h5py objects cannot be pickled

[lander1003]

你好，我在运行mainpro_CK+和mainpro_FER两个文件时均出现了h5py objects cannot be pickled这个报错情况，报错语句为这样for batch_idx, (inputs, targets) in enumerate(trainloader):，请问下怎么办啊，万分感谢

[WuJie1010]

您是否采用了python3,如果是的话，请按照py3的要求做相应调整

[lander1003]

好的，那如果用python2，用pytorch现在最新版本可以吗

[lander1003]

现在python2好像安装不了pytorch了，然而python3又不知道咋改，小白表示太难了

[victor-mageto]

@lander1003 did you get the solution, am also facing the same error.

[cimszw]

@lander1003 @victor-mageto It happens because h5py won't read from multiple processes. By omitting num_workers, you're setting it to the default of 0, which uses only the main process. You could also set it to 1 (on Linux, at least). So try to set :

1. num_workers=0.
2. datafile = h5py.File(datapath, 'w', swmr=True) it works for me.

[victor-mageto]

where did you define the datapath?

[lander1003]

@lander1003 did you get the solution, am also facing the same error.

I met the error when I use python3, so I change to python2

[lander1003]

@lander1003 @victor-mageto It happens because h5py won't read from multiple processes. By omitting num_workers, you're setting it to the default of 0, which uses only the main process. You could also set it to 1 (on Linux, at least). So try to set :

1. num_workers=0.
2. datafile = h5py.File(datapath, 'w', swmr=True) it works for me.

Can have your wechat ID? I really cannot solve it by myself, I'm really new in ML.Thank you!

[y894577]

我也是这个问题，有什么办法不回退到python2.7来解决这个问题么

[lander1003]

我也是这个问题，有什么办法不回退到python2.7来解决这个问题么

我之前就是退回2.7搞得，要是你用python3解决了可以分享一下嘛

[y894577]

我也是这个问题，有什么办法不回退到python2.7来解决这个问题么

我之前就是退回2.7搞得，要是你用python3解决了可以分享一下嘛

回退2.7会import不了torch，不知道是不是官方抛弃了2.7的问题，python3的代码也不知道从哪改，唉

[lander1003]

我也是这个问题，有什么办法不回退到python2.7来解决这个问题么

我之前就是退回2.7搞得，要是你用python3解决了可以分享一下嘛

回退2.7会import不了torch，不知道是不是官方抛弃了2.7的问题，python3的代码也不知道从哪改，唉

回退到python2可能需要在linux上才能安装torch老版本了，我记得当时也是这个问题

[lander1003]

我也是这个问题，有什么办法不回退到python2.7来解决这个问题么

我之前就是退回2.7搞得，要是你用python3解决了可以分享一下嘛

回退2.7会import不了torch，不知道是不是官方抛弃了2.7的问题，python3的代码也不知道从哪改，唉

可以价格wechat交流一下啊

[y894577]

我也是这个问题，有什么办法不回退到python2.7来解决这个问题么

我之前就是退回2.7搞得，要是你用python3解决了可以分享一下嘛

回退2.7会import不了torch，不知道是不是官方抛弃了2.7的问题，python3的代码也不知道从哪改，唉

可以价格wechat交流一下啊

可以啊，wechat和github同账户名

[ZangHuanyu]

@lander1003 @victor-mageto It happens because h5py won't read from multiple processes. By omitting num_workers, you're setting it to the default of 0, which uses only the main process. You could also set it to 1 (on Linux, at least). So try to set :

1. num_workers=0.
2. datafile = h5py.File(datapath, 'w', swmr=True) it works for me.

Could you plz tell me where should I apply "datafile = h5py.File(datapath, 'w', swmr=True)"? Do really appreciated!

[connorxian]

您是否采用了python3,如果是的话，请按照py3的要求做相应调整 请问是做怎么样的调整呢

[YELLOWCOOLS]

datafile = h5py.File(datapath, 'w') ，witch is not used in python3. You should try use 'pickle' to save data,like this: """ train_file=open('data/fer2013/train.txt','wb') pickle.dump(Training_x,train_file) """

[DaobinZhu]

代码没有人维护了吗。。我遇到了相同的问题