- Solve issue on which nested initial states are being copied from one run to another (#357);
- Created add_parameter_labels (#360);
- Add supress_print attribute to the Executor class for silencing prints and tqdm (#361);
- fix: easy supports list/array/sequence type as parameter (#348);
- Update setup.py to include tqdm package (#342);
- Debug M dictionairy unpacking in utils (#343);
- Add test for counting the parameters argument inside policies and SUFs (#345);
- Fix sweep_cartesian_product such that it works nicely with easy_run (#355);
- Update easy_run.py (#356);
- Fixed a bug where the single_proc mode would result in a missing simulation key (#332);
- Fixed a bug where the single_proc mode would not return results (#335);
- Fixed a wrong parameter order when doing single runs (#337).
- Added a Windows CI pipeline;
- Added toggle for enabling users to deactivate deepcopying. This is done by passing an additional object on the
ExecutionContext
, eg.ExecutionContext(mode, additional_objs={'deepcopy_off': True})
- A collection of type annotations for encapsuling
cadCAD
projects is now implemented through thecadCAD.types
submodules - Added
cadCAD.tools
as a submodule, which is originated from thecadCAD_tools
Python package. This submodule contains several helper functions for making the simulation experience more straightforward as well as a collection of performance profiling tools. - Added
cadCAD.diagram
as a submodule, which is originated from thecadCAD_diagram
Python package. This submodule contains functions for programatically generating block diagrams from existing models. - More informative error messages when policies and SUFs are wrongly implemented. (Issues #288 and #258)
- Merged repo with the
cadCAD_legacy_devel
, which includes performance improvements. In particular, simulations will start up faster due to code optimizations. cadCAD
now usespytest
as the testing framework. This was made possible by isolating the existing tests and wrapping them into functions.
- cadCAD is now Python 3.10+ compatible (Issue #306 and #301)
- Proper support for
ExecutionMode.single_mode
(Issue #253 and #254)
- ver. ≥
0.4.28
:-
-
- Configurations (
cadCAD.utils.Configuration
's) are now accessed via theconfigs
member ofcadCAD.configuration.Experiment
. Example:cadCAD.configuration.Experiment().configs
cadCAD.configs
has been re-included for backwards compatibility and has been assignedcadCAD.experiment.configs
- Configurations (
-
cadCAD.configuration.Experiment()
is unique representation of an experiment of one or more configured System Models.- The
cadCAD
module now contains a default Experiment objectcadCAD.experiment
as an instantiation ofcadCAD.configuration.Experiment()
- The
- An
Experiment
'sappend_model
method stores multiple system modelConfiguration
's for simulation execution withincadCAD.configuration.Experiment().configs
.cadCAD.configuration.Experiment().model_ids
contains system model labels and/or indexes forcadCAD.configuration.Experiment().configs
- The
Experiment
'sappend_model
method is defined withmodel_id
parameter that accepts a system model label.- If duplicate
model_id
's are specified, an index is appended to the label after the@
symbol. (Example:cadCAD.configuration.Experiment().model_ids = ['sys_model', 'sys_model@1', 'sys_model@2', ...]
) - If
model_id
's are not specified or duplicate, the label is auto-generated as a string indicating the system model index withincadCAD.configuration.Experiment().configs
. (Example of unspecified system models at indexes 1, 3, and 4:cadCAD.configuration.Experiment().model_ids = ['sys_model', '1', 'sys_model@2', '3', '4', ...]
)
- If duplicate
- The
-
-
Upgrade Guide: specific to feature changes / additions
-
- ver. ≥
0.4.27
:-
-
- Configurations (
cadCAD.utils.Configuration
's) as are no longer a part of thecadCAD
module (ascadCAD.configs
) and are now accessed via theconfigs
member ofcadCAD.configuration.Experiment
. Example:cadCAD.configuration.Experiment().configs
- Configurations (
-
cadCAD.configuration.Experiment
is unique representation of an experiment of one or more configured System Models. AnExperiment
'sappend_model
method stores multiple system modelConfiguration
's for simulation execution.- The
Experiment
'sappend_model
method requires amodel_id
parameter that is auto-created as'sys_model_#'
. - Requirements:
- Users must use different
model_id
's when appending multiple System Model Configurations. - Users can no longer use the
config_list
method ofcadCAD.configuration.Experiment
- Users must use different
- Backwards Compatibility: The
append_model
method ofcadCAD.configuration.Experiment
can also be used as theappend_configs
method.
- The
-
- Removed Nix
-
Upgrade Guide: specific to feature changes / additions
-
- Fixes:
- #248
- The previous release was returning partial results. An A/B test for this has been included and will be for future releases
- #242
- Parallelized simulations enabled with the re-inclusion of
ProcessPool
.
- Parallelized simulations enabled with the re-inclusion of
- #257
- ValueError for runs accepted by the
cadCAD.configuration.Experiment().append_model
via thesim_configs
no longer gives mis-leading error message if catching a non-related ValueError
- ValueError for runs accepted by the
- #252
- Jupyter lab and Jupyter notebook recognises cadCAD module
- #248
- ver. ≥
0.4.23
:- Hot-Fix: #203 (No Breaking Changes)
- Multi - System Model simulation results will no longer return truncated results (exclude the results of the last
cadCAD.configuration.Configuration
appended tocadCAD.configs
). - Issue: #195
- Multi - System Model simulation results will no longer return truncated results (exclude the results of the last
- Parameter Sweep value
M
(Params) requires up to a maximum of 2 distinct lengths
- Hot-Fix: #203 (No Breaking Changes)
cadCAD.configuration.Experiment
(Alpha) is in development and needed to be released to support the implementation of web applications and proprietary feature extensions. It is intended to represent a unique identifier of an experiment of one or more configured System Models. For this reason,append_configs
is a method ofcadCAD.configuration.Experiment
. As of now it does not support multi - system model simulation because configurations are still appended globally despiteappend_config
being a method of Experiment.
- ver. ≥
0.4.22
:from cadCAD.configuration import Experiment exp = Experiment() exp.append_configs(...)
- ver.
0.3.1
: Deprecatedfrom cadCAD.configuration import append_configs append_configs(...)
- Bug Fix: Multiprocessing error for Windows
- Local Execution Mode (Default): Implicit parallelization of Monte-Carlo / Stochastic simulations (Automatically
selects Multi-Threaded Mode if simulations are configured for more than a single run)
- Backwards Compatibility:
cadCAD.engine.ExecutionMode
accepts legacy execution modes from ver.0.3.1
- Backwards Compatibility:
-
ver. ≥
0.4.22
:from cadCAD.engine import ExecutionMode, ExecutionContext exec_mode = ExecutionMode() local_ctx = ExecutionContext(context=exec_mode.local_mode)
-
ver.
0.3.1
:Multi-Threaded:
from cadCAD.engine import ExecutionMode, ExecutionContext exec_mode = ExecutionMode() single_ctx = ExecutionContext(context=exec_mode.multi_proc)
Single-Thread:
from cadCAD.engine import ExecutionMode, ExecutionContext exec_mode = ExecutionMode() multi_ctx = ExecutionContext(context=exec_mode.single_proc)
- Single Result Dataset
as a 2 dimensional
list
- Returns a single dataset instead of multiple datasets per Monte Carlo simulation as in
0.3.1
:- New System Metrics as dataset attributes:
- Simulation (Alpha) is a unique identifier being developed to represent Experiments as stated above and
will be renamed accordingly
- Subset is a unique identifier of Monte-Carlo simulations produced by parameter sweeps
- Simulation (Alpha) is a unique identifier being developed to represent Experiments as stated above and
will be renamed accordingly
- New System Metrics as dataset attributes:
- Note: Returning a single dataset was originally specified during the project’s inception instead of multiple per simulation
- Returns a single dataset instead of multiple datasets per Monte Carlo simulation as in
-
ver. ≥
0.4.22
:import pandas as pd from tabulate import tabulate from cadCAD.engine import ExecutionMode, ExecutionContext, Executor import system_model_A, system_model_B from cadCAD import configs exec_mode = ExecutionMode() local_ctx = ExecutionContext(context=exec_mode.local_mode) simulation = Executor(exec_context=local_ctx, configs=configs) raw_result, sys_model, _ = simulation.execute() result = pd.DataFrame(raw_result) print(tabulate(result, headers='keys', tablefmt='psql'))
Results:
+----+------------+-----------+----+---------------------+------------+--------+-----+---------+----------+ | | s1 | s2 | s3 | timestamp | simulation | subset | run | substep | timestep | |----+------------+-----------+----+---------------------+------------+--------+-----+---------+----------| | 0 | 0.0 | 0.0 | 1 | 2018-10-01 15:16:24 | 0 | 0 | 1 | 0 | 0 | | 1 | 1.0 | 4 | 5 | 2018-10-01 15:16:25 | 0 | 0 | 1 | 1 | 1 | | 2 | 2.0 | 6 | 5 | 2018-10-01 15:16:25 | 0 | 0 | 1 | 2 | 1 | | 3 | 3.0 | [ 30 300] | 5 | 2018-10-01 15:16:25 | 0 | 0 | 1 | 3 | 1 | | 4 | 0 | 0 | 1 | 2018-10-01 15:16:24 | 1 | 0 | 1 | 0 | 0 | | 5 | 1 | 0 | 5 | 2018-10-01 15:16:25 | 1 | 0 | 1 | 1 | 1 | | 6 | a | 0 | 5 | 2018-10-01 15:16:25 | 1 | 0 | 1 | 2 | 1 | | 7 | ['c', 'd'] | [ 30 300] | 5 | 2018-10-01 15:16:25 | 1 | 0 | 1 | 3 | 1 | +----+------------+-----------+----+---------------------+------------+--------+-----+---------+----------+
-
ver.
0.3.1
:import pandas as pd from tabulate import tabulate from cadCAD.engine import ExecutionMode, ExecutionContext, Executor import system_model_A, system_model_B from cadCAD import configs exec_mode = ExecutionMode() multi_ctx = ExecutionContext(context=exec_mode.multi_proc) simulation = Executor(exec_context=multi_ctx, configs=configs) i = 0 config_names = ['sys_model_A', 'sys_model_B'] for raw_result, _ in simulation.execute(): result = pd.DataFrame(raw_result) print() print(f"{config_names[i]} Result: System Events DataFrame:") print(tabulate(result, headers='keys', tablefmt='psql')) print() i += 1
Results:
+----+------------+-----------+----+---------------------+-----+---------+----------+ | | s1 | s2 | s3 | timestamp | run | substep | timestep | |----+------------+-----------+----+---------------------+-----+---------+----------| | 0 | 0.0 | 0.0 | 1 | 2018-10-01 15:16:24 | 1 | 0 | 0 | | 1 | 1.0 | 4 | 5 | 2018-10-01 15:16:25 | 1 | 1 | 1 | | 2 | 2.0 | 6 | 5 | 2018-10-01 15:16:25 | 1 | 2 | 1 | | 3 | 3.0 | [ 30 300] | 5 | 2018-10-01 15:16:25 | 1 | 3 | 1 | | 4 | 0 | 0 | 1 | 2018-10-01 15:16:24 | 1 | 0 | 0 | | 5 | 1 | 0 | 5 | 2018-10-01 15:16:25 | 1 | 1 | 1 | | 6 | a | 0 | 5 | 2018-10-01 15:16:25 | 1 | 2 | 1 | | 7 | ['c', 'd'] | [ 30 300] | 5 | 2018-10-01 15:16:25 | 1 | 3 | 1 | +----+------------+-----------+----+---------------------+-----+---------+----------+
-
Flattened Configuration list: The
cadCAD.configs
(System Model Configuration)list
has been temporarily flattened to contain single runcadCAD.configuration.Configuration
objects to both fault-tolerant simulation and elastic workloads. This functionality will be restored in a subsequent release by a class that returnscadCAD.configs
's original representation in ver.0.3.1
.- The conversion utilities have been provided to restore its original representation of configurations with
runs >= 1
- System Configuration Conversions
- Configuration as list of Configuration Objects (as in ver.
0.3.1
) - New: System Configuration as a Pandas DataFrame
- New: System Configuration as list of Dictionaries
- Configuration as list of Configuration Objects (as in ver.
- System Configuration Conversions
- The conversion utilities have been provided to restore its original representation of configurations with
runs >= 1
-
Notes:
configs
is temporarily returned in a flattened format and reformatted into its intended formatConfiguration
objects at0x10790e470
and0x1143dd630
are reconstituted into objects at0x10790e7b8
and0x116268908
respectively.
-
ver. ≥
0.4.22
:from pprint import pprint from documentation.examples import sys_model_A, sys_model_B from cadCAD.configuration.utils import configs_as_objs, configs_as_dataframe, configs_as_dicts from cadCAD import configs flattened_configs = configs print('Flattened Format: Temporary') pprint(flattened_configs) print() print('Intended Format:') intended_configs = configs_as_objs(flattened_configs) pprint(intended_configs) print()
Result:
Flattened Format: Temporary [<cadCAD.configuration.Configuration object at 0x10790e470>, <cadCAD.configuration.Configuration object at 0x10790e7b8>, <cadCAD.configuration.Configuration object at 0x1143dd630>, <cadCAD.configuration.Configuration object at 0x116268908>] Intended Format: [<cadCAD.configuration.Configuration object at 0x10790e7b8>, <cadCAD.configuration.Configuration object at 0x116268908>]
- Enables the development of feature enhancements that involve the use of additional parameters without requiring users
to modify their update parameters spaces when upgrading to newer versions. For this reason state / policy update
examples in documentation include an additional
**kwargs
parameter.
-
ver. ≥
0.4.22
:def state_update(_params, substep, sH, s, _input, **kwargs): ... return 'state_variable_name', new_value def policy(_params, substep, sH, s, **kwargs): ... return {'signal_1': value_1, ..., 'signal_N': value_N}
-
ver.
0.3.1
:def state_update(_params, substep, sH, s, _input): ... return 'state_variable_name', new_value def policy(_params, substep, sH, s): ... return {'signal_1': value_1, ..., 'signal_N': value_N}
- Packaging: Add Nix derivation and shell for local development and distribution of cadCAD package using Nix. Nix is a powerful package manager for Linux and other Unix systems that makes package management reliable and reproducible, allowing you to share your development and build environments across different machines.