openrg / un-og-training Goto Github PK

There was a problem in running the glue command in 16.3.1 and 16.3.2. In the last line of the command for tabular output and plotting output the glue command does not work.

Specifically, these are the commands that do not work for each block of code, respectively:
(i) glue(table.to_markdown())
(ii) glue(fig.show())

The problem can be partially resolved by:
(i) installing the glue package: "conda install myst-nb"
(ii) importing the glue package: "from myst-nb import glue

However, the glue command still gives the following error in both blocks of code:
"NameError: name 'glue' is not defined"

Thank you.

Currently we are pinning our Python version to be >= 3.10, which resulted in my installing 3.11 in the un-og-training-dev environment.

I now want to install ogzaf to use in some example solutions, but that is only compatible with Python <= 3.10. Shall up change the pin in OG-ZAF or in the environment for this project?

How should we explain the calibration of pensions? Right now, OG-Core only parameterizes a social security style system.

In PR #24, I inserted the following code block into the OG/Ouput.md chapter:

import ogcore.output_tables as ot
from io import BytesIO
import pickle
import requests
path_dict = {
    "TPI": [
        "https://github.com/PSLmodels/OG-Core/blob/master/tests/test_io_data/TPI_vars_baseline.pkl?raw=true",
        "https://github.com/PSLmodels/OG-Core/blob/master/tests/test_io_data/TPI_vars_reform.pkl?raw=true"
            ],
    "Params": [
        "https://github.com/PSLmodels/OG-Core/blob/master/tests/test_io_data/model_params_baseline.pkl?raw=true",
        "https://github.com/PSLmodels/OG-Core/blob/master/tests/test_io_data/model_params_reform.pkl?raw=true"

    ]
}
output_dict = {
    "TPI": [],
    "Params": []
}
for key in path_dict.keys():
    for path in path_dict[key]:
        r = requests.get(path)
        output_dict[key].append(pickle.load(BytesIO(r.content)))
# make table
table = ot.macro_table(output_dict["TPI"][0], output_dict["Params"][0], output_dict["TPI"][1], output_dict["Params"][1], output_type="pct_diff", num_years=10, start_year= output_dict["Params"][0].start_year)

I have no issue executing this in Python 3.10. But with Python 3.11, there is an error with the pickle.load command when trying to read the pickle file containing the model parameters (which is a ogcore.parameters.Specifications object pickled with Cloudpickle). The traceback is:

TypeError                                 Traceback (most recent call last)
Cell In[2], line 24
     22     for path in path_dict[key]:
     23         r = requests.get(path)
---> 24         output_dict[key].append(pickle.load(BytesIO(r.content)))
     25 # make table
     26 table = ot.macro_table(output_dict["TPI"][0], output_dict["Params"][0], output_dict["TPI"][1], output_dict["Params"][1], output_type="pct_diff", num_years=10, start_year= output_dict["Params"][0].start_year)

TypeError: code() argument 13 must be str, not int

I've tried just using cloudpickle and pickle.loads to unpack and still get an error.

We should re-read chapters from the point of view of a novice and make sure there are sufficient details to get up and running with the software and custom packages.

How much detail should we add about calibrating tax functions? We must assume most countries won't have microsimulation models to link to. Given that, do we present just linear? Show how to estimate Gouveia-Strauss with aggregate data?

Thanks for filing an issue! Please remove any top-level headers that do not apply.

There was a minor typo in the second block of code for Chapter 15 (Simulating OG-Core), which prevented the code from running:
....
p_reform.update_specifications({'CIT_rate': [[0.15]]})
....

When I changed CIT to lower case, the code worked.

The push.yaml GitHub Action is not working. This GH Action is meant to run after a PR is merged (when changes are pushed to the main repository). It automatically updates the repository version and deploys the updated Jupyter Book documentation to the gh-pages branch for the documentation website. However, the "Checkout Repo" action is the first to run, and it is failing with the following error.

/usr/bin/git -c protocol.version=2 fetch --no-tags --prune --progress --no-recurse-submodules --depth=1 origin +d030ba2c5deb85175cf58e1e616c97ac7bdeab0f:refs/remotes/origin/main
  Error: fatal: could not read Username for 'https://github.com/': terminal prompts disabled
  The process '/usr/bin/git' failed with exit code 128

I think this must be some issue with the GitHub secret used in the GH Action. But I can't figure it out. This action works great in my fiscalsim-us repository.

cc: @jdebacker

We use the BYU Applied and Computational Math Emphasis open access Python labs in these training materials. We link to their 2021-2022 materials because those labs are individually packaged and the PDF format displays separately for each of them. It is a little less easy to use the most recent 2023-2024 materials, which are packaged as multi-chapter single PDF files (here and here) that have to be downloaded as .zip files that the students would have to search through. The updates to the 2023-2024 materials are fairly small, but they are significant in some places (e.g., discussion of fstrings in Exceptions and file I/O chapter). Below are some options of what we can do.

1. Current approach. Link to 2021-2021 labs and add supplementary exercises and content for areas that need to be updated.
2. Link to 2023-2024 labs and just have our users download the .zip files and search through to the correct chapters
3. Save the individual PDF files somewhere like Google Drive where they can be pulled up as full PDF files in users' browser.
4. Rewrite each ACME lab as is but as a .md file that can be rendered in the Jupyter Book. This is what the ACME labs should transition to anyway. Maybe this would help them.
5. Write our own .md training content and exercises.

cc: @jdebacker

Thanks for filing an issue! Please remove any top-level headers that do not apply.

After working through Chapter 15 (Simulating OG-Core), I was unable to work through any of the code in Chapter 17 (OG-Core Output). I think this may be due to the missing chapter in the training material, i.e., Chapter 16.

Thank you.

Should all exercises be colored green?

cc @rickecon

If we can add another level to the book's TOC, we should make each section of the Calibration chapter its own sub chapter.

On step 16.2 the calculation of non_stationary_Y fails because the two variables have different dimensions.

non_stationary_Y = TPI_vars["Y"] * np.cumprod(1 + p.g_n) * np.exp(p.g_y * np.arange(p.T))
print(non_stationary_Y

gives the error:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[16], line 1
----> 1 non_stationary_Y = TPI_vars["Y"] * np.cumprod(1 + p.g_n) * np.exp(p.g_y * np.arange(p.T))
      2 print(non_stationary_Y)

ValueError: operands could not be broadcast together with shapes (320,) (400,) 

The dimensions are:
TPI_vars["Y"].shape = (320,)
p.g_n.shape = (400,)

This breaks exercise 16.2.

cc: @sezdi0810 @GuBenda