donglinAI

Results 7 comments of donglinAI

这个问题是跟多GPU训练相关,如果是单块GPU训练的模型,需要将eval.py 20行附近的 # model = torch.nn.DataParallel(model)注释;如果是多块GPU训练,则需要去掉该行的注释。

@Dkili I find that tensorflow==1.4.1 (ubuntu16.04) is ok. Use `conda install tensorflow-gpu==1.4.1`. By commond `conda search tensorflow-gpu `, don't find 1.4.0 version , only have three version (1.3.0、1.4.1、1.5.0 ) nearby....

@Lemonyue I found that after downloading this project, the project'name is 'aster-master' instead of 'aster', but 'aster' in *.proto . So modify project'name to 'aster', and then run "protoc aster/protoc/*.proto...

The project need the successful installation of crnn, and the crnn's success needs the torch (isn't pytorch). So the solvement is to ellipsis the use of crnn

Could you share the code of generating Pyramid label?

@JingChaoLiu Thank you very much, this will give a lot of help, you're welcome! Best regards!

I find that if delete already existing logs , the problem that lhs shape not equal rhs shape will be solved @HoracceFeng