Skip to content

Chung neg visc test#1

Closed
baperry2 wants to merge 4 commits intodevelopmentfrom
chung-neg-visc-test
Closed

Chung neg visc test#1
baperry2 wants to merge 4 commits intodevelopmentfrom
chung-neg-visc-test

Conversation

@baperry2
Copy link
Copy Markdown
Owner

@baperry2 baperry2 commented Sep 7, 2023

test

@baperry2 baperry2 force-pushed the chung-neg-visc-test branch from 9db1865 to 10eed13 Compare September 7, 2023 21:38
@baperry2 baperry2 force-pushed the chung-neg-visc-test branch from 10eed13 to 050b80b Compare September 7, 2023 21:54
@baperry2 baperry2 closed this Sep 22, 2023
baperry2 pushed a commit that referenced this pull request May 3, 2024
Combined Source Term lookups
baperry2 added a commit that referenced this pull request Jul 8, 2024
…sport (AMReX-Combustion#492)

* manifold/table EOS support

* Convert tabs to spaces

* Initial framework for neural net based manifold EoS

Still need to generalize TabFuncParams and TabFuncData and work the neural net into the existing EoS framework

* Tentative start to NNFuncParams and NNFuncData

* Use generalize ManifoldFunc class instead of using TabFunc directly

* Migrate get_var_index; have TabFuncData inherit from ManFuncData

- Moved get_var_index to ManifoldFunc.H instead of EosParams.H
- Shortened get_var_index arglist

* Switch fully over to using generic ManFuncData struct when possible

* Mostly complete setup (I hope) for neural net manifold

- Still need to calculate source terms
- Still need to account for variables passed to input layer
- Maybe we can worry about GPUs later

* Set up neural net manifold to evaluate source terms

- This style of manifold will print all of the metadata properly, but SEGFAULTs at some point in EosEval/main.cpp

* Fixed SEGFAULT

* Non-crashing version of neural net manifold

* Switch manifold model selection to runtime parameter

* Add variable indices to NeuralNet manifold parameter defs

* Let manifold models read in mass fractions prefixed with Y-

* A few fixes to let IgnitionDelay run with neural net

* Make including training variables (mass fractions) in output variables optional

* Allow for reading of manifold biases from info file

* Fix an issue with calculating derivatives; add semicolon

* Switch to different initialization for iss

* Merge pull request #1 from baperry2/manifold

Combined Source Term lookups

* Use correct accessor dimension in loop bound

* Changes in Eos/ and Transport/ to permit home-rolled evaluation

* Add a few checks to ensure the input is valid

* Merge branch 'kiraneiden/manifold' of github.com:KiranEiden/PelePhysics into kiraneiden/manifold

* Avoid editing nnf_data, make indata_copy local again

* Implement 'get_all_values' function

* add inverse density option

* merge in updated pelephysics

* move new EOS files to Source/Eos

* other updates for EOS bringing in manifold

* constraints for physics combinations with manifold

* stuff for building with manifold transport

* bring in manifold transport

* changes to compile manifold stuff

* table lookup improvement

* EosEval test case

* clang-format new EOS stuff

* more formatting

* a few more missed files for formatting

* fix mangles make.package

* locate is a host device function

* address codespell issues

* move blackboxfunc to its own utility rather than part of eos

* gmake for moved blackboxfunc

* don't initialize idx_Wdot to avoid issues

* add some runtime gamma support, but warn because its risky

* dont compute wdot in test for gammalaw

* fix some bugs on cuda

* remove some hacks for manifold eos

* try running manifold in CI

* add_mani_src stuff change default in manifold EOS

* updates for Manifold in TranEval

* remove ProductionRateEval - EosEval tests the same thing

* partial removal of wdot stuff from black box funcs

* some sample network stuff

* new generic parameter stuff

* working template + Factory stuff

* updates to PeleParamsGeneric for templating

* move ManFuncData out of ManFuncParams

* improved GPU data management with templating

* remove old manfuncdata stuff

* remove a bunch of stupid test code

* start moving transport to new GPU data template

* trans stuff initialize using eos stuff

* clang-format and remove an old file

* remove unneeded file

* more clang-format

* fix spelling

* remove network evaluator using pytorch library for now

* clean up test cases; finish removing pytorch lib dep

* remove deprecated ManFuncParams class

* Add override tag to remove warning

* fix transport test in CI for manifold to use right chemistry

* add netork inputs file for transport

* no transparm in reacteval

* no transparm in ignition delay

* table dimensionality templates

* modify error checking in table

* compile time choice of network vs table for manifold

* add factory function for blackboxfunc

* move interpolation functions outside of table class for templating

* don't compile with Manifold for SYCL

final compilation change for no manifold with sycl

* rename ManifoldFunc to BlackBoxFunc

* 1D table if not manifold EOS

* revert some changes in ci workflow

* remove duplicate NUM_ELEMENTS in Null mech

* fix a bunch of compiler warnings

* fix a bunch more warnings

* streamline data storage for manifold models

* don't need changes in Make.package

* fix all clang-tidy stuff

* change species index from is to i in Eos loops

* style changes in EosParams

* use more verbose names in black box function stuff

---------

Co-authored-by: Kiran Eiden <[email protected]>
Co-authored-by: KiranEiden <[email protected]>
baperry2 pushed a commit that referenced this pull request Dec 19, 2024
ceptr: Added functionality for reading and converting PLOG reactions.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant