[REVIEW]: TypeError in plot_infotopo.py

[pitmonticone]

When running the example script examples/metrics/plot_infotopo.py I encountered the following error:

TypeError: unsupported type for timedelta seconds component: jaxlib.xla_extension.ArrayImpl

REVIEW ISSUE: REVIEW ISSUE: openjournals/joss-reviews#7360

[EtienneCmb]

Thank you for reporting this issue.

October, Jax dropped some internal callback functions that we used to display the progress bar (jax's changelog). I suspect that this error was due to the progress bar. We fixed this issue with commits 38cc6bc and feb9eb7. Tests are now passing for python>=3.8.

Could you please fetch the latest commits, update jax with pip install -U jax and running this example again?

[pitmonticone]

Fetched the latest commits and updated jax.

Here is the output:

Copnorm and demean the data
Get list of multiplets
   index  order       hoi           multiplet                                                                
0     72      5  0.182472     [0, 1, 2, 3, 6]
1     35      4  0.181906        [0, 1, 2, 3]
2      0      3  0.181892           [0, 1, 2]
3     37      4  0.181686        [0, 1, 2, 5]
4     38      4  0.181507        [0, 1, 2, 6]
5     84      5 -0.447083     [0, 3, 4, 5, 6]
6     69      4 -0.447269        [3, 4, 5, 6]
7     89      5 -0.453236     [1, 3, 4, 5, 6]
8     97      6 -0.453758  [1, 2, 3, 4, 5, 6]
9     95      6 -0.454479  [0, 1, 3, 4, 5, 6]
Compute entropy with gc
    Copnorm and demean the data
Get list of multiplets
Get list of multiplets                                                                                       
Traceback (most recent call last):
  File "/Users/pietromonticone/GitHub/hoi/examples/metrics/plot_infotopo.py", line 99, in <module>
    hoi = model.fit(minsize=3, method="gc")
          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/pietromonticone/GitHub/hoi/.venv/lib/python3.12/site-packages/hoi/metrics/infotopo.py", line 154, in fit
    _, hoi = jax.lax.scan(
             ^^^^^^^^^^^^^
  File "/Users/pietromonticone/GitHub/hoi/.venv/lib/python3.12/site-packages/hoi/utils/progressbar.py", line 62, in wrapper_progress_bar
    _update_progress_bar(iter_num)
  File "/Users/pietromonticone/GitHub/hoi/.venv/lib/python3.12/site-packages/hoi/utils/progressbar.py", line 151, in _update_progress_bar
    _ = jax.jax.lax.cond(
        ^^^^^^^^^^^^^^^^^
  File "/Users/pietromonticone/GitHub/hoi/.venv/lib/python3.12/site-packages/hoi/utils/progressbar.py", line 153, in <lambda>
    lambda _: host_callback.id_tap(
              ^^^^^^^^^^^^^^^^^^^^^
  File "/Users/pietromonticone/GitHub/hoi/.venv/lib/python3.12/site-packages/jax/experimental/host_callback.py", line 27, in call
    raise NotImplementedError(
NotImplementedError: jax.experimental.host_callback has been deprecated since March 2024 and is now no longer supported. See https://github.com/jax-ml/jax/issues/20385
--------------------
For simplicity, JAX has removed its internal frames from the traceback of the following exception. Set JAX_TRACEBACK_FILTERING=off to include these.

[EtienneCmb]

I believe there's an issue with your update because the latest version of hoi doesn't have the progress bar anymore. Now we rely on the package jax-tqdm. Have you installed hoi using pip or python setup.py?

[pitmonticone]

I have installed it in my local .venv using pip install -U hoi.

[pitmonticone]

Can you deploy the new version on pypi?

[EtienneCmb]

Yes, sure, I'll do it tomorrow morning ! I'll let you know once done 👍

[EtienneCmb]

Hi @pitmonticone , the new version is on pypi

[pitmonticone]

Yes, now it works fine:

  Copnorm and demean the data
Get list of multiplets
   index  order       hoi           multiplet                                                                                                                                                                 
0     36      4  0.203451        [0, 1, 2, 4]
1     73      5  0.203430     [0, 1, 2, 4, 5]
2     37      4  0.203373        [0, 1, 2, 5]
3      0      3  0.203372           [0, 1, 2]
4     72      5  0.202888     [0, 1, 2, 3, 6]
5     96      6 -0.420208  [0, 2, 3, 4, 5, 6]
6     84      5 -0.420437     [0, 3, 4, 5, 6]
7     95      6 -0.420715  [0, 1, 3, 4, 5, 6]
8     97      6 -0.421467  [1, 2, 3, 4, 5, 6]
9     89      5 -0.422012     [1, 3, 4, 5, 6]
Compute entropy with gc
    Copnorm and demean the data
Get list of multiplets
Get list of multiplets                                                                                                                                                                                        
Mutual information: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 99/99 [00:00<00:00, 75980.99it/s]
   index  order       hoi     multiplet
0      0      3  0.203372     [0, 1, 2]
1     69      4  0.094784  [3, 4, 5, 6]
2      4      3  0.000405     [0, 1, 6]
3     62      4  0.000385  [1, 3, 4, 6]
4     52      4  0.000381  [0, 3, 4, 6]
5     47      4 -0.000550  [0, 2, 3, 6]
6      8      3 -0.001450     [0, 2, 6]
7     33      3 -0.057861     [3, 5, 6]
8     31      3 -0.060821     [3, 4, 5]
9     32      3 -0.110925     [3, 4, 6]

Sorry for the pedantry, but during review I need to test the user experience.

I will close all the similar issues.

[pitmonticone]

I consider this issue resolved.

[EtienneCmb]

Yes, now it works fine:

  Copnorm and demean the data
Get list of multiplets
   index  order       hoi           multiplet                                                                                                                                                                 
0     36      4  0.203451        [0, 1, 2, 4]
1     73      5  0.203430     [0, 1, 2, 4, 5]
2     37      4  0.203373        [0, 1, 2, 5]
3      0      3  0.203372           [0, 1, 2]
4     72      5  0.202888     [0, 1, 2, 3, 6]
5     96      6 -0.420208  [0, 2, 3, 4, 5, 6]
6     84      5 -0.420437     [0, 3, 4, 5, 6]
7     95      6 -0.420715  [0, 1, 3, 4, 5, 6]
8     97      6 -0.421467  [1, 2, 3, 4, 5, 6]
9     89      5 -0.422012     [1, 3, 4, 5, 6]
Compute entropy with gc
    Copnorm and demean the data
Get list of multiplets
Get list of multiplets                                                                                                                                                                                        
Mutual information: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 99/99 [00:00<00:00, 75980.99it/s]
   index  order       hoi     multiplet
0      0      3  0.203372     [0, 1, 2]
1     69      4  0.094784  [3, 4, 5, 6]
2      4      3  0.000405     [0, 1, 6]
3     62      4  0.000385  [1, 3, 4, 6]
4     52      4  0.000381  [0, 3, 4, 6]
5     47      4 -0.000550  [0, 2, 3, 6]
6      8      3 -0.001450     [0, 2, 6]
7     33      3 -0.057861     [3, 5, 6]
8     31      3 -0.060821     [3, 4, 5]
9     32      3 -0.110925     [3, 4, 6]

Sorry for the pedantry, but during review I need to test the user experience.

I will close all the similar issues.

Great ! No problem, thank you for reporting this !