fabiencasenave/2D_Multiscale_Hyperelasticity_plaid_0.1.14

---
license: cc-by-sa-4.0
task_categories:
- graph-ml
pretty_name: 2D quasistatic non-linear structural mechanics with finite elasticity
and topology variations
tags:
- physics learning
- geometry learning
dataset_info:
features:
- name: Base_2_2/Zone
list:
list: int64
- name: Base_2_2/Zone/Elements_TRI_3/ElementConnectivity
list: int64
- name: Base_2_2/Zone/Elements_TRI_3/ElementRange
list: int64
- name: Base_2_2/Zone/GridCoordinates/CoordinateX
list: float32
- name: Base_2_2/Zone/GridCoordinates/CoordinateY
list: float32
- name: Base_2_2/Zone/VertexFields/P11
list: float32
- name: Base_2_2/Zone/VertexFields/P12
list: float32
- name: Base_2_2/Zone/VertexFields/P21
list: float32
- name: Base_2_2/Zone/VertexFields/P22
list: float32
- name: Base_2_2/Zone/VertexFields/psi
list: float32
- name: Base_2_2/Zone/VertexFields/u1
list: float32
- name: Base_2_2/Zone/VertexFields/u2
list: float32
- name: Base_2_2/Zone/ZoneBC/Ext_bound/PointList
list:
list: int32
- name: Base_2_2/Zone/ZoneBC/Holes/PointList
list:
list: int32
- name: Global/C11
list: float32
- name: Global/C12
list: float32
- name: Global/C22
list: float32
- name: Global/effective_energy
list: float32
splits:
- name: train
num_bytes: 360004988
num_examples: 764
- name: test
num_bytes: 117157624
num_examples: 376
download_size: 477234161
dataset_size: 477162612
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
---
<p align='center'>
<img src='https://i.ibb.co/JR7bWLNn/Logo-2-D-Multiscale-Hyperelasticity-2-consolas-100.png' alt='https://i.ibb.co/JR7bWLNn/Logo-2-D-Multiscale-Hyperelasticity-2-consolas-100.png' width='1000'/>
<img src='https://i.ibb.co/zHFQFnPR/2-D-Multiscale-Hyperelasticity.png' alt='https://i.ibb.co/zHFQFnPR/2-D-Multiscale-Hyperelasticity.png' width='1000'/>
</p>

```yaml
legal:
owner: Safran
license: cc-by-sa-4.0
data_production:
type: simulation
physics: 2D quasistatic non-linear structural mechanics, finite elasticity (large
strains), P1 elements, compressible hyperelastic material
simulator: fenics
num_samples:
train: 764
test: 376
storage_backend: hf_datasets
plaid:
version: 0.1.13.dev31+gda966e466

```
This dataset was generated with [`plaid`](https://plaid-lib.readthedocs.io/), we refer to this documentation for additional details on how to extract data from `plaid_sample` objects.

The simplest way to use this dataset is to first download it:
```python
from plaid.storage import download_from_hub

repo_id = "channel/dataset"
local_folder = "downloaded_dataset"

download_from_hub(repo_id, local_folder)
```

Then, to iterate over the dataset and instantiate samples:
```python
from plaid.storage import init_from_disk

local_folder = "downloaded_dataset"
split_name = "train"

datasetdict, converterdict = init_from_disk(local_folder)

dataset = datasetdict[split]
converter = converterdict[split]

for i in range(len(dataset)):
plaid_sample = converter.to_plaid(dataset, i)
```

It is possible to stream the data directly:
```python
from plaid.storage import init_streaming_from_hub

repo_id = "channel/dataset"

datasetdict, converterdict = init_streaming_from_hub(repo_id)

dataset = datasetdict[split]
converter = converterdict[split]

for sample_raw in dataset:
plaid_sample = converter.sample_to_plaid(sample_raw)
```

Plaid samples' features can be retrieved like the following:
```python
from plaid.storage import load_problem_definitions_from_disk
local_folder = "downloaded_dataset"
pb_defs = load_problem_definitions_from_disk(local_folder)

# or
from plaid.storage import load_problem_definitions_from_hub
repo_id = "channel/dataset"
pb_defs = load_problem_definitions_from_hub(repo_id)

pb_def = pb_defs[0]

plaid_sample = ... # use a method from above to instantiate a plaid sample

for t in plaid_sample.get_all_time_values():
for path in pb_def.get_in_features_identifiers():
plaid_sample.get_feature_by_path(path=path, time=t)
for path in pb_def.get_out_features_identifiers():
plaid_sample.get_feature_by_path(path=path, time=t)
```

For those familiar with HF's `datasets` library, raw data can be retrieved without using the `plaid` library:
```python
from datasets import load_dataset

repo_id = "channel/dataset"

datasetdict = load_dataset(repo_id)

for split_name, dataset in datasetdict.items():
for raw_sample in dataset:
for feat_name in dataset.column_names:
feature = raw_sample[feat_name]
```
Notice that raw data refers to the variable features only, with a specific encoding for time variable features.

### Dataset Sources

- **Papers:**
- [arxiv](https://arxiv.org/abs/2505.02974)