eval a batch=======================
the step 15902.0 takes 2.14595580101 loss 0.472916126251
I0809 01:34:08.393987 139943277836096 Chinese_OCR.py:215] the step 15902.0 takes 2.14595580101 loss 0.472916126251
the step 15903.0 takes 2.17172503471 loss 0.442270755768
I0809 01:34:10.566279 139943277836096 Chinese_OCR.py:215] the step 15903.0 takes 2.17172503471 loss 0.442270755768
the step 15904.0 takes 2.13369894028 loss 0.459591686726
I0809 01:34:12.700567 139943277836096 Chinese_OCR.py:215] the step 15904.0 takes 2.13369894028 loss 0.459591686726
the step 15905.0 takes 2.14362502098 loss 0.452670156956
I0809 01:34:14.844719 139943277836096 Chinese_OCR.py:215] the step 15905.0 takes 2.14362502098 loss 0.452670156956
the step 15906.0 takes 2.18492293358 loss 0.407150357962
I0809 01:34:17.030210 139943277836096 Chinese_OCR.py:215] the step 15906.0 takes 2.18492293358 loss 0.407150357962
the step 15907.0 takes 2.17844486237 loss 0.439693808556
I0809 01:34:19.209281 139943277836096 Chinese_OCR.py:215] the step 15907.0 takes 2.17844486237 loss 0.439693808556
the step 15908.0 takes 2.28083300591 loss 0.459465205669
I0809 01:34:21.490731 139943277836096 Chinese_OCR.py:215] the step 15908.0 takes 2.28083300591 loss 0.459465205669
the step 15909.0 takes 2.06452488899 loss 0.453502476215
I0809 01:34:23.555757 139943277836096 Chinese_OCR.py:215] the step 15909.0 takes 2.06452488899 loss 0.453502476215
the step 15910.0 takes 2.10654902458 loss 0.529790282249
I0809 01:34:25.662868 139943277836096 Chinese_OCR.py:215] the step 15910.0 takes 2.10654902458 loss 0.529790282249
the step 15911.0 takes 2.06580090523 loss 0.42560890317
I0809 01:34:27.729223 139943277836096 Chinese_OCR.py:215] the step 15911.0 takes 2.06580090523 loss 0.42560890317
the step 15912.0 takes 2.1419570446 loss 0.420587509871
I0809 01:34:29.871709 139943277836096 Chinese_OCR.py:215] the step 15912.0 takes 2.1419570446 loss 0.420587509871
the step 15913.0 takes 2.15385508537 loss 0.482629299164
I0809 01:34:32.026139 139943277836096 Chinese_OCR.py:215] the step 15913.0 takes 2.15385508537 loss 0.482629299164
the step 15914.0 takes 2.1051170826 loss 0.447850853205
I0809 01:34:34.131865 139943277836096 Chinese_OCR.py:215] the step 15914.0 takes 2.1051170826 loss 0.447850853205
the step 15915.0 takes 2.1156938076 loss 0.467434614897
I0809 01:34:36.248105 139943277836096 Chinese_OCR.py:215] the step 15915.0 takes 2.1156938076 loss 0.467434614897
the step 15916.0 takes 2.11783099174 loss 0.542848229408
I0809 01:34:38.366605 139943277836096 Chinese_OCR.py:215] the step 15916.0 takes 2.11783099174 loss 0.542848229408
the step 15917.0 takes 2.13679885864 loss 0.398986518383
I0809 01:34:40.503909 139943277836096 Chinese_OCR.py:215] the step 15917.0 takes 2.13679885864 loss 0.398986518383
the step 15918.0 takes 2.12822794914 loss 0.517429947853
I0809 01:34:42.632627 139943277836096 Chinese_OCR.py:215] the step 15918.0 takes 2.12822794914 loss 0.517429947853
the step 15919.0 takes 2.14565396309 loss 0.430533587933
I0809 01:34:44.778858 139943277836096 Chinese_OCR.py:215] the step 15919.0 takes 2.14565396309 loss 0.430533587933
the step 15920.0 takes 2.1857509613 loss 0.451907962561
I0809 01:34:46.965162 139943277836096 Chinese_OCR.py:215] the step 15920.0 takes 2.1857509613 loss 0.451907962561
the step 15921.0 takes 2.15857505798 loss 0.44280141592
I0809 01:34:49.124353 139943277836096 Chinese_OCR.py:215] the step 15921.0 takes 2.15857505798 loss 0.44280141592
the step 15922.0 takes 2.15956687927 loss 0.51418286562
I0809 01:34:51.284455 139943277836096 Chinese_OCR.py:215] the step 15922.0 takes 2.15956687927 loss 0.51418286562
the step 15923.0 takes 2.12021303177 loss 0.505261421204
I0809 01:34:53.405194 139943277836096 Chinese_OCR.py:215] the step 15923.0 takes 2.12021303177 loss 0.505261421204
the step 15924.0 takes 2.20005011559 loss 0.48279723525
I0809 01:34:55.605818 139943277836096 Chinese_OCR.py:215] the step 15924.0 takes 2.20005011559 loss 0.48279723525
the step 15925.0 takes 2.11905312538 loss 0.411675512791

[knavezl]

参考#17 方法，重新训练可解决

[mHacks2021]

一样的