PaddlePaddle / Paddle

PArallel Distributed Deep LEarning: Machine Learning Framework from Industrial Practice (『飞桨』核心框架,深度学习&机器学习高性能单机、分布式训练和跨平台部署)
http://www.paddlepaddle.org/
Apache License 2.0
22.29k stars 5.61k forks source link

在场景文字识别中,在训练事件中评估结果没有输出。 #11004

Closed yeyupiaoling closed 6 years ago

yeyupiaoling commented 6 years ago

我使用百度云的深度学习集群跑场景文字识别时,在训练过程中,没有输出评估器的结果:https://github.com/PaddlePaddle/models/blob/36e3fcd578e0e469d91b1257c27d627495c0b06b/scene_text_recognition/train.py#L75-L77

因为服务器没有warp-ctc,使用了百度开源的warp-ctc,地址如下:https://github.com/baidu-research/warp-ctc

输出结果如下:

Pass 0, batch 0, Samples 0, Cost 42.609291, Eval {}
...................................................................................................
Pass 0, batch 100, Samples 12800, Cost 0.494285, Eval {}
...................................................................................................
Pass 0, batch 200, Samples 25600, Cost 0.142084, Eval {}
...............
Test 0, Cost 0.105199, Eval {}
chengduoZH commented 6 years ago

你先检测一下模型的配置,在单机上运行有输出吗?

yeyupiaoling commented 6 years ago

@chengduoZH 因为我本地没有GPU,无法进行测试。

yeyupiaoling commented 6 years ago

@chengduoZH 可能是我使用了百度开源的warp-ctc,有没其他的warp-ctc?