♻️ Modernize the Hugging Face bridge

CHANGELOG.md

@@ -15,11 +15,13 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Changed
 
+- (huggingface bridge) restructuration: do not rely on binary blobs anymore, but exploit native arrow types by flattening cgns trees into constant and variable parts.
 - (sample) Restructuring of the Sample class to store a global (tensor of arbitrary order) at a given time step: replaces scalar and time_series. All Sample data are now stored in CGNS trees.
 
 ### Fixes
 
 - (meshes) fix `get_field_name`, could overwrite arguments during iteration over times, bases, zones and locations.
+- (docs) explain release process in Contributing page.
 
 ### Removed
 

examples/bridges/check_retrocomp_benchmarks.py

@@ -0,0 +1,15 @@
+"""This files serves to check if the main retrieval command in the PLAID Benchmarks
+is not returning an error."""
+
+from plaid.bridges import huggingface_bridge
+
+hf_dataset = huggingface_bridge.load_dataset_from_hub(
+    f"PLAID-datasets/Tensile2d", split="all_samples[:5]", num_proc=1
+)
+
+plaid_dataset, pb_def = huggingface_bridge.huggingface_dataset_to_plaid(
+    hf_dataset, processes_number=1, verbose=True
+)
+
+ids_train = pb_def.get_split('train_500')
+sample_train_0 = plaid_dataset[ids_train[0]]