{
  "id": 4411,
  "title": "Benchmark Can Not Run Correctly",
  "url": "/competitions/challenges-in-representation-learning-the-black-box-learning-challenge/discussion/4411",
  "author_name": "",
  "post_date": "2013-04-23T15:20:19.547Z",
  "votes": null,
  "comment_count": 4,
  "views": 1850,
  "content": "<p>All packages installed correctly on CentOS 6.4 x64, when I ran learn-zca.py, it shows:</p>\r\n<pre>/usr/lib/python2.6/site-packages/pylearn2-0.1dev-py2.6.egg/pylearn2/datasets/preprocessing.py:836: UserWarning: This ZCA preprocessor class is known to yield very different results on different platforms. If you plan to conduct experiments with this preprocessing on multiple machines, it is probably a good idea to do the preprocessing on a single machine and copy the preprocessed datasets to the others, rather than preprocessing the data independently in each location.<br>  warnings.warn(&quot;This ZCA preprocessor class is known to yield very &quot;<br>computing zca<br>Traceback (most recent call last):<br>  File &quot;learn_zca.py&quot;, line 10, in &lt;module&gt;<br>    zca.fit(extra.X)<br>  File &quot;/usr/lib/python2.6/site-packages/pylearn2-0.1dev-py2.6.egg/pylearn2/datasets/preprocessing.py&quot;, line 867, in fit<br>    eigs, eigv = linalg.eigh(np.dot(X.T, X) / X.shape[0])<br>MemoryError</pre>\r\n<pre>&nbsp;</pre>\r\n<pre>If I ran the train.py mlp.yaml, it showed</pre>\r\n<pre><br>  File &quot;/usr/lib/python2.6/site-packages/pylearn2-0.1dev-py2.6.egg/pylearn2/config/yaml_parse.py&quot;, line 238, in try_to_import<br>    raise ImportError(base_msg &#43; '. Original exception: '&#43;str(e))<br>ImportError: Could not import pylearn2.scripts.icml_2013_wrepl.black_box but could import pylearn2.scripts.icml_2013_wrepl. Original exception: No module named black_box.black_box_dataset</pre>",
  "messages": [
    {
      "id": "23365",
      "postDate": "04/23/2013 15:20:19",
      "content": "<p>All packages installed correctly on CentOS 6.4 x64, when I ran learn-zca.py, it shows:</p>\r\n<pre>/usr/lib/python2.6/site-packages/pylearn2-0.1dev-py2.6.egg/pylearn2/datasets/preprocessing.py:836: UserWarning: This ZCA preprocessor class is known to yield very different results on different platforms. If you plan to conduct experiments with this preprocessing on multiple machines, it is probably a good idea to do the preprocessing on a single machine and copy the preprocessed datasets to the others, rather than preprocessing the data independently in each location.<br>  warnings.warn(&quot;This ZCA preprocessor class is known to yield very &quot;<br>computing zca<br>Traceback (most recent call last):<br>  File &quot;learn_zca.py&quot;, line 10, in &lt;module&gt;<br>    zca.fit(extra.X)<br>  File &quot;/usr/lib/python2.6/site-packages/pylearn2-0.1dev-py2.6.egg/pylearn2/datasets/preprocessing.py&quot;, line 867, in fit<br>    eigs, eigv = linalg.eigh(np.dot(X.T, X) / X.shape[0])<br>MemoryError</pre>\r\n<pre>&nbsp;</pre>\r\n<pre>If I ran the train.py mlp.yaml, it showed</pre>\r\n<pre><br>  File &quot;/usr/lib/python2.6/site-packages/pylearn2-0.1dev-py2.6.egg/pylearn2/config/yaml_parse.py&quot;, line 238, in try_to_import<br>    raise ImportError(base_msg &#43; '. Original exception: '&#43;str(e))<br>ImportError: Could not import pylearn2.scripts.icml_2013_wrepl.black_box but could import pylearn2.scripts.icml_2013_wrepl. Original exception: No module named black_box.black_box_dataset</pre>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "23366",
      "postDate": "04/23/2013 15:34:24",
      "content": "<p>Regaring the MemoryError:</p>\r\n<p>How much RAM do you have on your machine? The unsupervised data is pretty big, so you might just not have enough RAM to run ZCA on it.</p>\r\n<p>Regariding the ImportError:</p>\r\n<p>This is weird, because if you were missing that module, it should have prevented you from getting far enough to get the MemoryError above. What happens if you run</p>\r\n<p>python -c &quot;from pylearn2.scripts.icml_2013_wrepl.black_box.black_box_dataset import BlackBoxDataset&quot;</p>\r\n<p>?<br>\r\n<br>\r\n</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "23398",
      "postDate": "04/24/2013 05:10:59",
      "content": "<p>Hi, I have another problem with the ZCA&#43;MLP benchmark</p>\r\n<p>I run the benchmark by following the instructor:</p>\r\n<p>1.&nbsp;<span style=\"line-height:1.4em\">python learn_zca.py</span></p>\r\n<p>2. train.py zca_mlp.yaml</p>\r\n<p>3.&nbsp;<span style=\"line-height:1.4em\">python make_submission.py zca_mlp_best.pkl zca_mlp_best.csv</span></p>\r\n<p>&nbsp;</p>\r\n<p><span style=\"line-height:1.4em\">The zca_mlp_best.pkl make it 0.26 and zca_mlp.pkl make it 0.30</span></p>\r\n<p>All seem strange...</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "23411",
      "postDate": "04/24/2013 14:01:58",
      "content": "<p>I'm pretty busy right now preparing to leave for a conference, and I don't know what could cause this off the top of my head. I'll see if any of the other pylearn developers are willing to look into it. If none of them can, I should be able to look into\r\n this on Sunday.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "23419",
      "postDate": "04/24/2013 17:03:09",
      "content": "<p>cc11, if you update pylearn2, the ImportError should go away. I just merged a pull request from Will Kurt that fixes it.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 23366,
      "author_name": "iangoodfellow",
      "author_url": "",
      "post_date": "04/23/2013 15:34:24",
      "content": "<p>Regaring the MemoryError:</p>\r\n<p>How much RAM do you have on your machine? The unsupervised data is pretty big, so you might just not have enough RAM to run ZCA on it.</p>\r\n<p>Regariding the ImportError:</p>\r\n<p>This is weird, because if you were missing that module, it should have prevented you from getting far enough to get the MemoryError above. What happens if you run</p>\r\n<p>python -c &quot;from pylearn2.scripts.icml_2013_wrepl.black_box.black_box_dataset import BlackBoxDataset&quot;</p>\r\n<p>?<br>\r\n<br>\r\n</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 23398,
      "author_name": "binghsu",
      "author_url": "",
      "post_date": "04/24/2013 05:10:59",
      "content": "<p>Hi, I have another problem with the ZCA&#43;MLP benchmark</p>\r\n<p>I run the benchmark by following the instructor:</p>\r\n<p>1.&nbsp;<span style=\"line-height:1.4em\">python learn_zca.py</span></p>\r\n<p>2. train.py zca_mlp.yaml</p>\r\n<p>3.&nbsp;<span style=\"line-height:1.4em\">python make_submission.py zca_mlp_best.pkl zca_mlp_best.csv</span></p>\r\n<p>&nbsp;</p>\r\n<p><span style=\"line-height:1.4em\">The zca_mlp_best.pkl make it 0.26 and zca_mlp.pkl make it 0.30</span></p>\r\n<p>All seem strange...</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 23411,
      "author_name": "iangoodfellow",
      "author_url": "",
      "post_date": "04/24/2013 14:01:58",
      "content": "<p>I'm pretty busy right now preparing to leave for a conference, and I don't know what could cause this off the top of my head. I'll see if any of the other pylearn developers are willing to look into it. If none of them can, I should be able to look into\r\n this on Sunday.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 23419,
      "author_name": "iangoodfellow",
      "author_url": "",
      "post_date": "04/24/2013 17:03:09",
      "content": "<p>cc11, if you update pylearn2, the ImportError should go away. I just merged a pull request from Will Kurt that fixes it.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "23365": "",
    "23366": "",
    "23398": "",
    "23411": "",
    "23419": ""
  },
  "source": "meta"
}