ModelFlowIb 2.72__tar.gz → 2.73__tar.gz
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- modelflowib-2.73/PKG-INFO +156 -0
- modelflowib-2.73/README.md +113 -0
- modelflowib-2.73/license.md +10 -0
- {modelflowib-2.72 → modelflowib-2.73}/modelBLfunk.py +1 -1
- {modelflowib-2.72 → modelflowib-2.73}/modelclass.py +203 -19
- {modelflowib-2.72 → modelflowib-2.73}/modelconstruct_estimation.py +2872 -2829
- {modelflowib-2.72 → modelflowib-2.73}/modeldekom.py +15 -8
- {modelflowib-2.72 → modelflowib-2.73}/modelestimator_new.py +2612 -2607
- modelflowib-2.73/modelflowib.egg-info/PKG-INFO +156 -0
- {modelflowib-2.72/ModelFlowIb.egg-info → modelflowib-2.73/modelflowib.egg-info}/SOURCES.txt +8 -17
- {modelflowib-2.72/ModelFlowIb.egg-info → modelflowib-2.73/modelflowib.egg-info}/top_level.txt +1 -13
- modelflowib-2.73/modelgrabgdx.py +584 -0
- modelflowib-2.73/modelhtml.py +606 -0
- {modelflowib-2.72 → modelflowib-2.73}/modelmanipulation.py +4 -4
- {modelflowib-2.72 → modelflowib-2.73}/modelnewton.py +124 -35
- {modelflowib-2.72 → modelflowib-2.73}/modelreport.py +15 -11
- {modelflowib-2.72 → modelflowib-2.73}/modelwidget.py +13 -0
- modelflowib-2.73/pyproject.toml +81 -0
- modelflowib-2.72/ModelFlowIb.egg-info/PKG-INFO +0 -33
- modelflowib-2.72/PKG-INFO +0 -33
- modelflowib-2.72/model_doable.py +0 -40
- modelflowib-2.72/model_ifrs9.py +0 -88
- modelflowib-2.72/model_run_numba.py +0 -15
- modelflowib-2.72/modelclass2.py +0 -463
- modelflowib-2.72/modelfeedback.py +0 -243
- modelflowib-2.72/modelgrabgdx.py +0 -292
- modelflowib-2.72/modelpatternrev.py +0 -72
- modelflowib-2.72/modelsandbox.py +0 -76
- modelflowib-2.72/modelsandbox_Mixin.py +0 -1565
- modelflowib-2.72/modeltodo.py +0 -31
- modelflowib-2.72/modeluserfunk_old.py +0 -90
- modelflowib-2.72/modelwidget_input_old.py +0 -1541
- modelflowib-2.72/modelwidget_shinypy.py +0 -935
- modelflowib-2.72/pyproject.toml +0 -46
- {modelflowib-2.72 → modelflowib-2.73}/model_Excel.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/model_cvx.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/model_dynare.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/model_financial_stability.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/model_latex.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/model_latex_class.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/model_parquet_mixin.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modelconstruct.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modeldash.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modeldashboot.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modeldashsidebar.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modeldiff.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modeldisplay.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modelestimation.py +0 -0
- {modelflowib-2.72/ModelFlowIb.egg-info → modelflowib-2.73/modelflowib.egg-info}/dependency_links.txt +0 -0
- {modelflowib-2.72/ModelFlowIb.egg-info → modelflowib-2.73/modelflowib.egg-info}/requires.txt +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modelgrab.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modelgrabwf2.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modelhelp.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modelinvert.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modeljupyter.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modeljupytermagic.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modelmacrograb.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modelmf.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modelnet.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modelnormalize.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modelpattern.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modeluserfunk.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modelvis.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/modelwidget_input.py +0 -0
- {modelflowib-2.72 → modelflowib-2.73}/setup.cfg +0 -0
|
@@ -0,0 +1,156 @@
|
|
|
1
|
+
Metadata-Version: 2.4
|
|
2
|
+
Name: modelflowib
|
|
3
|
+
Version: 2.73
|
|
4
|
+
Summary: A tool to solve and manage dynamic economic and financial models
|
|
5
|
+
Author-email: Ib Hansen <Ib.Hansen.Iv@gmail.com>
|
|
6
|
+
License-Expression: MIT
|
|
7
|
+
Project-URL: Homepage, https://github.com/IbHansen/ModelFlow2
|
|
8
|
+
Project-URL: Repository, https://github.com/IbHansen/ModelFlow2
|
|
9
|
+
Keywords: economics,dynamic models,BLL,numba,modelling
|
|
10
|
+
Classifier: Programming Language :: Python :: 3
|
|
11
|
+
Classifier: Operating System :: OS Independent
|
|
12
|
+
Requires-Python: >=3.9
|
|
13
|
+
Description-Content-Type: text/markdown
|
|
14
|
+
License-File: license.md
|
|
15
|
+
Requires-Dist: pandas
|
|
16
|
+
Requires-Dist: matplotlib
|
|
17
|
+
Requires-Dist: seaborn
|
|
18
|
+
Requires-Dist: sympy
|
|
19
|
+
Requires-Dist: numpy
|
|
20
|
+
Requires-Dist: networkx
|
|
21
|
+
Requires-Dist: tqdm
|
|
22
|
+
Requires-Dist: scipy
|
|
23
|
+
Requires-Dist: numexpr
|
|
24
|
+
Requires-Dist: bottleneck
|
|
25
|
+
Requires-Dist: XlsxWriter
|
|
26
|
+
Requires-Dist: pyarrow
|
|
27
|
+
Requires-Dist: tabulate
|
|
28
|
+
Requires-Dist: openpyxl
|
|
29
|
+
Requires-Dist: xarray
|
|
30
|
+
Requires-Dist: cvxopt
|
|
31
|
+
Requires-Dist: numba
|
|
32
|
+
Requires-Dist: dash
|
|
33
|
+
Requires-Dist: dash-bootstrap-components
|
|
34
|
+
Requires-Dist: dash-interactive-graphviz
|
|
35
|
+
Requires-Dist: lmfit
|
|
36
|
+
Requires-Dist: jupyter
|
|
37
|
+
Requires-Dist: ipywidgets
|
|
38
|
+
Requires-Dist: ipydatagrid
|
|
39
|
+
Provides-Extra: shiny
|
|
40
|
+
Requires-Dist: shiny; extra == "shiny"
|
|
41
|
+
Requires-Dist: shinywidgets; extra == "shiny"
|
|
42
|
+
Dynamic: license-file
|
|
43
|
+
|
|
44
|
+
# ModelFlow
|
|
45
|
+
A Python toolkit to manage models
|
|
46
|
+
|
|
47
|
+
The easy way to start is here [](https://mybinder.org/v2/gh/IbHansen/modelflow/master).
|
|
48
|
+
|
|
49
|
+
This will start **ModelFlow** as an online Jupyter notebook. Select one of the notebooks (files with the extension .ipynb). Some of the
|
|
50
|
+
notebooks are ModelFlow related. Other implement different model
|
|
51
|
+
|
|
52
|
+
- FRBUS will start the FRB/US model from the Federal reserve board.
|
|
53
|
+
- Q-JEM will start the Quarterly Japanese Economic Model from Bank of Japan
|
|
54
|
+
- ADAM will start the Anual Danish Aggregate Model for the Danish Statistical agency.
|
|
55
|
+
- Solow will implement a simple textybook Solow model.
|
|
56
|
+
|
|
57
|
+
You can also look at the python files of the system. Located at the **modelflow/** folder
|
|
58
|
+
|
|
59
|
+
Alternative look at the *getting started* section later in this file. It explains how to run ModelFlow localy.
|
|
60
|
+
|
|
61
|
+
The **Pandas** library is a great library to handle all kinds of datamanipulation and transformations.
|
|
62
|
+
|
|
63
|
+
However when it comes to models which contains lags or models which requires solving simultanous equations, Pandas is not quite helpful.
|
|
64
|
+
|
|
65
|
+
ModelFlow extends Pandas to handle a range of such models. And they can be large.
|
|
66
|
+
|
|
67
|
+
It requires you to specify the *model* specified as equations (the **business logic**) and place the *data* in a Pandas **DataFrame**. ModelFlow allows the model to meet the data and return the result as a new DataFrame.
|
|
68
|
+
|
|
69
|
+
A number of **analytical tools for model and result analytic** helps to understand the model and its results.
|
|
70
|
+
|
|
71
|
+
The user can **extend and modify the tools** to her or his needs.
|
|
72
|
+
|
|
73
|
+
**Onboarding models and combining models from different sources**. Creating a Macro prudential model often entails recycling several models specified in different ways: Excel, Latex, Dynare, Python or other languages. Python's ecosystem makes it possible to transform many different models into ModelFlow models or to wrap them into functions which can be called from ModelFlow models.
|
|
74
|
+
|
|
75
|
+
**Models can be specified in a high level Business logic language (a Domain Specific language)**. This allows the formulation of a model in a concise and expressive language which is close to the economic of the model. The user can concentrate on the economic or financial content - not the coding of the solution. The code for solving the model is generated by the tool. Then you can *solve the
|
|
76
|
+
simultaneous* (or *non-simultaneous* model) in an efficient way.
|
|
77
|
+
|
|
78
|
+
## Introduction
|
|
79
|
+
|
|
80
|
+
**ModelFlow is written in Python**. Python comes "batteries included" and is
|
|
81
|
+
the basis of a very rich ecosystem, which consists of a wide array of
|
|
82
|
+
libraries. ModelFlow is just another library. It supplements the existing
|
|
83
|
+
libraries regarding modeling language and solving and allows the use of
|
|
84
|
+
Python as a model management framework.
|
|
85
|
+
|
|
86
|
+
**Data handling and wrangling is done in the Pandas library**. This
|
|
87
|
+
library is the Swiss army knife of data science in Python. It can import and export data to most systems and it is very powerful in manipulating and transforming data.
|
|
88
|
+
The core
|
|
89
|
+
element of Pandas is the *Dataframe*. A Dataframe is a two-dimensional
|
|
90
|
+
tabular data structure. Each *column* consists of cells of the same type
|
|
91
|
+
-- it can be a number, a string, a matrix or another Python data object.This includes matrices and other dataframes. Each *row is indexed.* The index can basically be any type of variable
|
|
92
|
+
including dates, which is especially relevant for economic and financial models.
|
|
93
|
+
|
|
94
|
+
**ModelFlow gives the user tools for more than solving models**. This
|
|
95
|
+
includes:
|
|
96
|
+
|
|
97
|
+
- *Visualization* and comparison of results
|
|
98
|
+
|
|
99
|
+
- *Integration* of models from different sources
|
|
100
|
+
|
|
101
|
+
- *Analyze the logical structure of a model*. By applying graph theory,
|
|
102
|
+
ModelFlow can find data lineage, find a suitable calculating sequence and trace
|
|
103
|
+
causes of changes through the calculations.
|
|
104
|
+
|
|
105
|
+
- *Inverting* the model to calculating the necessary instruments to
|
|
106
|
+
achieve a desired target.
|
|
107
|
+
|
|
108
|
+
- Calculating the *attributions* from input to the results of a model.
|
|
109
|
+
|
|
110
|
+
- Calculating the *attribution* from input to the result of each
|
|
111
|
+
formula.
|
|
112
|
+
|
|
113
|
+
- Finding and calculating partial *derivatives* of formulas
|
|
114
|
+
|
|
115
|
+
- *Integrating user defined python functions* in the Business logic
|
|
116
|
+
language (like optimization, calculating risk weights or to make a matrices consistent with the RAS algorithm )
|
|
117
|
+
|
|
118
|
+
- *Wrap matlab* models so they can be used in the Business logic
|
|
119
|
+
language.
|
|
120
|
+
|
|
121
|
+
- *Speed up* solving using "Just in time compilation"
|
|
122
|
+
|
|
123
|
+
- Analyze the model structure through tools from graph theory
|
|
124
|
+
|
|
125
|
+
- Handle *large models.* 1,000,000 formulas is not a problem.
|
|
126
|
+
|
|
127
|
+
- Integrate model management in Jupyter notebooks for *agile and user
|
|
128
|
+
friendly model use*
|
|
129
|
+
|
|
130
|
+
|
|
131
|
+
**The core code of ModelFlow is small and
|
|
132
|
+
documented.** Thus it can easily be modified and expanded to the specific need of the user. *ModelFlow is a toolset*. It can handle models, which conform to the tools.
|
|
133
|
+
|
|
134
|
+
If you need a feature or have a model which can't be handled in ModelFlow,
|
|
135
|
+
you are encouraged to improve ModelFlow. Please share the
|
|
136
|
+
improvement, other users may have the same need, or can be inspired by
|
|
137
|
+
your work.
|
|
138
|
+
|
|
139
|
+
Also bear in mind that ModelFlow is experimental. It is provided ”as is”, without any representation or warranty of any kind either express or implied.
|
|
140
|
+
|
|
141
|
+
## Getting started
|
|
142
|
+
|
|
143
|
+
You need Python 3.7+ with asssociated libraries. The easy way is to install Anaconda Python is to download the distribution from [https://www.anaconda.com/distribution](https://www.anaconda.com/distribution)
|
|
144
|
+
|
|
145
|
+
Now you can install ModelFlow by running a command window from the Anaconda prompt and execute theese commands
|
|
146
|
+
```
|
|
147
|
+
conda install modelflow -c ibh -c defaults -c conda-forge
|
|
148
|
+
```
|
|
149
|
+
You will find the anaconda prompt by searching "anaconda" in the start menu search field
|
|
150
|
+
|
|
151
|
+
Try out one of the workbooks. to do this you:
|
|
152
|
+
```
|
|
153
|
+
click on the Adaconda prompth in the Anaconda folder
|
|
154
|
+
cd <The location you have downloaded the workbooks>
|
|
155
|
+
jupyter notebook
|
|
156
|
+
```
|
|
@@ -0,0 +1,113 @@
|
|
|
1
|
+
# ModelFlow
|
|
2
|
+
A Python toolkit to manage models
|
|
3
|
+
|
|
4
|
+
The easy way to start is here [](https://mybinder.org/v2/gh/IbHansen/modelflow/master).
|
|
5
|
+
|
|
6
|
+
This will start **ModelFlow** as an online Jupyter notebook. Select one of the notebooks (files with the extension .ipynb). Some of the
|
|
7
|
+
notebooks are ModelFlow related. Other implement different model
|
|
8
|
+
|
|
9
|
+
- FRBUS will start the FRB/US model from the Federal reserve board.
|
|
10
|
+
- Q-JEM will start the Quarterly Japanese Economic Model from Bank of Japan
|
|
11
|
+
- ADAM will start the Anual Danish Aggregate Model for the Danish Statistical agency.
|
|
12
|
+
- Solow will implement a simple textybook Solow model.
|
|
13
|
+
|
|
14
|
+
You can also look at the python files of the system. Located at the **modelflow/** folder
|
|
15
|
+
|
|
16
|
+
Alternative look at the *getting started* section later in this file. It explains how to run ModelFlow localy.
|
|
17
|
+
|
|
18
|
+
The **Pandas** library is a great library to handle all kinds of datamanipulation and transformations.
|
|
19
|
+
|
|
20
|
+
However when it comes to models which contains lags or models which requires solving simultanous equations, Pandas is not quite helpful.
|
|
21
|
+
|
|
22
|
+
ModelFlow extends Pandas to handle a range of such models. And they can be large.
|
|
23
|
+
|
|
24
|
+
It requires you to specify the *model* specified as equations (the **business logic**) and place the *data* in a Pandas **DataFrame**. ModelFlow allows the model to meet the data and return the result as a new DataFrame.
|
|
25
|
+
|
|
26
|
+
A number of **analytical tools for model and result analytic** helps to understand the model and its results.
|
|
27
|
+
|
|
28
|
+
The user can **extend and modify the tools** to her or his needs.
|
|
29
|
+
|
|
30
|
+
**Onboarding models and combining models from different sources**. Creating a Macro prudential model often entails recycling several models specified in different ways: Excel, Latex, Dynare, Python or other languages. Python's ecosystem makes it possible to transform many different models into ModelFlow models or to wrap them into functions which can be called from ModelFlow models.
|
|
31
|
+
|
|
32
|
+
**Models can be specified in a high level Business logic language (a Domain Specific language)**. This allows the formulation of a model in a concise and expressive language which is close to the economic of the model. The user can concentrate on the economic or financial content - not the coding of the solution. The code for solving the model is generated by the tool. Then you can *solve the
|
|
33
|
+
simultaneous* (or *non-simultaneous* model) in an efficient way.
|
|
34
|
+
|
|
35
|
+
## Introduction
|
|
36
|
+
|
|
37
|
+
**ModelFlow is written in Python**. Python comes "batteries included" and is
|
|
38
|
+
the basis of a very rich ecosystem, which consists of a wide array of
|
|
39
|
+
libraries. ModelFlow is just another library. It supplements the existing
|
|
40
|
+
libraries regarding modeling language and solving and allows the use of
|
|
41
|
+
Python as a model management framework.
|
|
42
|
+
|
|
43
|
+
**Data handling and wrangling is done in the Pandas library**. This
|
|
44
|
+
library is the Swiss army knife of data science in Python. It can import and export data to most systems and it is very powerful in manipulating and transforming data.
|
|
45
|
+
The core
|
|
46
|
+
element of Pandas is the *Dataframe*. A Dataframe is a two-dimensional
|
|
47
|
+
tabular data structure. Each *column* consists of cells of the same type
|
|
48
|
+
-- it can be a number, a string, a matrix or another Python data object.This includes matrices and other dataframes. Each *row is indexed.* The index can basically be any type of variable
|
|
49
|
+
including dates, which is especially relevant for economic and financial models.
|
|
50
|
+
|
|
51
|
+
**ModelFlow gives the user tools for more than solving models**. This
|
|
52
|
+
includes:
|
|
53
|
+
|
|
54
|
+
- *Visualization* and comparison of results
|
|
55
|
+
|
|
56
|
+
- *Integration* of models from different sources
|
|
57
|
+
|
|
58
|
+
- *Analyze the logical structure of a model*. By applying graph theory,
|
|
59
|
+
ModelFlow can find data lineage, find a suitable calculating sequence and trace
|
|
60
|
+
causes of changes through the calculations.
|
|
61
|
+
|
|
62
|
+
- *Inverting* the model to calculating the necessary instruments to
|
|
63
|
+
achieve a desired target.
|
|
64
|
+
|
|
65
|
+
- Calculating the *attributions* from input to the results of a model.
|
|
66
|
+
|
|
67
|
+
- Calculating the *attribution* from input to the result of each
|
|
68
|
+
formula.
|
|
69
|
+
|
|
70
|
+
- Finding and calculating partial *derivatives* of formulas
|
|
71
|
+
|
|
72
|
+
- *Integrating user defined python functions* in the Business logic
|
|
73
|
+
language (like optimization, calculating risk weights or to make a matrices consistent with the RAS algorithm )
|
|
74
|
+
|
|
75
|
+
- *Wrap matlab* models so they can be used in the Business logic
|
|
76
|
+
language.
|
|
77
|
+
|
|
78
|
+
- *Speed up* solving using "Just in time compilation"
|
|
79
|
+
|
|
80
|
+
- Analyze the model structure through tools from graph theory
|
|
81
|
+
|
|
82
|
+
- Handle *large models.* 1,000,000 formulas is not a problem.
|
|
83
|
+
|
|
84
|
+
- Integrate model management in Jupyter notebooks for *agile and user
|
|
85
|
+
friendly model use*
|
|
86
|
+
|
|
87
|
+
|
|
88
|
+
**The core code of ModelFlow is small and
|
|
89
|
+
documented.** Thus it can easily be modified and expanded to the specific need of the user. *ModelFlow is a toolset*. It can handle models, which conform to the tools.
|
|
90
|
+
|
|
91
|
+
If you need a feature or have a model which can't be handled in ModelFlow,
|
|
92
|
+
you are encouraged to improve ModelFlow. Please share the
|
|
93
|
+
improvement, other users may have the same need, or can be inspired by
|
|
94
|
+
your work.
|
|
95
|
+
|
|
96
|
+
Also bear in mind that ModelFlow is experimental. It is provided ”as is”, without any representation or warranty of any kind either express or implied.
|
|
97
|
+
|
|
98
|
+
## Getting started
|
|
99
|
+
|
|
100
|
+
You need Python 3.7+ with asssociated libraries. The easy way is to install Anaconda Python is to download the distribution from [https://www.anaconda.com/distribution](https://www.anaconda.com/distribution)
|
|
101
|
+
|
|
102
|
+
Now you can install ModelFlow by running a command window from the Anaconda prompt and execute theese commands
|
|
103
|
+
```
|
|
104
|
+
conda install modelflow -c ibh -c defaults -c conda-forge
|
|
105
|
+
```
|
|
106
|
+
You will find the anaconda prompt by searching "anaconda" in the start menu search field
|
|
107
|
+
|
|
108
|
+
Try out one of the workbooks. to do this you:
|
|
109
|
+
```
|
|
110
|
+
click on the Adaconda prompth in the Anaconda folder
|
|
111
|
+
cd <The location you have downloaded the workbooks>
|
|
112
|
+
jupyter notebook
|
|
113
|
+
```
|
|
@@ -0,0 +1,10 @@
|
|
|
1
|
+
Copyright © 2019 Danmarks Nationalbank and ECB
|
|
2
|
+
|
|
3
|
+
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Soft-ware"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
|
|
4
|
+
|
|
5
|
+
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
|
|
6
|
+
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
|
|
7
|
+
|
|
8
|
+
Except as contained in this notice, the name of Danmarks Nationalbank and ECB shall not be used in advertising or otherwise to promote the sale, use or other dealings in this Software without prior written authorization from Danmarks Nationalbank and ECB.
|
|
9
|
+
|
|
10
|
+
Based on MIT License.
|
|
@@ -7,7 +7,7 @@ Created on Fri Mar 2 17:01:49 2018
|
|
|
7
7
|
Functions placed here are included in the Pyfs business language
|
|
8
8
|
|
|
9
9
|
"""
|
|
10
|
-
from math import exp, log, sqrt, tanh, erf
|
|
10
|
+
from math import exp, log, sqrt, tanh, erf,isclose
|
|
11
11
|
from numpy import transpose , array
|
|
12
12
|
from scipy.stats import norm,lognorm
|
|
13
13
|
from scipy.stats import gamma
|
|
@@ -49,6 +49,7 @@ from collections import defaultdict, namedtuple
|
|
|
49
49
|
from itertools import groupby, chain, zip_longest,accumulate
|
|
50
50
|
import re
|
|
51
51
|
import pandas as pd
|
|
52
|
+
from pandas.api.types import is_numeric_dtype
|
|
52
53
|
import sys
|
|
53
54
|
import networkx as nx
|
|
54
55
|
import fnmatch
|
|
@@ -104,6 +105,7 @@ import modelvis as mv
|
|
|
104
105
|
import modelpattern as pt
|
|
105
106
|
from modelnet import draw_adjacency_matrix
|
|
106
107
|
from modelnewton import newton_diff
|
|
108
|
+
from modelsolver_ng import Solver_ng_Mixin
|
|
107
109
|
import modeljupyter as mj
|
|
108
110
|
from modelhelp import cutout, update_var
|
|
109
111
|
from modelnormalize import normal
|
|
@@ -146,7 +148,7 @@ class BaseModel():
|
|
|
146
148
|
tabcomplete=True, previousbase=False, use_preorder=True, normalized=True,safeorder= False,
|
|
147
149
|
var_description={}, model_description = '',
|
|
148
150
|
var_groups = {}, reports = {}, equations_latex='',
|
|
149
|
-
eviews_dict = {},use_fbmin=
|
|
151
|
+
eviews_dict = {},use_fbmin= True,
|
|
150
152
|
substitution = {},
|
|
151
153
|
**kwargs):
|
|
152
154
|
''' initialize a model'''
|
|
@@ -1405,11 +1407,17 @@ class Org_model_Mixin():
|
|
|
1405
1407
|
|
|
1406
1408
|
def __getattr__(self, name):
|
|
1407
1409
|
'''To execute the . operator
|
|
1408
|
-
|
|
1409
|
-
|
|
1410
|
-
|
|
1410
|
+
|
|
1411
|
+
|
|
1412
|
+
|
|
1411
1413
|
'''
|
|
1412
|
-
|
|
1414
|
+
# Guard against infinite recursion: __getattr__ is only called when normal
|
|
1415
|
+
# lookup fails. Dunder/internal names and a not-yet-initialised (or empty)
|
|
1416
|
+
# model lack 'allvar_set'; reading it via self.allvar_set would re-enter
|
|
1417
|
+
# __getattr__ endlessly, so resolve it from __dict__ directly.
|
|
1418
|
+
if name.startswith('__') and name.endswith('__'):
|
|
1419
|
+
raise AttributeError(name)
|
|
1420
|
+
|
|
1413
1421
|
if name.startswith('model_var_'):
|
|
1414
1422
|
wishfor= name[len('model_var_'):]
|
|
1415
1423
|
return(self.var_with_frmlname(wishfor))
|
|
@@ -1418,10 +1426,11 @@ class Org_model_Mixin():
|
|
|
1418
1426
|
return(self.frml_with_frmlname(wishfor))
|
|
1419
1427
|
|
|
1420
1428
|
|
|
1421
|
-
else:
|
|
1422
|
-
|
|
1429
|
+
else:
|
|
1430
|
+
allvar_set = self.__dict__.get('allvar_set')
|
|
1431
|
+
if allvar_set is not None and name.upper() in allvar_set:
|
|
1423
1432
|
return mv.varvis(model=self, var=name.upper())
|
|
1424
|
-
else:
|
|
1433
|
+
else:
|
|
1425
1434
|
raise AttributeError(f'The specification:"{name}" did not match a method, property or variable name')
|
|
1426
1435
|
|
|
1427
1436
|
def __dir__(self):
|
|
@@ -4778,10 +4787,10 @@ class Display_Mixin():
|
|
|
4778
4787
|
.set_sticky(axis='index')\
|
|
4779
4788
|
.set_table_attributes('class="table"')
|
|
4780
4789
|
|
|
4781
|
-
if
|
|
4782
|
-
result = result
|
|
4783
|
-
else:
|
|
4790
|
+
if all(is_numeric_dtype(dt) for dt in df.dtypes):
|
|
4784
4791
|
result = result.format('{:,.'+xdec+'f}'+xpct)
|
|
4792
|
+
else:
|
|
4793
|
+
result = result
|
|
4785
4794
|
|
|
4786
4795
|
if use_tooltip:
|
|
4787
4796
|
try:
|
|
@@ -4843,10 +4852,14 @@ class Display_Mixin():
|
|
|
4843
4852
|
|
|
4844
4853
|
result = df.style.set_table_styles(styles)
|
|
4845
4854
|
|
|
4846
|
-
|
|
4847
|
-
|
|
4848
|
-
|
|
4855
|
+
# Only apply the numeric format when every column is numeric.
|
|
4856
|
+
# NB: pandas >=3.0 stores strings as the 'str' dtype (not 'object'),
|
|
4857
|
+
# so a "not object" test wrongly treats pre-formatted string tables
|
|
4858
|
+
# (df_str) as numeric and crashes formatting them. Test for numeric.
|
|
4859
|
+
if all(is_numeric_dtype(dt) for dt in df.dtypes):
|
|
4849
4860
|
result = result.format('{:,.'+xdec+'f}'+xpct+' ')
|
|
4861
|
+
else:
|
|
4862
|
+
result = result
|
|
4850
4863
|
|
|
4851
4864
|
if use_tooltip:
|
|
4852
4865
|
try:
|
|
@@ -6460,6 +6473,8 @@ class Solver_Mixin():
|
|
|
6460
6473
|
|
|
6461
6474
|
solver : str, optional
|
|
6462
6475
|
Specifies the solver to be used. The default solver is chosen based on the model's properties.
|
|
6476
|
+
A next-generation (modelsolver_ng) solver can be selected by passing its
|
|
6477
|
+
``<name>_ng`` name, e.g. ``solver='newton_ng'`` or ``solver='sim_ng'``.
|
|
6463
6478
|
|
|
6464
6479
|
silent : bool, optional
|
|
6465
6480
|
If True, the solver runs silently without printing output to the console. Default is True.
|
|
@@ -6537,7 +6552,7 @@ class Solver_Mixin():
|
|
|
6537
6552
|
solver = newkwargs.get('solver', solverguess)
|
|
6538
6553
|
silent = newkwargs.get('silent', True)
|
|
6539
6554
|
self.model_solver = getattr(self, solver)
|
|
6540
|
-
|
|
6555
|
+
|
|
6541
6556
|
# print(f'solver:{solver},solverkwargs:{newkwargs}')
|
|
6542
6557
|
# breakpoint()
|
|
6543
6558
|
outdf = self.model_solver(*args, **newkwargs)
|
|
@@ -6638,7 +6653,9 @@ class Solver_Mixin():
|
|
|
6638
6653
|
# breakpoint()
|
|
6639
6654
|
# if we import from a cache, we assume that the dataframe is in the same order
|
|
6640
6655
|
if transpile_reset or not hasattr(self, f'pro_{jitname}'):
|
|
6641
|
-
|
|
6656
|
+
# sanitize: the model name can contain characters that
|
|
6657
|
+
# are illegal in a module name / file path (e.g. 'FRB/US')
|
|
6658
|
+
jitfilename= f"modelsource/{re.sub(r'[^0-9a-zA-Z_]', '_', jitname)}_jitsolver.py"
|
|
6642
6659
|
jitfile = Path(jitfilename)
|
|
6643
6660
|
jitfile.parent.mkdir(parents=True, exist_ok=True)
|
|
6644
6661
|
if not silent:
|
|
@@ -9345,7 +9362,7 @@ class Report_Mixin:
|
|
|
9345
9362
|
|
|
9346
9363
|
from model_parquet_mixin import Parquet_Mixin
|
|
9347
9364
|
|
|
9348
|
-
class model(Parquet_Mixin, Zip_Mixin, Json_Mixin, Model_help_Mixin, Solver_Mixin, Display_Mixin, Graph_Draw_Mixin, Graph_Mixin,
|
|
9365
|
+
class model(Parquet_Mixin, Zip_Mixin, Json_Mixin, Model_help_Mixin, Solver_Mixin, Solver_ng_Mixin, Display_Mixin, Graph_Draw_Mixin, Graph_Mixin,
|
|
9349
9366
|
Dekomp_Mixin, Org_model_Mixin, BaseModel, Description_Mixin, Excel_Mixin, Dash_Mixin, Modify_Mixin,
|
|
9350
9367
|
Fix_Mixin,Stability_Mixin,Report_Mixin):
|
|
9351
9368
|
'''This is the main model definition'''
|
|
@@ -9633,8 +9650,7 @@ frml <CALC_ADJUST> b_a = a-(c+b)$'''
|
|
|
9633
9650
|
# zz = model.modelload(r'https://raw.githubusercontent.com/IbHansen/Modelflow2/master/Examples/ADAM/baseline.pcim')
|
|
9634
9651
|
mpak,baseline = model.modelload(r'pak.pcim',run=1,use_fbmin=True)
|
|
9635
9652
|
alternative = baseline.upd("<2020 2100> PAKGGREVCO2CER PAKGGREVCO2GER PAKGGREVCO2OER = 30")
|
|
9636
|
-
result = mpak(alternative,2020,2100,keep='Carbon tax nominal 30',silent=0) # simulates the model
|
|
9637
|
-
diff_level, att_level, att_pct, diff_growth, att_growth = tup = mpak.dekomp('PAKNECONOTHRXN',lprint=False,start=2020,end=2027)
|
|
9653
|
+
result = mpak(alternative,2020,2100,keep='Carbon tax nominal 30',silent=0,solver='sim',nonlin=6) # simulates the model
|
|
9638
9654
|
|
|
9639
9655
|
|
|
9640
9656
|
fb_nodes,dag_nodes = mpak.get_minimal_feedback_set(mpak.endograph)
|
|
@@ -9642,3 +9658,171 @@ frml <CALC_ADJUST> b_a = a-(c+b)$'''
|
|
|
9642
9658
|
print(f'Number of variables in "minimum" feedback set: {len(fb_nodes)}')
|
|
9643
9659
|
print(f'Number of variables in "DAG" graph : {len(dag_nodes)}')
|
|
9644
9660
|
# mtest,bk = model.modelload(r'ibstestnozip')
|
|
9661
|
+
|
|
9662
|
+
#%% test next-generation (ng) solvers against the legacy solvers
|
|
9663
|
+
if 1:
|
|
9664
|
+
ngmodel = model(smallmodel)
|
|
9665
|
+
ngdf = pd.DataFrame(
|
|
9666
|
+
{'X': [0.2, 0.2, 0.2], 'C': [2.0, 3., 5.], 'R': [2., 1., 0.4],
|
|
9667
|
+
'PPP': [0.4, 0., 0.4]})
|
|
9668
|
+
|
|
9669
|
+
# Gauss-Seidel: legacy .sim vs new .sim_ng
|
|
9670
|
+
old_sim = ngmodel.sim(ngdf, silent=1)
|
|
9671
|
+
new_sim = ngmodel.sim_ng(ngdf, silent=1)
|
|
9672
|
+
d_sim = float((old_sim - new_sim).abs().max().max())
|
|
9673
|
+
print(f'sim vs sim_ng max abs diff: {d_sim:.3e}')
|
|
9674
|
+
|
|
9675
|
+
# 1-D stuffed Gauss-Seidel: legacy .sim1d vs new .sim1d_ng
|
|
9676
|
+
old_sim1d = ngmodel.sim1d(ngdf, silent=1)
|
|
9677
|
+
new_sim1d = ngmodel.sim1d_ng(ngdf, silent=1)
|
|
9678
|
+
d_sim1d = float((old_sim1d - new_sim1d).abs().max().max())
|
|
9679
|
+
print(f'sim1d vs sim1d_ng max abs diff: {d_sim1d:.3e}')
|
|
9680
|
+
|
|
9681
|
+
# Newton needs a well-posed simultaneous model: the smallmodel above is
|
|
9682
|
+
# singular for D3 (D3 = ... + D3 -> zero on the Jacobian diagonal once the
|
|
9683
|
+
# -I term is applied), so use a small non-degenerate linear one here.
|
|
9684
|
+
# No recursive tail on purpose: per-period Newton solves only the
|
|
9685
|
+
# simultaneous core (endovar = coreorder), so a recursive var would be left
|
|
9686
|
+
# unsolved and legitimately disagree with the stacked solver (all endo).
|
|
9687
|
+
goodmodel = model('frml <> a = 0.5*b + x $ '
|
|
9688
|
+
'frml <> b = 0.3*a + c $ ')
|
|
9689
|
+
gdf = pd.DataFrame({'X': [1.0, 2., 3.], 'C': [2.0, 2., 2.]})
|
|
9690
|
+
|
|
9691
|
+
# unified per-period Newton: legacy .newton_implicit vs new .newton_ng
|
|
9692
|
+
old_newton = goodmodel.newton_implicit(gdf, silent=1)
|
|
9693
|
+
new_newton = goodmodel.newton_ng(gdf, silent=1)
|
|
9694
|
+
d_newton = float((old_newton - new_newton).abs().max().max())
|
|
9695
|
+
print(f'newton_implicit vs newton_ng max abs diff: {d_newton:.3e}')
|
|
9696
|
+
|
|
9697
|
+
# unified stacked Newton: legacy .newtonstack_implicit vs new .newtonstack_ng
|
|
9698
|
+
old_stack = goodmodel.newtonstack_implicit(gdf, silent=1)
|
|
9699
|
+
new_stack = goodmodel.newtonstack_ng(gdf, silent=1)
|
|
9700
|
+
d_stack = float((old_stack - new_stack).abs().max().max())
|
|
9701
|
+
print(f'newtonstack_implicit vs newtonstack_ng max abs diff: {d_stack:.3e}')
|
|
9702
|
+
|
|
9703
|
+
# 'newtonstack_implicit_ng' is an alias for 'newtonstack_ng'
|
|
9704
|
+
d_alias = float((goodmodel.newtonstack_implicit_ng(gdf, silent=1)
|
|
9705
|
+
- new_stack).abs().max().max())
|
|
9706
|
+
|
|
9707
|
+
# per-period and stacked Newton must reach the same (well-posed) solution
|
|
9708
|
+
d_cross = float((new_newton - new_stack).abs().max().max())
|
|
9709
|
+
print(f'newton_ng vs newtonstack_ng max abs diff: {d_cross:.3e}')
|
|
9710
|
+
|
|
9711
|
+
# single-pass DAG evaluator: legacy .xgenr vs new .xgenr_ng. (This toy
|
|
9712
|
+
# model is actually simultaneous, so xgenr's single sweep is not a real
|
|
9713
|
+
# solution -- but both implementations must produce the SAME sweep.)
|
|
9714
|
+
old_xgenr = ngmodel.xgenr(ngdf, silent=1)
|
|
9715
|
+
new_xgenr = ngmodel.xgenr_ng(ngdf, silent=1)
|
|
9716
|
+
d_xgenr = float((old_xgenr - new_xgenr).abs().max().max())
|
|
9717
|
+
print(f'xgenr vs xgenr_ng max abs diff: {d_xgenr:.3e}')
|
|
9718
|
+
|
|
9719
|
+
# dispatcher form (well-posed model)
|
|
9720
|
+
disp = goodmodel.solve_ng(gdf, solver='newton', silent=1)
|
|
9721
|
+
d_disp = float((disp - new_newton).abs().max().max())
|
|
9722
|
+
print(f'solve_ng(newton) matches newton_ng : {d_disp < 1e-9}')
|
|
9723
|
+
|
|
9724
|
+
# __call__ can select an ng solver via its '<name>_ng' name, with the
|
|
9725
|
+
# unchanged getattr(self, solver) dispatch. The surrounding __call__
|
|
9726
|
+
# behaviour is identical, so solver='sim' and solver='sim_ng' must match.
|
|
9727
|
+
call_classic = ngmodel(ngdf, solver='sim', silent=1, reset_options=True)
|
|
9728
|
+
call_ng = ngmodel(ngdf, solver='sim_ng', silent=1, reset_options=True)
|
|
9729
|
+
d_call = float((call_classic - call_ng).abs().max().max())
|
|
9730
|
+
print(f"__call__(solver='sim') vs (solver='sim_ng') max abs diff: {d_call:.3e}")
|
|
9731
|
+
|
|
9732
|
+
# opt-in residual recording: keep_residual=True populates model.ng_residual
|
|
9733
|
+
# (Gauss builds a no-jit res model only then; default solves skip it).
|
|
9734
|
+
goodmodel.sim_ng(gdf, silent=1, keep_residual=True)
|
|
9735
|
+
r_gauss = float(goodmodel.ng_residual.abs().max().max())
|
|
9736
|
+
goodmodel.newton_ng(gdf, silent=1, keep_residual=True)
|
|
9737
|
+
r_newton = float(goodmodel.ng_residual.abs().max().max())
|
|
9738
|
+
print(f'keep_residual: sim_ng max|F(y)-y| {r_gauss:.3e}, newton_ng {r_newton:.3e}')
|
|
9739
|
+
|
|
9740
|
+
assert (d_sim < 1e-6 and d_sim1d < 1e-6 and d_xgenr < 1e-6
|
|
9741
|
+
and d_newton < 1e-6 and d_stack < 1e-6 and d_alias < 1e-12
|
|
9742
|
+
and d_cross < 1e-6 and d_call < 1e-6
|
|
9743
|
+
and r_gauss < 1e-4 and r_newton < 1e-4), \
|
|
9744
|
+
'ng solver diverges from legacy'
|
|
9745
|
+
print('ng solver test: OK')
|
|
9746
|
+
|
|
9747
|
+
#%% experiment: solve the PAK carbon-tax scenario with each ng solver and
|
|
9748
|
+
# check every solver reaches the same solution as the default solve.
|
|
9749
|
+
if 1:
|
|
9750
|
+
mpak,baseline = model.modelload(r'pak.pcim',run=1,use_fbmin=True,ljit=False)
|
|
9751
|
+
# The reference is the default solve computed above:
|
|
9752
|
+
# result = mpak(alternative, 2020, 2100, keep='Carbon tax nominal 30')
|
|
9753
|
+
# Excluded: 'res' (an internal residual-codegen flavour, not a stand-alone
|
|
9754
|
+
# solver) and 'xgenr' (a single topological sweep -- only valid for a DAG,
|
|
9755
|
+
# not this simultaneous model).
|
|
9756
|
+
# ('newtonstack_implicit' is an alias of 'newtonstack' -- same solver.)
|
|
9757
|
+
ng_experiment = ['sim', 'sim1d', 'newton', 'newtonstack','newton_fbmin','newtonstack_fbmin']
|
|
9758
|
+
|
|
9759
|
+
# The Newton-family solvers refresh the Jacobian periodically (nonlin) for
|
|
9760
|
+
# this nonlinear model; Gauss-Seidel needs no options. Any converged solver
|
|
9761
|
+
# must reach the same fixed point as the default solve.
|
|
9762
|
+
newton_opts = dict(nonlin=6)
|
|
9763
|
+
solver_opts = {
|
|
9764
|
+
'sim': {}, 'sim1d': {},
|
|
9765
|
+
'newton': newton_opts, 'newtonstack': newton_opts,
|
|
9766
|
+
'newton_fbmin': newton_opts, 'newtonstack_fbmin': newton_opts
|
|
9767
|
+
}
|
|
9768
|
+
|
|
9769
|
+
endo = sorted(mpak.endogene)
|
|
9770
|
+
ref = result[endo] # default (auto-guessed) solution, endo columns
|
|
9771
|
+
|
|
9772
|
+
print('\nPAK carbon-tax scenario -- ng solver experiment (vs default solve):')
|
|
9773
|
+
print(f'{"solver":26s}{"max |diff|":>16s}{"max rel diff":>16s} result')
|
|
9774
|
+
diffs = [] # converged but disagrees -> real problem
|
|
9775
|
+
failed = [] # raised / did not converge -> solver limitation
|
|
9776
|
+
for sname in ng_experiment:
|
|
9777
|
+
try:
|
|
9778
|
+
r = mpak(alternative, 2020, 2100, solver=f'{sname}_ng',
|
|
9779
|
+
silent=1, reset_options=True,ljit=False,
|
|
9780
|
+
keep=f'Carbon tax 30 [{sname}_ng]',
|
|
9781
|
+
**solver_opts[sname])
|
|
9782
|
+
except Exception as e:
|
|
9783
|
+
print(f'{sname + "_ng":26s}{"solve raised":>32s} {type(e).__name__}')
|
|
9784
|
+
failed.append(f'{sname} ({type(e).__name__})')
|
|
9785
|
+
continue
|
|
9786
|
+
cur = r[endo]
|
|
9787
|
+
absdiff = (cur - ref).abs()
|
|
9788
|
+
maxabs = float(absdiff.max().max())
|
|
9789
|
+
relerr = float((absdiff / (ref.abs() + 1e-8)).max().max())
|
|
9790
|
+
same = np.allclose(cur.values, ref.values, rtol=1e-2, atol=1e-2)
|
|
9791
|
+
print(f'{sname + "_ng":26s}{maxabs:16.3e}{relerr:16.3e} '
|
|
9792
|
+
f'{"same" if same else "DIFFERENT"}')
|
|
9793
|
+
if not same:
|
|
9794
|
+
diffs.append(f'{sname} (max rel diff {relerr:.2e})')
|
|
9795
|
+
|
|
9796
|
+
# Report only -- this is a real-world demonstration, not a unit test.
|
|
9797
|
+
# Gauss-Seidel (sim/sim1d) must match the default solve; the Newton family
|
|
9798
|
+
# should match too now that per-period Newton also solves the recursive
|
|
9799
|
+
# prolog/epilog blocks. Only Gauss-Seidel mismatches are asserted.
|
|
9800
|
+
gs_problems = [d for d in diffs if d.split()[0] in ('sim', 'sim1d')]
|
|
9801
|
+
if failed:
|
|
9802
|
+
print(f'PAK solver experiment: did not converge / raised: {failed}')
|
|
9803
|
+
if diffs:
|
|
9804
|
+
print(f'PAK solver experiment: differ from default solve: {diffs}')
|
|
9805
|
+
if not failed and not diffs:
|
|
9806
|
+
print('PAK solver experiment: all ng solvers match the default solve')
|
|
9807
|
+
assert not gs_problems, \
|
|
9808
|
+
f'Gauss-Seidel ng solver differs from default solve: {gs_problems}'
|
|
9809
|
+
#%%
|
|
9810
|
+
with mpak.timer('load') :
|
|
9811
|
+
mpak,baseline = model.modelload(r'pak.pcim',run=1,use_fbmin=True,ljit=False,solver='sim1d')
|
|
9812
|
+
with mpak.timer('update'):
|
|
9813
|
+
alternative = baseline.upd("<2020 2100> PAKGGREVCO2CER PAKGGREVCO2GER PAKGGREVCO2OER = 30")
|
|
9814
|
+
#%%
|
|
9815
|
+
with mpak.timer('newton_ng'):
|
|
9816
|
+
result = mpak(alternative,2020,2100,keep='Carbon tax nominal 30',silent=0,solver='newton_ng',nonlin=20,max_iterations=100,ljit=True,jacobian='fd') # simulates the model
|
|
9817
|
+
with mpak.timer('newton_ng jit'):
|
|
9818
|
+
result = mpak(alternative,2020,2100,keep='Carbon tax nominal 30',silent=1,solver='newton_fbmin_ng',max_iterations=100,ljit=True,jacobian='fd') # simulates the model
|
|
9819
|
+
with mpak.timer('sim'):
|
|
9820
|
+
result = mpak(alternative,2020,2100,keep='Carbon tax nominal 30',silent=1,solver='sim_ng',nonlin=23,ljit=False) # simulates the model
|
|
9821
|
+
with mpak.timer('sim jit'):
|
|
9822
|
+
result = mpak(alternative,2020,2100,keep='Carbon tax nominal 30',silent=1,solver='sim_ng',nonlin=23,ljit=True) # simulates the model
|
|
9823
|
+
#%%
|
|
9824
|
+
with mpak.timer('newtonstack_fbmin_ng'):
|
|
9825
|
+
# reset_options: __call__ options are sticky, so jacobian='fd' from the
|
|
9826
|
+
# newton_fbmin cells above would otherwise leak in (a fd build here is
|
|
9827
|
+
# n_fb_stacked+1 full-span sweeps -- minutes of silence)
|
|
9828
|
+
result = mpak(alternative,2020,2100,keep='Carbon tax nominal 30',silent=1,solver='newtonstack_fbmin_ng',nonlin=5,max_iterations=100,ljit=False,reset_options=True) # simulates the model
|