[ML][Python] Optimize eager laoding setup by removing redundant RDF triggers

Author: siliataider

bindings/pyroot/pythonizations/python/ROOT/_pythonization/_ml_dataloader.py

@@ -175,19 +175,6 @@ def __init__(
                     given value is {validation_split}"
             )
 
-        if load_eager:
-            # TODO: overhead, check if we can improve the following lines
-            if sampling_type == "undersampling" and not replacement:
-                rdf_0 = rdataframes[0].Count().GetValue()
-                rdf_1 = rdataframes[1].Count().GetValue()
-                rdf_minor = min(rdf_0, rdf_1)
-                rdf_major = max(rdf_0, rdf_1)
-                if rdf_major < rdf_minor / sampling_ratio:
-                    raise ValueError(
-                        f"The sampling_ratio is too low: not enough entries in the majority class to sample from. \n \
-                        Choose sampling_ratio > {round(rdf_minor / rdf_major, 3)} or set replacement to False."
-                    )
-
         if not hasattr(rdataframes, "__iter__"):
             rdataframes = [rdataframes]
         self.noded_rdfs = [RDF.AsRNode(rdf) for rdf in rdataframes]

tree/ml/inc/ROOT/ML/RSampler.hxx

@@ -103,6 +103,14 @@ public:
       fNumMajor = fDatasets[fMajor].GetRows();
       fNumMinor = fDatasets[fMinor].GetRows();
       fNumResampledMajor = static_cast<std::size_t>(fNumMinor / fSampleRatio);
+      if (!fReplacement && fNumResampledMajor > fNumMajor) {
+         auto minRatio = std::to_string(std::round(double(fNumMinor) / double(fNumMajor) * 100.0) / 100.0);
+         minRatio.erase(minRatio.find('.') + 3);
+         throw std::invalid_argument(
+            "The sampling_ratio is too low: not enough entries in the majority class to sample from.\n"
+            "Choose sampling_ratio > " +
+            minRatio + " or set replacement to True.");
+      }
       fNumEntries = fNumMinor + fNumResampledMajor;
    }