WIP: New JSON Interface #231

iamveronika · 2024-08-26T07:02:48Z

Command line interface
julia scripts/runfit.jl -p fitting_params.json

Results:

bash-4.4$ julia scripts/runfit.jl -p fitting_params.json
  Activating project at `/z1-vekondra/vekondra/ACEP`
[ Info: making ACEmodel
       |      E    |    F  
 train | 2.69e-03  |  2.91e-01  
  test | 8.07e-03  |  3.13e-01

New results:

bash-4.4$ julia runfit.jl -p fitting_params.json
  Activating project at `/z1-vekondra/vekondra/tmp/ACEpotentials.jl`
[ Info: making ACEmodel
┌───────────────┬──────────┬───────┬────┬─────┬─────┐
│          Type │ #Configs │ #Envs │ #E │  #F │  #V │
├───────────────┼──────────┼───────┼────┼─────┼─────┤
│ isolated_atom │        1 │     1 │  1 │   3 │   0 │
│           dia │       25 │    50 │ 25 │ 150 │ 150 │
│            bt │       25 │    50 │ 25 │ 150 │ 150 │
│           liq │        2 │   128 │  2 │ 384 │  12 │
├───────────────┼──────────┼───────┼────┼─────┼─────┤
│         total │       53 │   229 │ 53 │ 687 │ 312 │
│       missing │        0 │     0 │  0 │   0 │   6 │
└───────────────┴──────────┴───────┴────┴─────┴─────┘
[ Info: Assembling linear problem.
[ Info:   - Creating feature matrix with size (1052, 120).
[ Info:   - Beginning assembly with processor count:  1.
Progress: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| Time: 0:00:41
[ Info:   - Assembly completed.
[ Info: Assembling full weight vector.
Progress: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| Time: 0:00:00
┌ Warning: Need to apply preconditioner in LSQR.
└ @ ACEfit /z1-vekondra/vekondra/tmp/ACEfit.jl/src/solvers.jl:95
damp  0.005
atol  1.0e-6
maxiter  100000
Converged after 20 iterations.
relative RMS error  0.09083234883494287
[ Info: RMSE Table
┌───────────────┬────────────┬──────────┬───────────┐
│          Type │    E [meV] │ F [eV/A] │   V [meV] │
├───────────────┼────────────┼──────────┼───────────┤
│ isolated_atom │ 158544.968 │    0.000 │     0.000 │
│           dia │    153.972 │    0.346 │  5052.524 │
│           liq │     66.949 │    4.232 │ 11817.642 │
│            bt │    406.986 │    2.670 │  4291.416 │
├───────────────┼────────────┼──────────┼───────────┤
│           set │  21779.881 │    3.405 │  5147.665 │
└───────────────┴────────────┴──────────┴───────────┘
[ Info: MAE Table
┌───────────────┬────────────┬──────────┬──────────┐
│          Type │    E [meV] │ F [eV/A] │  V [meV] │
├───────────────┼────────────┼──────────┼──────────┤
│ isolated_atom │ 158544.968 │    0.000 │    0.000 │
│           dia │    103.150 │    0.278 │ 3548.744 │
│           liq │     65.941 │    3.447 │ 8521.847 │
│            bt │    306.851 │    1.889 │ 2838.204 │
├───────────────┼────────────┼──────────┼──────────┤
│           set │   3187.300 │    2.400 │ 3398.412 │
└───────────────┴────────────┴──────────┴──────────┘
[ Info: RMSE Table
┌───────────────┬────────────┬──────────┬───────────┐
│          Type │    E [meV] │ F [eV/A] │   V [meV] │
├───────────────┼────────────┼──────────┼───────────┤
│ isolated_atom │ 158544.968 │    0.000 │     0.000 │
│           dia │    153.972 │    0.346 │  5052.524 │
│           liq │     66.949 │    4.232 │ 11817.642 │
│            bt │    406.986 │    2.670 │  4291.416 │
├───────────────┼────────────┼──────────┼───────────┤
│           set │  21779.881 │    3.405 │  5147.665 │
└───────────────┴────────────┴──────────┴───────────┘
[ Info: MAE Table
┌───────────────┬────────────┬──────────┬──────────┐
│          Type │    E [meV] │ F [eV/A] │  V [meV] │
├───────────────┼────────────┼──────────┼──────────┤
│ isolated_atom │ 158544.968 │    0.000 │    0.000 │
│           dia │    103.150 │    0.278 │ 3548.744 │
│           liq │     65.941 │    3.447 │ 8521.847 │
│            bt │    306.851 │    1.889 │ 2838.204 │
├───────────────┼────────────┼──────────┼──────────┤
│           set │   3187.300 │    2.400 │ 3398.412 │
└───────────────┴────────────┴──────────┴──────────┘

cortner · 2024-08-26T14:52:16Z

scripts/json_interface.jl

@@ -0,0 +1,54 @@
+using ACEpotentials


should this be a sub-module of ACEpotentials?

... answering myself. Yes, I think that would be better.

cortner · 2024-08-26T14:53:25Z

scripts/runfit.jl

+    pyimport("pprint")["pprint"](controller.select(user_api="blas").info())
+end
+
+include("../src/ace1_compat.jl")


This include should definitely not be here. Use ACEpotentials.ACE1compat to access the submodule.

cortner · 2024-08-26T14:55:02Z

Project.toml

@@ -8,6 +8,7 @@ ACE1x = "5cc4c08c-8782-4a30-af6d-550b302e9707"
 ACEbase = "14bae519-eb20-449c-a949-9c58ed33163e"


There is a lengthy comment about maintaining a separate project inside scripts in the other PR. Can you copy that over please?

I also don't like this activation here. In general a user should create a working directory for their project, where ACEpotentials is added (to the local Project.toml). This will then be the activated project.

cortner · 2024-08-26T14:55:31Z

thank you - looks much better but still some re-organization to be discussed. Before our meeting can you please also take a look at the documentation.

iamveronika · 2024-08-27T18:39:53Z

Just a few quesions:

Do we want a dry run?
We don't have multithreaded assembly yet. Do we want to include this in our first example?
Weight for lsq is not yet "by config_type" @cortner

cortner · 2024-08-27T18:54:43Z

Do we want a dry run?

yes. The point is that the user can check how big the problem is.

We don't have multithreaded assembly yet. Do we want to include this in our first example?

I guess not. Just get anything to run, we can add the distributed and multi-threaded aspects later.

Weight for lsq is not yet "by config_type"

I don't understand this. Can you say more please?

iamveronika · 2024-08-27T19:19:51Z

I mean that here (in scripts/runfit,jl) weight should be also depending on config_type
wE = 30.0; wF = 1.0; wV = 1.0
weights = (wE = wE/u"eV", wF = wF / u"eV/Å", )

cortner · 2024-08-28T00:05:16Z

Yes, weights should be provided in the "old" style. My little hacked together assembly was just for a quick demo. The docs and tests should show how weights are correctly specified. E.g. look at test_silicon

cortner · 2024-08-28T00:05:46Z

Btw - the assembly with the new kernels is essentially ready. we can discuss that tomorrow.

…/updt # Conflicts: # Project.toml

cortner · 2024-08-31T04:54:18Z

@iamveronika okease update the PR again by merging main into your branch

# Conflicts: # Project.toml # src/atoms_data.jl # src/models/calculators.jl

# Conflicts: # Project.toml

cortner · 2024-09-10T00:40:25Z

I think this is ready to merge. That will enable us to make further progress through separate smaller PRs.

Veronika Kondratenko and others added 5 commits August 16, 2024 14:46

updated runfits and acecompat

74715d2

removed unnecessary 'show'

d9b7caa

Removed accidentally committed fitting_params.json file

cd023bc

Merge branch 'ACEsuit:main' into main

0ea104a

Corrected dir and acecompat

88fcb14

cortner reviewed Aug 26, 2024

View reviewed changes

iamveronika mentioned this pull request Aug 26, 2024

Updated runfits and acecompat #223

Closed

iamveronika changed the title ~~Updated and corrected errors~~ WIP: New JSON Interface Aug 26, 2024

Veronika Kondratenko added 2 commits August 27, 2024 10:47

Cleaned runfits

82c1745

Relocated json_inteface

048bb12

Merge branch 'co/fits' of github.com:ACEsuit/ACEpotentials.jl into ve…

2b31791

…/updt # Conflicts: # Project.toml

Veronika Kondratenko and others added 12 commits September 1, 2024 17:36

Updated runfits and saved the results to disk

6861cd6

Deleted json file and folder with results

408b477

Merged main into ve/updt

a0da813

For merge

5b112a7

Merge commit 'b427bcbb97e3092451e04b6e65b2c808a172d172' into ve/updt

ddcdeb4

# Conflicts: # Project.toml # src/atoms_data.jl # src/models/calculators.jl

Added saving results

44a25d1

Updated runfits

3037bbd

Updated runfits and json file

b6ca83c

Updated runfits

4c699aa

Merge commit 'e0ea1e110b82cd64809e81d0db02dd23abd55f76' into ve/updt

afca49a

# Conflicts: # Project.toml

cleanup

4dc91e2

cleanup; runfit test

f7c3cd1

cortner merged commit 0807f85 into ACEsuit:main Sep 10, 2024
2 checks passed

cortner deleted the ve/updt branch September 10, 2024 00:40

cortner mentioned this pull request Sep 10, 2024

Revive and redesign JSON interface #218

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

WIP: New JSON Interface #231

WIP: New JSON Interface #231

iamveronika commented Aug 26, 2024 •

edited

Loading

cortner Aug 26, 2024

cortner Aug 26, 2024

cortner Aug 26, 2024

cortner Aug 26, 2024

CheukHinHoJerry Aug 26, 2024

cortner commented Aug 26, 2024 •

edited

Loading

iamveronika commented Aug 27, 2024

cortner commented Aug 27, 2024

iamveronika commented Aug 27, 2024

cortner commented Aug 28, 2024

cortner commented Aug 28, 2024

cortner commented Aug 31, 2024

cortner commented Sep 10, 2024

		@@ -8,6 +8,7 @@ ACE1x = "5cc4c08c-8782-4a30-af6d-550b302e9707"
		ACEbase = "14bae519-eb20-449c-a949-9c58ed33163e"

WIP: New JSON Interface #231

WIP: New JSON Interface #231

Conversation

iamveronika commented Aug 26, 2024 • edited Loading

cortner Aug 26, 2024

Choose a reason for hiding this comment

cortner Aug 26, 2024

Choose a reason for hiding this comment

cortner Aug 26, 2024

Choose a reason for hiding this comment

cortner Aug 26, 2024

Choose a reason for hiding this comment

CheukHinHoJerry Aug 26, 2024

Choose a reason for hiding this comment

cortner commented Aug 26, 2024 • edited Loading

iamveronika commented Aug 27, 2024

cortner commented Aug 27, 2024

iamveronika commented Aug 27, 2024

cortner commented Aug 28, 2024

cortner commented Aug 28, 2024

cortner commented Aug 31, 2024

cortner commented Sep 10, 2024

iamveronika commented Aug 26, 2024 •

edited

Loading

cortner commented Aug 26, 2024 •

edited

Loading