Replacing PyMC3 plots w/ Arviz plots & sigma Param change [Part 4] #24

CloudChaoszero · 2021-01-19T03:18:00Z

Description

The following is a large PR breakdown of PR #16.

Replace PyMC3 dependent plots with arviz plots in case studies & examples.

The issue is a 3-part step for the following:
- Initiative via Issue Remove Plots & Diagnostics from PyMC Code Base pymc#4371 (Leave open)

Replace parameter sd with sigma (e.g. some examples have pm.Normal(...sd=...)

Closes issue Use sigma instead of sd #2

review-notebook-app · 2021-01-19T03:18:05Z

Check out this pull request on

See visual diffs & provide feedback on Jupyter Notebooks.

Powered by ReviewNB

review-notebook-app · 2021-02-07T21:33:37Z

View / edit / reply to this conversation on ReviewNB

twiecki commented on 2021-02-07T21:33:37Z
----------------------------------------------------------------

plot is missing

CloudChaoszero commented on 2021-02-14T08:51:31Z
----------------------------------------------------------------

There appears to be a number of errors in this notebook. I display some in this (personal) PR.

One error in particular that I first come across is

GLM-model-selection KeyError: 'var names: "['sd_log__'] are not present" in dataset'

It's noted in this raised issue

CloudChaoszero commented on 2021-02-14T08:52:58Z
----------------------------------------------------------------

Hmm, should we just revert this Notebook to the upstream/main state, thoughts?

michaelosthege commented on 2021-03-08T11:29:30Z
----------------------------------------------------------------

Try calling pyplot.show().

The interactive plots are always a bit dangerous, because they often get lost when saving the notebook. Maybe you can find out how to tell nbconvert to save the widget state and then run the notebook from the command line?

This does NOT save the widget state: jupyter nbconvert --ExecutePreprocessor.kernel_name="python3" --ExecutePreprocessor.timeout=14000 --execute --inplace {}

michaelosthege commented on 2021-03-08T11:35:28Z
----------------------------------------------------------------

The {} is where the notebook filename goes

CloudChaoszero commented on 2021-03-14T05:05:12Z
----------------------------------------------------------------

@michaelosthege thanks for the response! I'm somewhat confused. I get this error

var names: "['sd_log__']

I created a gist of showing the example. :/

https://gist.github.com/CloudChaoszero/09b566b857190cb04391c615dc08c129

michaelosthege commented on 2021-03-14T11:16:03Z
----------------------------------------------------------------

The key belongs to the transform of sd, but it looks like sd is sampled untransformed. I'm not familiar with this notebook and find the pandas stuff happening above quite confusing. @twiecki do you know if this model selection can be simplified with newer ArviZ features? Having to work with straces directly is not something we should need to teach 😬

Multiprocess sampling (2 chains in 2 jobs) 
NUTS: [sd, np.power(x, 5), np.power(x, 4), np.power(x, 3), np.power(x, 2), x, Intercept]

review-notebook-app · 2021-02-07T21:33:38Z

View / edit / reply to this conversation on ReviewNB

twiecki commented on 2021-02-07T21:33:38Z
----------------------------------------------------------------

plot is missing

CloudChaoszero commented on 2021-02-14T08:52:26Z
----------------------------------------------------------------

For errors in this notebook, see the other mentioned comment about issues in this notebook

review-notebook-app · 2021-02-07T21:33:39Z

View / edit / reply to this conversation on ReviewNB

twiecki commented on 2021-02-07T21:33:38Z
----------------------------------------------------------------

plot is missing

review-notebook-app · 2021-02-07T21:33:39Z

View / edit / reply to this conversation on ReviewNB

twiecki commented on 2021-02-07T21:33:39Z
----------------------------------------------------------------

table is missing

review-notebook-app · 2021-02-07T21:33:40Z

View / edit / reply to this conversation on ReviewNB

twiecki commented on 2021-02-07T21:33:40Z
----------------------------------------------------------------

plot is missing

review-notebook-app · 2021-02-07T21:33:41Z

View / edit / reply to this conversation on ReviewNB

twiecki commented on 2021-02-07T21:33:41Z
----------------------------------------------------------------

not sure what annotate now is but this is an error

CloudChaoszero commented on 2021-02-14T08:30:59Z
----------------------------------------------------------------

Looks like the library deprecated the annotate function, [seen in this Seaborn Issue](https://github.com/mwaskom/seaborn/issues/2306) :/

review-notebook-app · 2021-02-07T21:33:41Z

View / edit / reply to this conversation on ReviewNB

twiecki commented on 2021-02-07T21:33:41Z
----------------------------------------------------------------

colors here are mixed up in the 2D plot, should match marginals

review-notebook-app · 2021-02-07T21:33:42Z

View / edit / reply to this conversation on ReviewNB

twiecki commented on 2021-02-07T21:33:42Z
----------------------------------------------------------------

ugh, damn, this convergence is really bad. maybe try with target_accept=0.99

CloudChaoszero commented on 2021-02-14T08:28:51Z
----------------------------------------------------------------

Sounds good. I added a target_accept = 0.99

…h sigma ⏪ Revert changes due to improper metadata updates 🎨 Fix divergence in last cells and outdated errors due to variou package changes

CloudChaoszero · 2021-02-14T08:28:52Z

Sounds good. I added a target_accept = 0.99

View entire conversation on ReviewNB

CloudChaoszero · 2021-02-14T08:31:00Z

Looks like the library deprecated the annotate function, [seen in this Seaborn Issue](https://github.com/mwaskom/seaborn/issues/2306) :/

View entire conversation on ReviewNB

CloudChaoszero · 2021-02-14T08:51:32Z

There appears to be a number of errors in this notebook. I display some in this (personal) PR.

One error in particular that I first come across is

GLM-model-selection KeyError: 'var names: "['sd_log__'] are not present" in dataset'

It's noted in this raised issue

View entire conversation on ReviewNB

CloudChaoszero · 2021-02-14T08:52:27Z

For errors in this notebook, see the other mentioned comment about issues in this notebook

View entire conversation on ReviewNB

CloudChaoszero · 2021-02-14T08:52:58Z

Hmm, should we just revert this Notebook to the upstream/main state, thoughts?

View entire conversation on ReviewNB

michaelosthege · 2021-03-08T11:29:31Z

Try calling pyplot.show().

The interactive plots are always a bit dangerous, because they often get lost when saving the notebook. Maybe you can find out how to tell nbconvert to save the widget state and then run the notebook from the command line?

This does NOT save the widget state: jupyter nbconvert --ExecutePreprocessor.kernel_name="python3" --ExecutePreprocessor.timeout=14000 --execute --inplace {}

View entire conversation on ReviewNB

review-notebook-app · 2021-03-08T11:32:21Z

View / edit / reply to this conversation on ReviewNB

michaelosthege commented on 2021-03-08T11:32:21Z
----------------------------------------------------------------

In most other notebooks we have the watermark at the end?

michaelosthege · 2021-03-08T11:35:29Z

The {} is where the notebook filename goes

View entire conversation on ReviewNB

OriolAbril · 2021-03-08T16:36:50Z

In most other notebooks we have the watermark at the end?

Yes, all notebooks are supposed to have the watermark at the end.

The interactive plots are always a bit dangerous, because they often get lost when saving the notebook

Agreed, however, I don't think there are interactive plots in the notebook

KeyError: 'var names: "['sd_log__'] are not present" in dataset'

This is because ArviZ does not store the transformed variables in the posterior group. You can either use the transform argument in the plotting functions (I think they all have it but not sure). Or manually add the transformed variable to the dataset. See one example here: arviz-devs/arviz#641

CloudChaoszero · 2021-03-14T05:05:13Z

@michaelosthege thanks for the response! I'm somewhat confused. I get this error

var names: "['sd_log__']

I created a gist of showing the example. :/

https://gist.github.com/CloudChaoszero/09b566b857190cb04391c615dc08c129

View entire conversation on ReviewNB

michaelosthege · 2021-03-14T11:16:04Z

The key belongs to the transform of sd, but it looks like sd is sampled untransformed. I'm not familiar with this notebook and find the pandas stuff happening above quite confusing. @twiecki do you know if this model selection can be simplified with newer ArviZ features? Having to work with straces directly is not something we should need to teach 😬

Multiprocess sampling (2 chains in 2 jobs) 
NUTS: [sd, np.power(x, 5), np.power(x, 4), np.power(x, 3), np.power(x, 2), x, Intercept]

View entire conversation on ReviewNB

OriolAbril · 2021-03-14T12:21:36Z

Maybe we can create an issue as a remainder to fix this but for now go ahead and merge the PR?

CloudChaoszero · 2021-03-21T02:25:19Z

Maybe we can create an issue as a remainder to fix this but for now go ahead and merge the PR?

Sounds good @OriolAbril ! I made this issue

OriolAbril · 2021-03-21T06:05:43Z

I saw the issue, thanks! I don't see the comments of this PR on the issue yet, especially the ones about the glm comparison one, can you add them?

CloudChaoszero · 2021-03-21T20:19:52Z

I saw the issue, thanks! I don't see the comments of this PR on the issue yet, especially the ones about the glm comparison one, can you add them?

Sounds good. @OriolAbril ! I added it at the last section, of the original description, and linked one of this PR's comments

* change workflows directory * change workflows directory * use last version of pymc * update tests, update aesara import

🎨 Update plots from pymc to arviz dependencies & replace param sd wit…

9662276

…h sigma ⏪ Revert changes due to improper metadata updates 🎨 Fix divergence in last cells and outdated errors due to variou package changes

CloudChaoszero force-pushed the replace-pymc3-arviz-plots_part4 branch from db1afed to 9662276 Compare February 14, 2021 08:28

CloudChaoszero added 2 commits March 20, 2021 19:03

🔀 Prep to resolve merge conflict on GLM Linear notebook

5297098

🔀 GLM Linear notebook update resolve merge conflict

105ac55

CloudChaoszero mentioned this pull request Mar 21, 2021

PyMC3 Examples Known Errors & Notes (March 2021) #43

Open

5 tasks

OriolAbril merged commit 885fb46 into pymc-devs:main Mar 21, 2021

OriolAbril mentioned this pull request Mar 21, 2021

Update GLM hierarchical binomial model to best practices #45

Merged

CloudChaoszero deleted the replace-pymc3-arviz-plots_part4 branch April 5, 2021 01:53

farhanreynaldo mentioned this pull request Nov 13, 2021

Update Hierarchical binomial notebook #248

Merged

twiecki pushed a commit that referenced this pull request Jan 17, 2023

change workflows directory (#24)

0b02af8

* change workflows directory * change workflows directory * use last version of pymc * update tests, update aesara import

Uh oh!

Replacing PyMC3 plots w/ Arviz plots & sigma Param change [Part 4] #24

Replacing PyMC3 plots w/ Arviz plots & sigma Param change [Part 4] #24

Uh oh!

Conversation

CloudChaoszero commented Jan 19, 2021

Description

Uh oh!

review-notebook-app bot commented Jan 19, 2021

Uh oh!

review-notebook-app bot commented Feb 7, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

review-notebook-app bot commented Feb 7, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

review-notebook-app bot commented Feb 7, 2021

Uh oh!

review-notebook-app bot commented Feb 7, 2021

Uh oh!

review-notebook-app bot commented Feb 7, 2021

Uh oh!

review-notebook-app bot commented Feb 7, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

review-notebook-app bot commented Feb 7, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

review-notebook-app bot commented Feb 7, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

CloudChaoszero commented Feb 14, 2021

Uh oh!

CloudChaoszero commented Feb 14, 2021

Uh oh!

CloudChaoszero commented Feb 14, 2021

Uh oh!

CloudChaoszero commented Feb 14, 2021

Uh oh!

CloudChaoszero commented Feb 14, 2021

Uh oh!

michaelosthege commented Mar 8, 2021

Uh oh!

review-notebook-app bot commented Mar 8, 2021

Uh oh!

michaelosthege commented Mar 8, 2021

Uh oh!

OriolAbril commented Mar 8, 2021

Uh oh!

CloudChaoszero commented Mar 14, 2021

Uh oh!

michaelosthege commented Mar 14, 2021

Uh oh!

OriolAbril commented Mar 14, 2021

Uh oh!

CloudChaoszero commented Mar 21, 2021

Uh oh!

OriolAbril commented Mar 21, 2021

Uh oh!

CloudChaoszero commented Mar 21, 2021

Uh oh!

Uh oh!

review-notebook-app bot commented Feb 7, 2021 •

edited

Loading

review-notebook-app bot commented Feb 7, 2021 •

edited

Loading

review-notebook-app bot commented Feb 7, 2021 •

edited

Loading

review-notebook-app bot commented Feb 7, 2021 •

edited

Loading

review-notebook-app bot commented Feb 7, 2021 •

edited

Loading