用32核的 CPU 跑了几天了, 800 张训练而以

[coolksj]

Instructions for updating: Use standard file APIs to delete files with this prefix. INFO:tensorflow:Step: 400 Time: 5.623 sec/batch, Cost = 1.21803749, BatchSize: 64, Shape[1]: 19 INFO:tensorflow:Step: 500 Time: 5.737 sec/batch, Cost = 0.95884866, BatchSize: 64, Shape[1]: 19 WARNING:tensorflow:From D:\orcs\captcha_trainer\validation.py:55: The name tf.logging.info is deprecated. Please use tf.compat.v1.logging.info instead.

INFO:tensorflow:0 4 4 [12, 31, 35, 1] [28, 14, 11, 35, 37] --> ['b', 'u', 'y', '0'] ['r', 'd', 'a', 'y', '-'] INFO:tensorflow:1 4 3 [25, 23, 31, 35] [20, 23, 16, 37, 37] --> ['o', 'm', 'u', 'y'] ['j', 'm', 'f', '-', '-'] INFO:tensorflow:2 4 4 [21, 26, 21, 28] [35, 27, 16, 13, 37] --> ['k', 'p', 'k', 'r'] ['y', 'q', 'f', 'c', '-'] INFO:tensorflow:3 4 4 [14, 12, 37, 12, 26] [14, 25, 35, 25, 37] --> ['d', 'b', '-', 'b', 'p'] ['d', 'o', 'y', 'o', '-'] INFO:tensorflow:4 4 4 [18, 25, 2, 26] [25, 3, 26, 27, 37] --> ['h', 'o', '1', 'p'] ['o', '2', 'p', 'q', '-'] ERROR:tensorflow:[{"origin": "buy0", "decode": "rday-"}, {"origin": "omuy", "decode": "jmf--"}, {"origin": "kpkr", "decode": "yqfc-"}, {"origin": "db-bp", "decode": "doyo-"}, {"origin": "ho1p", "decode": "o2pq-"}] INFO:tensorflow:Epoch: 42, Step: 500, Accuracy = 0.0000, Cost = 0.95885, Time = 4.327 sec/batch, LearningRate: 1.747833266563248e-05 INFO:tensorflow:Step: 600 Time: 5.744 sec/batch, Cost = 0.90788847, BatchSize: 64, Shape[1]: 19 INFO:tensorflow:Step: 700 Time: 5.696 sec/batch, Cost = 0.75557369, BatchSize: 64, Shape[1]: 19

为什么输出里面 orgin 会有 db-bp 这样的源呢? 根本没有这样的标注啊, 为什么会识别出来 " - " 号呢?

[ghost]

同问，我已经跑了20000times了，还是遥遥无期

[kerlomz]

-是空白占位符请直接无视，其次，根本原因是样本太少无法收敛

[ghost]

-是空白占位符请直接无视，其次，根本原因是样本太少无法收敛

大佬为何我家群不让通过啊

[kerlomz]

-是空白占位符请直接无视，其次，根本原因是样本太少无法收敛

大佬为何我家群不让通过啊

是不是小号加群？