[MRG + 1] ENH: try making HNN GUI with ipywidgets #76

jasmainak · 2019-09-14T03:30:26Z

Here is a demo of the GUI with ipywidgets.

Some advantages of this approach:

Easier to maintain. We can probably get 95% of the functionality with 1000 lines of clean code as opposed to maintaining 8000 lines of messy code
Easy to install/deploy in classroom environment
You can easily add new features (e.g., Nick's work related to connectivity).
Ability to spin up HNN from the mobile and host a demo on Brown website using JupyterHub :-)
Potential to integrate with MNE source localization 'notebook' backend

To run this example, you'll need ipywidgets and voila. First make sure you have the latest version of jupyter

$ conda install jupyter

Then install the dependencies:

$ pip install ipywidgets voila

Finally do:

$ voila hnn_widget.ipynb

voila is not strictly necessary. You can also do:

$ jupyter-notebook hnn_widget.ipynb

and run the first cell.

You should see:

current dipole	spikes

PS: If we want interactive plots, we could potentially leverage ipympl but I would say, let's start with something simple

HNN-core examples tested

Evoked example
Rhythmic (alpha) example
Gamma example

TODO / Wishlist

codecov-io · 2019-09-14T03:38:52Z

Codecov Report

Merging #76 (18b98a6) into master (00ebb0f) will decrease coverage by 0.26%.
The diff coverage is n/a.

❗ Current head 18b98a6 differs from pull request most recent head cd7bb0d. Consider uploading reports for the commit cd7bb0d to get more accurate results

@@            Coverage Diff             @@
##           master      #76      +/-   ##
==========================================
- Coverage   90.69%   90.43%   -0.27%     
==========================================
  Files          14       14              
  Lines        2365     2425      +60     
==========================================
+ Hits         2145     2193      +48     
- Misses        220      232      +12

Impacted Files	Coverage Δ
hnn_core/network.py	`93.05% <0.00%> (-1.44%)`	⬇️
hnn_core/parallel_backends.py	`82.70% <0.00%> (-0.87%)`	⬇️
hnn_core/network_builder.py	`92.16% <0.00%> (+0.27%)`	⬆️

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 00ebb0f...cd7bb0d. Read the comment docs.

jasmainak · 2021-03-31T03:34:25Z

Now that HNN-core is cleaned up, I'd like to revisit this for 0.2. A dashboard-GUI might be much easier to maintain in the long run.

jasmainak · 2021-04-13T19:55:27Z

@stephanie-r-jones @ntolley @rythorpe @cjayb

related to today's discussion, I want to bring to your attention this GUI code I wrote long time ago. It's less than 100 lines of code at the moment and can already do this:

I'm sure the left panel can be modified to add some tabs for modifying different kinds of parameters.

And then you can use voila to hide the code. If someone wants to pair up and try to improve this, I'd be up for it. Clone this branch and give it a shot I would say!

ntolley · 2021-04-14T15:10:45Z

@jasmainak this is awesome! Once I wrap up my open PR's I'd be interested in testing this out. Given the simplicity of ipywidgets, I think it will provide a nice space to improve on the essential functionality for the GUI (taking a lot from the existing version).

jasmainak · 2021-04-14T22:02:14Z

wonderful, thanks for the encouragement! I pushed another tiny update to the UI

codecov-commenter · 2021-04-17T03:12:01Z

Codecov Report

Merging #76 (471058f) into master (bab8196) will decrease coverage by 2.88%.
The diff coverage is 63.63%.

@@            Coverage Diff             @@
##           master      #76      +/-   ##
==========================================
- Coverage   90.51%   87.63%   -2.89%     
==========================================
  Files          18       19       +1     
  Lines        3415     3792     +377     
==========================================
+ Hits         3091     3323     +232     
- Misses        324      469     +145

Impacted Files	Coverage Δ
hnn_core/gui/gui.py	`62.53% <62.53%> (ø)`
hnn_core/params.py	`90.83% <88.23%> (-0.08%)`	⬇️
hnn_core/parallel_backends.py	`81.49% <0.00%> (-0.83%)`	⬇️

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update bab8196...471058f. Read the comment docs.

ntolley · 2021-04-19T14:06:23Z

hnn_core/gui.py

+    sliders = list()
+    for d in param_keys:
+        slider = FloatLogSlider(
+            value=params[d], min=-5, max=1, step=0.2,


The dynamic range may differ between neurotransmitters, might be worth considering different min/maxes for GABA vs. AMPA

Note this in log scale! But yes totally open to suggestions to improve the interface here

ntolley · 2021-04-19T14:10:15Z

hnn_core/gui.py

+               _get_sliders(params,
+               ['gbar_L2Pyr_L5Pyr', 'gbar_L2Basket_L5Pyr'])]
+
+    # accordians to group local-connectivity by cel type


Suggested change

# accordians to group local-connectivity by cel type

# accordians to group local-connectivity by cell type

ntolley · 2021-04-19T14:19:10Z

Playing around with the GUI, personally I'm a big fan of the dashboard layout. Quick note, the drop down for the drives defaults to a blank value, is this intentional?

jasmainak · 2021-04-19T14:21:43Z

yeah, it's a bit intentional for now. You need to change the dropdown for a drive to be added. I think we need to change it ultimately to be a radio + "add" button. I'm just trying to get an MVP working first. Did you try the mybinder link above??

ntolley · 2021-04-19T14:38:21Z

Just tried it now! Really awesome to have that working on binder. The barrier for entry is getting even lower now that there is a pip install and an immediately usable GUI from the the web.

rythorpe · 2021-04-19T15:49:25Z

This is super cool @jasmainak! +100 for the dashboard GUI configuration. The code is a lot simpler with a modular design.

When X-forwarding the current hnn-gui hosted on a computing cluster, I've often noticed that having multiple application windows open can cause my system to freeze up with fluctuations in my internet speed/bandwidth. I'm curious to see if using ipywidgets will improve this.

jasmainak · 2021-04-19T19:52:07Z

You should give it a try! Here are two links to get you started:

https://docs.anaconda.com/anaconda/user-guide/tasks/remote-jupyter-notebook/
https://voila.readthedocs.io/en/stable/using.html#as-a-jupyter-server-extension

I think it should also be pretty doable to configure it for an entire class. I have worked with Isabel for the MNE workshops and she set up a JupyterHub in Brown. It worked pretty well.

jasmainak · 2021-04-25T22:39:26Z

hnn_core/gui.py

+        variables['net'].connectivity = list()
+        # XXX: hack until there is a proper API
+        variables['net'].self._set_default_connections()


@ntolley food for thought for you what should be the right API here

EDIT: maybe Network shouldn't be instantiated in GUI until the run button is clicked ?

jasmainak · 2022-07-09T18:34:17Z

When I click on "Add drive" I get the following error:

jasmainak · 2022-07-09T18:47:29Z

okay I'm done for now. Let me know once the error is fixed and I'll do another round :)

chenghuzi · 2022-07-10T03:50:26Z

okay I'm done for now. Let me know once the error is fixed and I'll do another round :)

Fixed in 3d9082d

chenghuzi · 2022-07-10T03:51:27Z

@chenghuzi I'm undoing the Black code formatting everywhere. Please disable it in your editor ... it's too distracting.

Sure, as you like :)

MANIFEST.in

README.rst

rythorpe · 2022-07-10T20:18:10Z

hnn_core/gui/gui.py

+    default_data = {
+        'weights_ampa': {
+            'L5_pyramidal': 0.,
+            'L2_pyramidal': 0.,
+            'L5_basket': 0.,
+            'L2_basket': 0.
+        },
+        'weights_nmda': {
+            'L5_pyramidal': 0.,
+            'L2_pyramidal': 0.,
+            'L5_basket': 0.,
+            'L2_basket': 0.
+        },
+        'delays': {
+            'L5_pyramidal': 0.1,
+            'L2_pyramidal': 0.1,
+            'L5_basket': 0.1,
+            'L2_basket': 0.1
+        },
+    }


This is perhaps opening a large can of worms, but what happens when we want to add an hnn-core model that contain more or less cell types than the ones we currently have? Will we need to update the hardcoded cell types here?

yep. But the widgets will also need to be updated ... no point in generalizing just this.

Unless I'm missing something crucial here, I feel like the GUI code should be as independent from specific network attributes as possible. If we hardcode these values in now, we're just repeating the same mistakes of the past GUI.

As a solution, I propose that we first have the user specify which network model to load into the GUI environment, and then instantiate cell specific widgets based on the cell types and properties contained in that Network instance.

I agree that would be a great feature to have ... at the moment, we don't yet allow the user to modify cell properties (e.g., the constants in the mechanisms). I think we should discuss what would be a reasonable and "easy" thing to fix in the code. I propose that we merge this PR and after that spend a week on maintenance -- writing tests, refactoring, making it more future proof etc. I highly recommend pulling the PR and trying to play around with the code to get a sense of what is possible and what is not.

jasmainak · 2022-07-10T22:40:08Z

@chenghuzi I still have this problem:

Note how the button next to proximal can't be selected. Make sure to pull my last commit before fixing it.

jasmainak · 2022-07-10T22:41:10Z

another issue is that the location is not aligned with the rest of the parameters. See:

hnn_core/gui/gui.py

jasmainak · 2022-07-10T23:02:06Z

we're close @chenghuzi ! I can see the light for this PR :) Just some more tweaks and we can merge. Also please set the PR title to MRG when it's ready on your end

jasmainak · 2022-07-10T23:10:00Z

one more issue. There should be some check in the callback that returns if the simulation has not yet been run (don't use try/except)

chenghuzi · 2022-07-11T04:59:57Z

another issue is that the location is not aligned with the rest of the parameters. See:

Now I put that in the title, which should make it more clear

chenghuzi · 2022-07-11T05:02:02Z

@chenghuzi I still have this problem:

Note how the button next to proximal can't be selected. Make sure to pull my last commit before fixing it.

I tried on a very small browser window and it didn't happen. This is weird... Anyway, I'll change the alignment from HBox to Vbox, which should provide enough width for each of them.

chenghuzi · 2022-07-11T05:24:19Z

one more issue. There should be some check in the callback that returns if the simulation has not yet been run (don't use try/except)

Fixed

jasmainak · 2022-07-11T14:53:04Z

So it turns out I can't approve my own PR ;-) I think @rythorpe or @ntolley you'll have to be the judges here and merge when ready. Please merge by "rebase and merge" to keep the commits.

rythorpe · 2022-07-11T17:38:03Z

I'll give it another run through and (hopefully!) rebase and merge by this afternoon. Stay tuned!

rythorpe

This looks great @chenghuzi and @jasmainak! I've created a follow issue with some enhancement nitpicks that we can discuss and/or address in the next few PRs.

rythorpe · 2022-07-11T23:11:53Z

Thanks @chenghuzi and @jasmainak for pushing this forward. We've now got ourselves a new GUI 🎉

jasmainak · 2022-07-11T23:32:55Z

Great work @chenghuzi ! 🥳 🥳 🥳

jasmainak added this to the 0.2 milestone Mar 31, 2021

jasmainak force-pushed the hnn_gui branch from b100540 to 8ecba13 Compare April 13, 2021 19:18

jasmainak force-pushed the hnn_gui branch from 9eed708 to c741113 Compare April 16, 2021 04:14

jasmainak changed the title ~~ENH: try making HNN GUI with ipywidgets~~ [WIP] ENH: try making HNN GUI with ipywidgets Apr 16, 2021

jasmainak force-pushed the hnn_gui branch 2 times, most recently from 1533641 to 1e5dbca Compare April 16, 2021 14:06

jasmainak force-pushed the hnn_gui branch from 1c1e43d to a9bc0b6 Compare April 17, 2021 18:58

jasmainak mentioned this pull request Apr 18, 2021

constructing network object is too slow #317

Closed

jasmainak force-pushed the hnn_gui branch 4 times, most recently from 4356129 to 259d7c5 Compare April 19, 2021 14:00

ntolley reviewed Apr 19, 2021

View reviewed changes

jasmainak force-pushed the hnn_gui branch from 259d7c5 to 45e6ad8 Compare April 19, 2021 14:08

ntolley reviewed Apr 19, 2021

View reviewed changes

jasmainak force-pushed the hnn_gui branch 2 times, most recently from c0db431 to 353d8c0 Compare April 21, 2021 22:02

jasmainak commented Apr 25, 2021

View reviewed changes

jasmainak force-pushed the hnn_gui branch from 3d58401 to 7207a0f Compare April 27, 2021 21:47

Fix error

3d9082d

Fix formatting error

fde873c

rythorpe reviewed Jul 10, 2022

View reviewed changes

MANIFEST.in Show resolved Hide resolved

rythorpe reviewed Jul 10, 2022

View reviewed changes

README.rst Show resolved Hide resolved

rythorpe reviewed Jul 10, 2022

View reviewed changes

MAINT: add connecting to callbacks at the end

e857f24

jasmainak force-pushed the hnn_gui branch from fc24bad to e857f24 Compare July 10, 2022 22:39

jasmainak commented Jul 10, 2022

View reviewed changes

hnn_core/gui/gui.py Outdated Show resolved Hide resolved

jasmainak commented Jul 10, 2022

View reviewed changes

hnn_core/gui/gui.py Outdated Show resolved Hide resolved

Fix layout

471058f

chenghuzi changed the title ~~[WIP] ENH: try making HNN GUI with ipywidgets~~ [MRG] ENH: try making HNN GUI with ipywidgets Jul 11, 2022

jasmainak changed the title ~~[MRG] ENH: try making HNN GUI with ipywidgets~~ [MRG + 1] ENH: try making HNN GUI with ipywidgets Jul 11, 2022

rythorpe mentioned this pull request Jul 11, 2022

follow-up GUI improvements #497

Closed

8 tasks

rythorpe approved these changes Jul 11, 2022

View reviewed changes

rythorpe merged commit 9850f20 into jonescompneurolab:master Jul 11, 2022

	# accordians to group local-connectivity by cel type
	# accordians to group local-connectivity by cell type

[MRG + 1] ENH: try making HNN GUI with ipywidgets #76

[MRG + 1] ENH: try making HNN GUI with ipywidgets #76

Conversation

jasmainak commented Sep 14, 2019 • edited Loading

HNN-core examples tested

TODO / Wishlist

codecov-io commented Sep 14, 2019 • edited Loading

Codecov Report

jasmainak commented Mar 31, 2021

jasmainak commented Apr 13, 2021 • edited Loading

ntolley commented Apr 14, 2021

jasmainak commented Apr 14, 2021 • edited Loading

codecov-commenter commented Apr 17, 2021 • edited Loading

Codecov Report

ntolley Apr 19, 2021 • edited Loading

Choose a reason for hiding this comment

jasmainak Apr 19, 2021

Choose a reason for hiding this comment

ntolley Apr 19, 2021

Choose a reason for hiding this comment

ntolley commented Apr 19, 2021

jasmainak commented Apr 19, 2021

ntolley commented Apr 19, 2021

rythorpe commented Apr 19, 2021

jasmainak commented Apr 19, 2021

jasmainak Apr 25, 2021 • edited Loading

Choose a reason for hiding this comment

jasmainak commented Jul 9, 2022

jasmainak commented Jul 9, 2022

chenghuzi commented Jul 10, 2022

chenghuzi commented Jul 10, 2022

rythorpe Jul 10, 2022

Choose a reason for hiding this comment

jasmainak Jul 10, 2022

Choose a reason for hiding this comment

rythorpe Jul 11, 2022

Choose a reason for hiding this comment

jasmainak Jul 11, 2022

Choose a reason for hiding this comment

jasmainak commented Jul 10, 2022

jasmainak commented Jul 10, 2022

jasmainak commented Jul 10, 2022

jasmainak commented Jul 10, 2022

chenghuzi commented Jul 11, 2022

chenghuzi commented Jul 11, 2022

chenghuzi commented Jul 11, 2022

jasmainak commented Jul 11, 2022

rythorpe commented Jul 11, 2022

rythorpe left a comment

Choose a reason for hiding this comment

rythorpe commented Jul 11, 2022

jasmainak commented Jul 11, 2022

jasmainak commented Sep 14, 2019 •

edited

Loading

codecov-io commented Sep 14, 2019 •

edited

Loading

jasmainak commented Apr 13, 2021 •

edited

Loading

jasmainak commented Apr 14, 2021 •

edited

Loading

codecov-commenter commented Apr 17, 2021 •

edited

Loading

ntolley Apr 19, 2021 •

edited

Loading

jasmainak Apr 25, 2021 •

edited

Loading