dead in train

[buaacyh]

你好xingyi，我尝试用你的pose-hg-2d进行训练mpii数据集，我把所有的config都配置好了，数据也下载了。前期导入都没有问题，但是每次都在train.py中的第21行出bug，“for i, (input, target, meta) in enumerate(dataLoader):”。我在这句的后面加print函数，都不能显示出来。后来我在MPii.py的getitem()，LoadImage()和GetPartInfo()加了一些print，结果如下： (env_py27) ccc@bj:~/case/hourglass/pytorch-pose-hg-2d/src$ python main.py -expID 2D -GPU 3 ~/anaconda3/envs/env_py27/lib/python2.7/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype fromfloattonp.floatingis deprecated. In future, it will be treated asnp.float64 == np.dtype(float).type`. from ._conv import register_converters as _register_converters

Using GPU 3

==> initializing 2D val data.

Loaded 2D val 2958 samples

==> initializing 2D train data.

Loaded 2D train 22246 samples

('len(dataLoader) is %d', 3708)

prepare to in dataLoader

~/case/hourglass/Datasets/mpii/images/085460756.jpg 9384 pts.shape is (16, 2) 9384 c.shape is (2,) 9384 s.shape is () self.split is train

~/case/hourglass/Datasets/mpii/images/019372879.jpg 18030 pts.shape is (16, 2) 18030 c.shape is (2,) 18030 s.shape is () self.split is train` 每次都只显示2张或4张图片信息，不管我设置trainBatch是1还是5还是10。曾经有两次我通过了“for i, (input, target, meta) in enumerate(dataLoader):”，但是在后面停止，不能训练了。不知道你之前遇到过这个问题吗？ 谢谢，不胜感激！

[xingyizhou]

Hi, It seems you are using Anaconda3, while our code has only been tested on Python2. And I can't find the bug in your logs.

[buaacyh]

Thank you very much! It's that problem! When I changed the conda environment, everything is going well. Thanks again, it has troubled me for days. Thanks again.