Fixes for broken examples (#202)

* Try to fix gaas-map

* Added retrying logic for hamiltonian example

* Fix linters

* Try to fix pet-mad-uq

Author: ceriottm

examples/gaas-map/gaas-map.py

@@ -35,6 +35,7 @@
 from sklearn.linear_model import RidgeCV
 from skmatter.decomposition import PCovR
 from skmatter.preprocessing import StandardFlexibleScaler
+from urllib3.util.retry import Retry
 
 
 # %%
@@ -45,13 +46,31 @@
 # train a ML potential to describe the full composition range.
 #
 
-filename = "gaas_training.xyz"
-if not os.path.exists(filename):
-    url = f"https://zenodo.org/records/10566825/files/{filename}"
-    response = requests.get(url)
+
+def fetch_dataset(filename, base_url, local_path=""):
+    """Helper function to load the pre-computed examples"""
+
+    local_file = local_path + filename
+    if os.path.isfile(local_file):
+        return
+
+    # Retry strategy: wait 1s, 2s, 4s, 8s, 16s on 429/5xx errors
+    retry_strategy = Retry(
+        total=5, backoff_factor=1, status_forcelist=[429, 500, 502, 503, 504]
+    )
+    session = requests.Session()
+    session.mount("https://", requests.adapters.HTTPAdapter(max_retries=retry_strategy))
+
+    # Fetch with automatic retry and error raising
+    response = session.get(base_url + filename)
     response.raise_for_status()
-    with open(filename, "wb") as f:
-        f.write(response.content)
+
+    with open(local_file, "wb") as file:
+        file.write(response.content)
+
+
+filename = "gaas_training.xyz"
+fetch_dataset(filename, "https://zenodo.org/records/10566825/files/")
 
 structures = ase.io.read(filename, ":")
 energy = np.array([f.info["energy"] for f in structures])

examples/hamiltonian-qm7/hamiltonian-qm7.py

@@ -87,6 +87,7 @@
 from mlelec.features.acdc import compute_features_for_target
 from mlelec.targets import drop_zero_blocks  # noqa: F401
 from mlelec.utils.plot_utils import plot_losses
+from urllib3.util.retry import Retry
 
 
 os.environ["PYSCFAD_BACKEND"] = "torch"
@@ -188,15 +189,33 @@ def save_parameters(file_path, **params):
 # We first download the data for the two examples from Zenodo
 # and unzip the downloaded datafile.
 
-if not os.path.exists("hamiltonian-qm7-data"):
-    url = r"https://zenodo.org/records/15524259/files/hamiltonian-qm7-data.zip"
-    response = requests.get(url)
+
+def fetch_dataset(filename, base_url, local_path=""):
+    """Helper function to load the pre-computed examples"""
+
+    local_file = local_path + filename
+    if os.path.isfile(local_file):
+        return
+
+    # Retry strategy: wait 1s, 2s, 4s, 8s, 16s on 429/5xx errors
+    retry_strategy = Retry(
+        total=5, backoff_factor=1, status_forcelist=[429, 500, 502, 503, 504]
+    )
+    session = requests.Session()
+    session.mount("https://", requests.adapters.HTTPAdapter(max_retries=retry_strategy))
+
+    # Fetch with automatic retry and error raising
+    response = session.get(base_url + filename)
     response.raise_for_status()
-    with open("hamiltonian-qm7-data.zip", "wb") as f:
-        f.write(response.content)
 
-    with ZipFile("hamiltonian-qm7-data.zip", "r") as zObject:
-        zObject.extractall(path=".")
+    with open(local_file, "wb") as file:
+        file.write(response.content)
+
+
+fetch_dataset("hamiltonian-qm7-data.zip", "https://zenodo.org/records/15524259/files/")
+
+with ZipFile("hamiltonian-qm7-data.zip", "r") as zObject:
+    zObject.extractall(path=".")
 
 # %%
 # Prepare the Dataset for ML Training

examples/pet-mad-uq/pet-mad-uq.py

@@ -87,8 +87,7 @@
 import subprocess
 from urllib.request import urlretrieve
 
-import ase.cell
-import ase.ga.utilities
+import ase.geometry.rdf
 import matplotlib.pyplot as plt
 import numpy as np
 import pandas as pd
@@ -363,7 +362,7 @@
 for atoms in frames:
 
     # Compute H-H distances
-    bins, xs = ase.ga.utilities.get_rdf(  # type: ignore
+    bins, xs = ase.geometry.rdf.get_rdf(  # type: ignore
         atoms, 4.5, num_bins, elements=[1, 1]
     )
 
@@ -378,7 +377,7 @@
     rdfs_hh.append(bins * 3.0 / 2.0)  # correct ASE normalization
 
     # Compute O-O distances
-    bins, xs = ase.ga.utilities.get_rdf(  # type: ignore
+    bins, xs = ase.geometry.rdf.get_rdf(  # type: ignore
         atoms, 4.5, num_bins, elements=[8, 8]
     )
     bins[2:-2] = (