Skip to content

Commit ad756de

Browse files
ashwinmaclaude
ashwinma
and
claude
authored
Add ACKNOWLEDGEMENTS.md and wire citation requirement into add-model workflow (#34)
* add ACKNOWLEDGEMENTS.md and wire citation requirement into add-model workflow - Create ACKNOWLEDGEMENTS.md with per-model attribution: upstream paper, ROCm blog authors (AMD Silo AI), and collaboration callouts (AstraZeneca × AMD) - Update _template/README.md to require an Acknowledgements/Citation section in every new model entry - Update add-model command to require citation info in README and an ACKNOWLEDGEMENTS.md entry as mandatory steps Co-Authored-By: Claude Sonnet 4 <noreply@anthropic.com> * ack: fix AMD Silo AI link to use amd.com page Co-Authored-By: Claude Sonnet 4 <noreply@anthropic.com> --------- Co-authored-by: Claude Sonnet 4 <noreply@anthropic.com>
1 parent c0dafba commit ad756de

3 files changed

Lines changed: 136 additions & 2 deletions

File tree

‎.claude/commands/add-model.md‎

Lines changed: 10 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -18,6 +18,12 @@ Add a new model to this repository following all repo conventions.
1818
- Task description
1919
- License (SPDX id or direct link)
2020
- Upstream code repo and paper links
21+
- **Acknowledgements and citation section** (required — see template):
22+
- Upstream repo URL
23+
- Paper citation (author, title, venue, DOI or arXiv id)
24+
- BibTeX or DOI-based "cite as" block copied from the upstream repo or model card
25+
- ROCm blog URL and author names (AMD Silo AI) if a blog post exists
26+
- Any named collaboration (e.g. AstraZeneca × AMD, ORNL × AMD)
2127

2228
5. **Add recipes** under `<slug>/recipes/`. Use one subfolder per task (`recipes/inference/`, `recipes/finetune/`, etc.). Each recipe subfolder needs a `README.md`.
2329

@@ -42,9 +48,11 @@ Add a new model to this repository following all repo conventions.
4248
- `healthcare/` (Healthcare & Life Sciences): add research/engineering-only disclaimer; no PHI; copy intended-use and limitations from the model card.
4349
- `physics_simulation/`: state physical domain (fluid dynamics, plasma, etc.), dataset format (HDF5, NetCDF), and HPC/multi-node requirements.
4450

45-
9. **Do not commit** large checkpoints, datasets, `.env` files, or tokens — document how users obtain them instead.
51+
9. **Update `ACKNOWLEDGEMENTS.md`** at the repo root — add a per-model entry under the appropriate domain section following the existing format. Include paper citation, upstream repo, ROCm blog + author (if applicable), and any collaboration callout.
4652

47-
10. **Git workflow** — always branch, never commit directly to `main`:
53+
10. **Do not commit** large checkpoints, datasets, `.env` files, or tokens — document how users obtain them instead.
54+
55+
11. **Git workflow** — always branch, never commit directly to `main`:
4856
1. `git fetch origin && git checkout -b <your-username>/<model-slug> origin/main`
4957
2. Create all files, set scripts `chmod +x`.
5058
3. `git add <domain>/models/<slug>/` and commit.

‎ACKNOWLEDGEMENTS.md‎

Lines changed: 112 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,112 @@
1+
# Acknowledgements
2+
3+
AI4Science Studio curates recipes for open AI-for-science models. The models themselves, their weights, and the ideas behind them are the work of the original authors listed below. AMD Silo AI contributed ROCm/AMD-specific recipes and blog posts for many of these models.
4+
5+
---
6+
7+
## Earth Science
8+
9+
### StormCast
10+
- **Upstream:** NVIDIA Earth-2 Studio — [`NVIDIA/earth2studio`](https://github.com/NVIDIA/earth2studio)
11+
- **Paper:** Bodnar et al., *Kilometer-Scale Convection Allowing Model Emulation using Generative Diffusion Modeling*, arXiv:2408.10958
12+
- **ROCm blog (inference):** [Running StormCast on AMD Instinct](https://rocm.blogs.amd.com/artificial-intelligence/stormcast-inference/README.html) — Pauli Pihajoki (AMD Silo AI)
13+
- **ROCm blog (ensembles):** [StormCast Ensemble Forecasting on AMD Instinct](https://rocm.blogs.amd.com/artificial-intelligence/stormcast-ensembles/README.html) — Pauli Pihajoki (AMD Silo AI)
14+
15+
### ORBIT-2
16+
- **Upstream:** [`XiaoWang-Github/ORBIT-2`](https://github.com/XiaoWang-Github/ORBIT-2); weights on [Hugging Face](https://huggingface.co/jychoi-hpc/ORBIT-2)
17+
- **Paper:** Wang et al., *ORBIT-2: Scaling Exascale Vision Foundation Models for Weather and Climate Downscaling*, arXiv:2505.04802
18+
- **Dataset DOI:** [10.13139/OLCF/2589526](https://doi.org/10.13139/OLCF/2589526)
19+
- **Maintained by:** Oak Ridge National Laboratory (ORNL)
20+
21+
### ArchesWeather
22+
- **Upstream:** [`gcouairon/ArchesWeather`](https://huggingface.co/gcouairon/ArchesWeather); recipe code from [`silogen/ai-samples`](https://github.com/silogen/ai-samples)
23+
- **Paper:** Couairon et al., *ArchesWeather & ArchesWeatherGen: efficient AI weather forecasting*, arXiv:2412.12971
24+
- **ROCm blog:** [Training ArchesWeather on AMD Instinct](https://rocm.blogs.amd.com/artificial-intelligence/geoarches-training/README.html) — Luka Tsabadze, Rahul Biswas, Pauli Pihajoki, Daniel Warna, Baiqiang Xia, Sopiko Kurdadze (AMD Silo AI)
25+
26+
### Aurora
27+
- **Upstream:** [`microsoft/aurora`](https://github.com/microsoft/aurora); recipe code from [`silogen/ai-samples`](https://github.com/silogen/ai-samples)
28+
- **Paper:** Bodnar et al., *A foundation model of the Earth system*, Nature 2025, https://doi.org/10.1038/s41586-025-08897-0
29+
- **ROCm blog:** [Running SOTA AI-based Weather Forecasting models on AMD Instinct](https://rocm.blogs.amd.com/artificial-intelligence/ai-weather-forecasting/README.html) — Luka Tsabadze, Rahul Biswas, Pauli Pihajoki, Daniel Warna, Baiqiang Xia (AMD Silo AI)
30+
31+
### GenCast
32+
- **Upstream:** [`google-deepmind/graphcast`](https://github.com/google-deepmind/graphcast); recipe code from [`silogen/ai-samples`](https://github.com/silogen/ai-samples)
33+
- **Paper:** Price et al., *GenCast: Diffusion-based ensemble weather forecasting for improved prediction accuracy and uncertainty quantification*, Nature 2025, https://doi.org/10.1038/s41586-024-08252-9
34+
- **ROCm blog:** [Running SOTA AI-based Weather Forecasting models on AMD Instinct](https://rocm.blogs.amd.com/artificial-intelligence/ai-weather-forecasting/README.html) — Luka Tsabadze, Rahul Biswas, Pauli Pihajoki, Daniel Warna, Baiqiang Xia (AMD Silo AI)
35+
36+
### NeuralGCM
37+
- **Upstream:** [`google-research/neuralgcm`](https://github.com/google-research/neuralgcm); weights via Google Cloud Storage
38+
- **Papers:**
39+
- Kochkov et al., *Neural General Circulation Models for Weather and Climate*, arXiv:2311.07222
40+
- Kochkov et al., *Neural GCMs optimized to predict satellite-based precipitation*, arXiv:2412.11973
41+
42+
### PanguWeather
43+
- **Upstream:** [`198808xc/Pangu-Weather`](https://github.com/198808xc/Pangu-Weather); recipe code from [`silogen/ai-samples`](https://github.com/silogen/ai-samples)
44+
- **Paper:** Bi et al., *Accurate medium-range global weather forecasting with 3D neural networks*, Nature 2023, https://doi.org/10.1038/s41586-023-06185-3
45+
- **ROCm blog:** [Running SOTA AI-based Weather Forecasting models on AMD Instinct](https://rocm.blogs.amd.com/artificial-intelligence/ai-weather-forecasting/README.html) — Luka Tsabadze, Rahul Biswas, Pauli Pihajoki, Daniel Warna, Baiqiang Xia (AMD Silo AI)
46+
47+
---
48+
49+
## Material Science
50+
51+
### MatterGen
52+
- **Upstream:** [`microsoft/mattergen`](https://github.com/microsoft/mattergen)
53+
- **Paper:** Zeni et al., *MatterGen: a generative model for inorganic materials design*, Nature 2025, https://doi.org/10.1038/s41586-025-08628-5
54+
- **ROCm blog:** [Running MatterGen on AMD Instinct](https://rocm.blogs.amd.com/artificial-intelligence/mattergen/README.html) — Sopiko Kurdadze (AMD Silo AI)
55+
56+
### HydraGNN
57+
- **Upstream:** [`ORNL/HydraGNN`](https://github.com/ORNL/HydraGNN) (branch `Predictive_GFM_2024`); weights on [Hugging Face](https://huggingface.co/mlupopa/HydraGNN_Predictive_GFM_2024)
58+
- **Dataset DOI:** [10.13139/OLCF/2474799](https://doi.org/10.13139/OLCF/2474799)
59+
- **Citation:** M. Lupo Pasini et al., *HydraGNN_Predictive_GFM_2024 — Ensemble of predictive graph foundation models for group state atomistic materials modeling*, DOI 10.13139/OLCF/2474799
60+
- **Maintained by:** Oak Ridge National Laboratory (ORNL)
61+
62+
---
63+
64+
## Healthcare & Life Sciences
65+
66+
### GP-MoLFormer
67+
- **Upstream:** [`IBM/gp-molformer`](https://github.com/IBM/gp-molformer)
68+
- **Paper:** Ross et al., *Generative Pre-trained Transformer for De Novo Drug Design and Molecular Property Optimization*, arXiv:2302.07432
69+
- **ROCm blog:** [Running GP-MoLFormer on AMD Instinct](https://rocm.blogs.amd.com/artificial-intelligence/gp-molformer/README.html) — Sopiko Kurdadze (AMD Silo AI)
70+
71+
### SwinUNETR
72+
- **Upstream:** [`Project-MONAI/research-contributions`](https://github.com/Project-MONAI/research-contributions); recipe code from [`silogen/ai-samples`](https://github.com/silogen/ai-samples)
73+
- **Paper:** Hatamizadeh et al., *Swin UNETR: Swin Transformers for Semantic Segmentation of Brain Tumors in MRI Images*, arXiv:2201.01266
74+
- **ROCm blog (training):** [Running SwinUNETR on AMD Instinct](https://rocm.blogs.amd.com/artificial-intelligence/running-swinunetr-amd/README.html) — Joaquin Rives Gambin (AMD Silo AI)
75+
- **ROCm blog (inference optimization):** [SwinUNETR Inference Optimization on AMD Instinct](https://rocm.blogs.amd.com/artificial-intelligence/swinunetr-inference-optimization/README.html) — Joaquin Rives Gambin, Vasumathi Neralla, David Björelind, Rui Sampaio (AMD Silo AI / AstraZeneca × AMD collaboration)
76+
- **AstraZeneca × AMD collaboration:** https://www.amd.com/en/blogs/2025/astrazeneca-improved-life-sciences-model-training-time.html
77+
78+
### SemlaFlow
79+
- **Upstream:** [`rssrwn/semla-flow`](https://github.com/rssrwn/semla-flow)
80+
- **Paper:** Morehead et al., *SemlaFlow: Efficient 3D Molecular Generation with Latent Attention and Equivariant Flow Matching*, OpenReview: https://openreview.net/forum?id=bee2G6pEh0
81+
- **ROCm blog:** [Running SemlaFlow on AMD Instinct](https://rocm.blogs.amd.com/artificial-intelligence/semlaflow/README.html) — Vasumathi Neralla, Rui Sampaio (AMD Silo AI / AstraZeneca × AMD collaboration)
82+
- **AstraZeneca × AMD collaboration:** https://www.amd.com/en/blogs/2025/astrazeneca-improved-life-sciences-model-training-time.html
83+
84+
### REINVENT4
85+
- **Upstream:** [`MolecularAI/REINVENT4`](https://github.com/MolecularAI/REINVENT4)
86+
- **Paper:** Loeffler et al., *REINVENT4: Modern AI-driven generative molecule design*, J Cheminformatics 2024, https://doi.org/10.1186/s13321-024-00812-5
87+
- **ROCm blog:** [Running REINVENT4 on AMD Instinct](https://rocm.blogs.amd.com/artificial-intelligence/running-reinvent4-amd/README.html) — David Björelind, Rui Sampaio (AMD Silo AI / AstraZeneca × AMD collaboration)
88+
- **AstraZeneca × AMD collaboration:** https://www.amd.com/en/blogs/2025/astrazeneca-improved-life-sciences-model-training-time.html
89+
90+
---
91+
92+
## Physics Simulation
93+
94+
### MATEY
95+
- **Upstream:** [`ORNL/MATEY`](https://github.com/ORNL/MATEY)
96+
- **Paper:** Subramanian et al., *MATEY: multiscale adaptive foundation models for spatiotemporal physical systems*, arXiv:2412.20601
97+
- **Maintained by:** Oak Ridge National Laboratory (ORNL)
98+
99+
### Walrus
100+
- **Upstream:** [`PolymathicAI/walrus`](https://github.com/PolymathicAI/walrus); weights on [Hugging Face](https://huggingface.co/polymathic-ai/walrus)
101+
- **Paper:** McCabe et al., *Walrus: A Cross-Domain Foundation Model for Continuum Dynamics*, arXiv:2511.15684
102+
- **Maintained by:** Polymathic AI
103+
104+
---
105+
106+
## AMD Silo AI
107+
108+
ROCm-specific recipes for many models in this repository were authored by the [AMD Silo AI](https://www.amd.com/en/solutions/ai/silo-ai.html) team and published on the [ROCm Blogs](https://rocm.blogs.amd.com) platform. Contributors across these recipes include: Pauli Pihajoki, Luka Tsabadze, Rahul Biswas, Sopiko Kurdadze, Daniel Warna, Baiqiang Xia, Joaquin Rives Gambin, Vasumathi Neralla, David Björelind, and Rui Sampaio.
109+
110+
## AstraZeneca × AMD Collaboration
111+
112+
The recipes for REINVENT4, SemlaFlow, and the SwinUNETR inference optimization were developed as part of a collaboration between AstraZeneca and AMD. See the [collaboration blog post](https://www.amd.com/en/blogs/2025/astrazeneca-improved-life-sciences-model-training-time.html) for background. The SwinUNETR training recipe and GP-MoLFormer are AMD Silo AI contributions independent of this collaboration.

‎_template/README.md‎

Lines changed: 14 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -16,3 +16,17 @@ Rename the parent folder from `_template` to your **model slug** (see the domain
1616
Put training, fine-tuning, inference, or evaluation scripts (or step-by-step docs) under `recipes/`. Prefer one subfolder per task, e.g. `recipes/inference/`, `recipes/finetune/`.
1717

1818
Do not commit large checkpoints or datasets; document how to obtain them instead.
19+
20+
## Acknowledgements and citation
21+
22+
Every model entry must include attribution. Fill in this section before opening a PR:
23+
24+
```
25+
- **Upstream repo:** [org/repo](https://github.com/org/repo)
26+
- **Paper:** Author et al., *Title*, Venue Year, https://doi.org/... (or arXiv:XXXX.XXXXX)
27+
- **Cite as:** paste BibTeX or DOI-based citation from the upstream repo/model card
28+
- **ROCm blog (if applicable):** [Title](https://rocm.blogs.amd.com/...) — Author Name (AMD Silo AI)
29+
- **Collaboration (if applicable):** e.g. AstraZeneca × AMD, ORNL × AMD
30+
```
31+
32+
Also add an entry to [`ACKNOWLEDGEMENTS.md`](../../ACKNOWLEDGEMENTS.md) at the repo root following the existing per-model format.

0 commit comments

Comments
 (0)