More complete the README?(3D Blocks World)

[LHWsleeve]

Hi I am trying to run the source code of this paper.（Learning Interpretable Spatial Operations in a Rich 3D Blocks World）, but I have encountered many problems. If you have time, could you complete the README? More detailed description of the environment required to run the code. THANKS

such as bugs, incorrect parameters, etc.

  _File "/home/sleeve/GroundedLanguage/SpatialReasoning/train.py", line 25, in <module>
    Training = data.Data(config, train=True)
TypeError: __init__() got an unexpected keyword argument 'train'_

_File "/home/sleeve/GroundedLanguage/SpatialReasoning/Language_Pipeline.py", line 12, in <module>
    np.set_printoptions(threshold=np.nan)
  File "/home/sleeve/.conda/envs/tf01/lib/python2.7/site-packages/numpy/core/arrayprint.py", line 246, in set_printoptions
    floatmode, legacy)
  File "/home/sleeve/.conda/envs/tf01/lib/python2.7/site-packages/numpy/core/arrayprint.py", line 93, in _make_options_dict
    raise ValueError("threshold must be numeric and non-NAN, try "
ValueError: threshold must be numeric and non-NAN, try sys.maxsize for untruncated representation_

[ybisk]

Hi,

Can you try 753072850d05ffe1669bec8c105b289e00f8bd95 and see if this fixes things for you. It appears to be running when I try it with TF 1.14 in python2.7

Thanks

[LHWsleeve]

Thank you very much for your reply and help, and I got the training results.

And, there are some questions I would like to ask. 0.The most important question: Does the code only consider the start and end points? Do you consider the trajectory problem? What if I encounter obstacles while moving the block? (I don't seem to see the moving track code for the block).

1.simple_eval is the configuration file for verification and testing?

2. I tried to run the simple_eval file. I don't understand the result. Could you explain it?

I don't know what the numbers in the first three columns mean. Why not output location information like a training set (perhaps my level is not good enough to find information)?.

I am not sure what (a, p, g) represents in the coordinates?

And is the instruction behind (a, p, g) a predicted moving block instruction?

The code part is as follows：

Thanks again for your help, I am sorry if I disturb you.

------------------ 原始邮件 ------------------ 发件人: "Yonatan Bisk"notifications@github.com; 发送时间: 2019年6月27日(星期四) 上午6:05 收件人: "ybisk/GroundedLanguage"GroundedLanguage@noreply.github.com; 抄送: "1421147715"1421147715@qq.com;"Author"author@noreply.github.com; 主题: Re: [ybisk/GroundedLanguage] More complete the README?(3D BlocksWorld) (#5)

Hi,

Can you try 7530728 and see if this fixes things for you. It appears to be running when I try it with TF 1.14 in python2.7

Thanks

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.

[ybisk]

0. It is just predicting a target. If you're interested in a trajectory, I recommend looking into Dipendra's work -- e.g. https://github.com/lil-lab/ciff

1. Yes, this can be used for both, just train and validate on dev and then load and predict on test

2. I think there were meant to be screenshots but I don't see any so I'm making a guess here but

https://github.com/ybisk/GroundedLanguage/blob/master/SpatialReasoning/Language_Pipeline.py#L228 --> distances, angles, utterances, gold target, predicted target

sorry for the terse variable names!