eval mode has bad performance 测试与训练模式性能差别很大

[woodyyyyyyyy]

during training, the train accuracy was 96% or higher, the eval accuracy was only 10%. Finally,for a single picture, the train mode will predict the length well even though it shouldn't be accurate. However, for this picture , the eval mode will only predict one character maybe not. I use nn.CTCLoss, and my char_set has numbers, letters, and 6 Chinese characters. 训练模式有很高准确率，验证模式准确率很低。测试一张图片无论是验证集还是测试集，训练模式最起码位数差不多，但是验证模式只能有一两个预测。

[woodyyyyyyyy]

感谢回复，thanks

[chnzhero]

运行环境是linux还是win啊请问大佬

[woodyyyyyyyy]

萌新用的是win