Skip to content

Latest commit

 

History

History
308 lines (264 loc) · 15.3 KB

CHANGELOG.md

File metadata and controls

308 lines (264 loc) · 15.3 KB
Raw

Changelog:

0.4.29.1

Changes

  • Parallel executor uses the context manager handling the Process Pool lifetime

0.4.29

  • Merged repo with the cadCAD_tweaked, which includes performance improvements
  • Python 3.10 compatible

September 28, 2021

New Features:

  • ver. ≥ 0.4.28:
    • Experiments
      • System Model Configuration
        • Configurations (cadCAD.utils.Configuration's) are now accessed via the configs member of cadCAD.configuration.Experiment. Example: cadCAD.configuration.Experiment().configs
        • cadCAD.configs has been re-included for backwards compatibility and has been assigned cadCAD.experiment.configs
      • Experiments
        • cadCAD.configuration.Experiment() is unique representation of an experiment of one or more configured System Models.
          • The cadCAD module now contains a default Experiment object cadCAD.experiment as an instantiation of cadCAD.configuration.Experiment()
        • An Experiment's append_model method stores multiple system model Configuration's for simulation execution within cadCAD.configuration.Experiment().configs. cadCAD.configuration.Experiment().model_ids contains system model labels and/or indexes for cadCAD.configuration.Experiment().configs
          • The Experiment's append_model method is defined with model_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 within cadCAD.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', ...])
    • Upgrade Guide: specific to feature changes / additions

August 25, 2021

New Features:

  • ver. ≥ 0.4.27:
    • Experiments
      • System Model Configurations
        • Configurations (cadCAD.utils.Configuration's) as are no longer a part of the cadCAD module (as cadCAD.configs) and are now accessed via the configs member of cadCAD.configuration.Experiment. Example: cadCAD.configuration.Experiment().configs
      • Experiments
        • cadCAD.configuration.Experiment is unique representation of an experiment of one or more configured System Models. An Experiment's append_model method stores multiple system model Configuration's for simulation execution.
          • The Experiment's append_model method requires a model_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 of cadCAD.configuration.Experiment
          • Backwards Compatibility: The append_model method of cadCAD.configuration.Experiment can also be used as the append_configs method.
    • 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.
    • #257
      • ValueError for runs accepted by the cadCAD.configuration.Experiment().append_model via the sim_configs no longer gives mis-leading error message if catching a non-related ValueError
    • #252
      • Jupyter lab and Jupyter notebook recognises cadCAD module

September 22, 2020

Multi - System Model Execution
  • 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 to cadCAD.configs).
      • Issue: #195
    • Parameter Sweep value M (Params) requires up to a maximum of 2 distinct lengths

August 5, 2020

Experiments
  • 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 of cadCAD.configuration.Experiment. As of now it does not support multi - system model simulation because configurations are still appended globally despite append_config being a method of Experiment.
Examples:
  • ver. ≥ 0.4.22: 
    from cadCAD.configuration import Experiment
exp = Experiment()
exp.append_configs(...)
  • ver. 0.3.1: Deprecated 
    from cadCAD.configuration import append_configs
append_configs(...)

June 22, 2020

  • Bug Fix: Multiprocessing error for Windows

June 19, 2020

New Features:

Local Execution Mode (Default)
  • 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
Examples:

  • 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)
cadCAD Post-Processing Enhancements / Modifications
  • 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
    • Note: Returning a single dataset was originally specified during the project’s inception instead of multiple per simulation
Examples:

  • 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 run cadCAD.configuration.Configuration objects to both fault-tolerant simulation and elastic workloads. This functionality will be restored in a subsequent release by a class that returns cadCAD.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
Examples:

  • Notes:

    • configs is temporarily returned in a flattened format and reformatted into its intended format
    • Configuration objects at 0x10790e470 and 0x1143dd630 are reconstituted into objects at 0x10790e7b8 and 0x116268908 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>]
Expandable state and policy update parameter space:
  • 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.
Examples:

  • 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}

May 29, 2020

  • 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.