Graphnetcore dataload compat

.github/workflows/Documenter.yml

@@ -36,7 +36,7 @@ jobs:
           cp ./README.md ./docs/src/index.md
 
       - name: "Install dependencies"
-        run: julia --project=docs/ -e 'using Pkg; Pkg.add(url="https://github.com/una-auxme/GraphNetCore.jl", rev="GNS-small-graph"); Pkg.develop(PackageSpec(path=pwd())); Pkg.instantiate()'
+        run: julia --project=docs/ -e 'using Pkg; Pkg.develop(PackageSpec(path=pwd())); Pkg.instantiate()'
       - name: "Build and deploy"
         env:
           GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.gitignore

@@ -16,6 +16,8 @@ deps/src/
 # Build artifacts for creating documentation generated by the Documenter package
 docs/build/
 docs/site/
+# Generated from README.md at build time (see docs/make.jl / Documenter.yml)
+docs/src/index.md
 
 # File generated by Pkg, the package manager, based on a corresponding Project.toml
 # It records a fixed state of all packages used by the project. As such, it should not be

LICENSE

@@ -1,6 +1,6 @@
 MIT License
 
-Copyright (c) 2026 Josef Kircher <Josef.Kircher@uni-a.de> and contributors
+Copyright (c) 2026 Josef Jouaux <Josef.Kircher@uni-a.de> and contributors
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

Project.toml

@@ -1,7 +1,7 @@
 name = "GraphNetSim"
 uuid = "5ff66f56-808c-48e7-ac84-dd29877231f8"
 version = "0.1.0"
-authors = ["JK <Josef.Kircher@uni-a.de", "JT <julian.trommer@informatik.uni-augsburg.de>"]
+authors = ["Josef Jouaux <Josef.Kircher@uni-a.de>", "JT <julian.trommer@uni-a.de>"]
 
 [deps]
 Adapt = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"
@@ -12,12 +12,14 @@ ComponentArrays = "b0b7db55-cfe3-40fc-9ded-d10e2dbeff66"
 DataFrames = "a93c6f00-e57d-5684-b7b6-d8193f3e46c0"
 DataInterpolations = "82cc6244-b520-54b8-b5a6-8a565e85f1d0"
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
+GPUCompiler = "61eb1bfa-7361-4325-ad38-22787b887f55"
 GraphNetCore = "7809f980-de1b-4f9a-8451-85f041491431"
 HDF5 = "f67ccb44-e63f-5c2f-98bd-6dc0ccc4ba2f"
 JLD2 = "033835bb-8acc-5ee8-8aae-3f567f8a3819"
 JSON = "682c06a0-de6a-54ab-a142-c8b1cf79cde6"
 JuliaFormatter = "98e50ef6-434e-11e9-1051-2b60c6c9e899"
 KernelAbstractions = "63c18a36-062a-441e-b654-da1e3ab1ce7c"
+LightXML = "9c8b4983-aa76-5018-a973-4c85ecc9e179"
 Lux = "b2108857-7c20-44ae-9111-449ecde12c47"
 LuxCUDA = "d0bbae9a-e099-4d5b-a835-1c6931763bda"
 MLUtils = "f1d291b0-491e-4a28-83b9-f70985020b54"
@@ -29,33 +31,34 @@ PreallocationTools = "d236fae5-4411-538c-8e31-a6e3d9e00b46"
 Printf = "de0858da-6303-5e67-8744-51eddeeeb8d7"
 ProgressMeter = "92933f4c-e287-5a05-a399-4b506db050ca"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
+ReadVTK = "dc215faf-f008-4882-a9f7-a79a826fadc3"
 RecursiveArrayTools = "731186ca-8d62-57ce-b412-fbd966d074cd"
 SciMLBase = "0bca4576-84f4-4d90-8ffe-ffa030f20462"
 SciMLSensitivity = "1ed8b502-d754-442c-8d5d-10ac956f44a1"
+Setfield = "efcf1570-3423-57d1-acb7-fd33fddbac46"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
 cuDNN = "02a925ec-e4fe-4b08-9a7e-0d78e3d38ccd"
 
-[sources]
-GraphNetCore = {rev = "GNS-small-graph", url = "https://github.com/una-auxme/GraphNetCore.jl"}
-
 [compat]
 Adapt = "4.3.0"
 Aqua = "0.8"
 CSV = "0.10.15"
-CUDA = "5.5.2"
+CUDA = "5.9.6"
+GPUCompiler = "~1.9"
 ChainRulesCore = "1.16.0 - 1"
 ComponentArrays = "0.15"
 DataFrames = "1.6"
 DataInterpolations = "8.6.0"
 Distributions = "0.25"
-GraphNetCore = "0.3.1"
+GraphNetCore = "0.4.2"
 HDF5 = "0.17"
-JLD2 = "0.4 - 0"
+JLD2 = "0.6"
 JSON = "0.21"
-JuliaFormatter = "2.2.1"
+JuliaFormatter = "~2.2"
 KernelAbstractions = "0.9.39"
-Lux = "1.4 - 1"
+LightXML = "0.9"
+Lux = "1.13 - 1"
 LuxCUDA = "0.3"
 MLUtils = "0.4.4 - 0.4"
 Optimisers = "0.4, 1"
@@ -66,19 +69,20 @@ PreallocationTools = "0.4.34"
 Printf = "1"
 ProgressMeter = "1.7.0 - 1"
 Random = "1"
+ReadVTK = "0.2.6"
 RecursiveArrayTools = "3.36.0"
 SciMLBase = "2.7.0 - 2"
 SciMLSensitivity = "7.76 - 7"
+Setfield = "1.1.2"
 Statistics = "1"
 Test = "1"
-Zygote = "0.6"
+Zygote = "0.6, 0.7"
 cuDNN = "1.4 - 1"
-julia = "1.10"
+julia = "1.11"
 
 [extras]
 Aqua = "4c88cf16-eb10-579e-8560-4a9242c79595"
-Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 
 [targets]
-test = ["Aqua", "Statistics", "Test"]
+test = ["Aqua", "Test"]

README.md

@@ -1,5 +1,9 @@
 # GraphNetSim.jl
 
+<p align="center">
+  <img src="https://raw.githubusercontent.com/una-auxme/GraphNetSim.jl/main/docs/src/assets/logo.svg" alt="GraphNetSim.jl logo" width="200"/>
+</p>
+
 [![Docs](https://img.shields.io/badge/docs-dev-blue.svg)](https://una-auxme.github.io/GraphNetSim.jl/dev)
 [![ColPrac: Contributor's Guide on Collaborative Practices for Community Packages](https://img.shields.io/badge/ColPrac-Contributor's%20Guide-blueviolet)](https://github.com/SciML/ColPrac)
 [![Code Style: Blue](https://img.shields.io/badge/code%20style-blue-4495d1.svg)](https://github.com/JuliaDiff/BlueStyle)
@@ -17,6 +21,13 @@
 
 The package is build upon [**GraphNetCore.jl**](https://github.com/una-auxme/GraphNetCore.jl) for the underlying graph neural network architecture.
 
+## Requirements
+
+- **Julia ≥ 1.11**
+- Built on [GraphNetCore.jl](https://github.com/una-auxme/GraphNetCore.jl) **v0.4**, which uses a
+  [Lux.jl](https://github.com/LuxDL/Lux.jl) `TrainState` and pulls in a CUDA-capable stack.
+- A CUDA-capable GPU is recommended for training (falls back to CPU when CUDA is unavailable).
+
 ## Installation
 
 To add GraphNetSim.jl to your Julia environment, use:
@@ -62,16 +73,18 @@ min_loss = train_network(
     training_strategy=DerivativeTraining()  # Training strategy
 )
 
-# Evaluate the trained network with long-term rollouts
+# Evaluate the trained network with long-term rollouts.
+# Note: the ODE solver is the 4th *positional* argument (not a keyword).
 eval_network(
     ds_path,
     cp_path,
-    "./results";               # Output directory
-    solver=Tsit5(),            # ODE solver
+    "./results",               # Output directory
+    Tsit5();                   # ODE solver (positional)
     start=0.0f0,
     stop=1.0f0,
-    saves=0.0:0.01:1.0,        # Time points to save
-    mse_steps=0.0:0.1:1.0      # Time points for error metrics
+    dt=0.01f0,                 # fixed timestep (omit for adaptive stepping)
+    saves=0.0f0:0.01f0:1.0f0,  # Time points to save
+    mse_steps=collect(0.0f0:0.1f0:1.0f0),  # Time points for error metrics
 )
 ```
 
@@ -93,7 +106,7 @@ The metadata file defines feature dimensions, node types, graph connectivity, an
 
 ## Key Features
 
-- **Multiple Training Strategies**: Choose between `BatchingTraining` and `DerivativeTraining` to suit your problem
+- **Multiple Training Strategies**: `DerivativeTraining` (fast, no ODE solve), plus the ODE-based `BatchingStrategy`, `SingleShooting`, and `MultipleShooting` for fine-tuning
 - **GPU-accelerated Training**: Automatic CUDA detection and memory management
 - **Flexible Architecture**: Configurable message passing steps, layer sizes, and hidden layers
 - **Progress Monitoring**: Built-in progress bars and logging for training and validation
@@ -136,30 +149,41 @@ See the [full API documentation](https://una-auxme.github.io/GraphNetSim.jl/dev/
 
 ## Visualization
 
-Export predicted trajectories as VTK files for visualization:
+Export the evaluation results as VTK HDF files for visualization in ParaView. The
+`eval_network` output is written under `<out_path>/<solver>/trajectories.h5`:
 
 ```julia
-visualize(
-    "trajectories.h5",           # Results file from eval_network
-    "./vtk_output",              # Output directory
-    "pos",                        # Position dataset name
-    "prediction";                 # Subgroup to visualize
-    Trajectorys=1:5              # Trajectory indices
+visualize_eval(
+    "./results/tsit5/trajectories.h5",  # trajectories.h5 written by eval_network
+    "./results/vtkhdf/",                # output directory for the VTK HDF files
 )
 ```
 
+The lower-level `visualize` function is also available for exporting a single group of a
+custom HDF5 file (see the [API reference](https://una-auxme.github.io/GraphNetSim.jl/dev/api/)).
+
 ## Related Packages
 
 - [**PointNeighbors.jl**](https://github.com/una-auxme/PointNeighbors.jl): Efficient spatial indexing for neighbor queries
+- [**Octopus.jl**](https://github.com/una-auxme/Octopus.jl): Fast octree neighborhood search on CPU and NVIDIA GPUs
 - [**GraphNetCore.jl**](https://github.com/una-auxme/GraphNetCore.jl): Core GNN architecture and normalization strategies
-- [**DifferentialEquations.jl**](https://github.com/SciML/DifferentialEquations.jl): ODE solvers for trajectory integration
 
 ## References
 
 This package is inspired by the Graph Network-based Simulator (GNS) framework:
 
 - Sanchez-Gonzalez, A., Godwin, J., Pfaff, T., et al. (2020). "Learning to Simulate Complex Physics with Graph Networks." *Proceedings of the 37th International Conference on Machine Learning (ICML)*.
 
+## License and Attribution
+
+GraphNetSim.jl is distributed under the [MIT License](https://github.com/una-auxme/GraphNetSim.jl/blob/main/LICENSE).
+
+It includes / is derived from portions of DeepMind's
+[`learning_to_simulate`](https://github.com/google-deepmind/deepmind-research/tree/master/learning_to_simulate)
+(Copyright 2020 DeepMind Technologies Limited), which is licensed under the Apache License,
+Version 2.0. The attribution and the full Apache 2.0 license text are provided in
+[`THIRD_PARTY_NOTICES.md`](https://github.com/una-auxme/GraphNetSim.jl/blob/main/THIRD_PARTY_NOTICES.md).
+
 ## Contributing
 
 We welcome contributions to GraphNetSim.jl! Please follow the [ColPrac](https://github.com/SciML/ColPrac) guidelines for collaborative practices.