Update BubbleML_2.py

BubbleML_2.py

@@ -19,9 +19,8 @@ from datasets import (
 
 _CITATION = ""  # optional
 _DESCRIPTION = """
-BubbleML: high-fidelity boiling simulations for 3 Liquids- (FC72 & R515B & LN2)
-and Flow Boiling Regimes.
-Pre-defined train/test splits across all benchmarks.
+BubbleML: high-fidelity single-bubble boiling simulations (FC72 & R515B).
+Pre-defined train/test splits across two directories.
 """
 
 class BubbleMLConfig(BuilderConfig):
@@ -48,7 +47,7 @@ class BubbleMLDataset(GeneratorBasedBuilder):
         BubbleMLConfig(
             name="single-bubble",
             description="Single-bubble (FC72 & R515B) train/test split",
-            data_dir="",  # repo root when loading remotely; overridden by --data_dir in CLI
+            data_dir="",  # use repo root locally; if empty, remote download is used
             data_files={
                 "train": [
                     # FC72 train
@@ -85,7 +84,6 @@ class BubbleMLDataset(GeneratorBasedBuilder):
     ]
 
     def _info(self) -> DatasetInfo:
-        # Nested Sequence to represent 4D arrays (time, channel, H, W) of floats
         features = Features({
             "input": Sequence(Sequence(Sequence(Sequence(Value("float32"))))),
             "output": Sequence(Sequence(Sequence(Sequence(Value("float32"))))),
@@ -101,60 +99,69 @@ class BubbleMLDataset(GeneratorBasedBuilder):
         )
 
     def _split_generators(self, dl_manager: datasets.DownloadManager):
-        base_dir = (
-            self.config.data_dir
-            if self.config.data_dir
-            else os.path.dirname(os.path.abspath(__file__))
-        )
+        cfg = self.config
 
-        def resolve(split_name: str):
-            out = []
-            for rel in self.config.data_files[split_name]:
-                full = os.path.join(base_dir, rel)
-                if not os.path.isfile(full):
-                    raise FileNotFoundError(f"Expected data file at {full}, but not found.")
-                out.append(full)
-            return out
+        if cfg.data_dir:
+            # Local testing: use the provided data_dir
+            base_dir = cfg.data_dir
+            def resolve_local(split):
+                paths = []
+                for rel in cfg.data_files[split]:
+                    full = os.path.join(base_dir, rel)
+                    if not os.path.isfile(full):
+                        raise FileNotFoundError(f"Expected {full}, but it was not found.")
+                    paths.append((full, full.replace(".hdf5", ".json")))
+                return paths
+            train_pairs = resolve_local("train")
+            test_pairs  = resolve_local("test")
+        else:
+            # Remote download: fetch each .hdf5 and .json from the Hub
+            base_url = "https://huggingface.co/datasets/hpcforge/BubbleML_2/resolve/main/"
+            # build map of rel -> url
+            url_map = {}
+            for split in ["train", "test"]:
+                for rel in cfg.data_files[split]:
+                    url_map[rel] = base_url + rel
+                    meta = rel.replace(".hdf5", ".json")
+                    url_map[meta] = base_url + meta
+            # download all
+            downloaded = dl_manager.download(url_map)
+            # resolve into pairs
+            train_pairs = []
+            test_pairs  = []
+            for split in ["train", "test"]:
+                for rel in cfg.data_files[split]:
+                    h5_path   = downloaded[rel]
+                    json_path = downloaded[rel.replace(".hdf5", ".json")]
+                    train_pairs.append((h5_path, json_path)) if split == "train" else test_pairs.append((h5_path, json_path))
 
         return [
-            datasets.SplitGenerator(
-                name=Split.TRAIN, gen_kwargs={"files": resolve("train")}
-            ),
-            datasets.SplitGenerator(
-                name=Split.TEST, gen_kwargs={"files": resolve("test")}
-            ),
+            datasets.SplitGenerator(name=Split.TRAIN, gen_kwargs={"file_pairs": train_pairs}),
+            datasets.SplitGenerator(name=Split.TEST,  gen_kwargs={"file_pairs": test_pairs}),
         ]
 
-    def _generate_examples(self, files):
-        """Yield examples as dicts with input, output, fluid_params, filename."""
-        for idx, path in enumerate(files):
-            # Load HDF5 arrays
-            with h5py.File(path, "r") as h5f:
+    def _generate_examples(self, file_pairs):
+        """Yield examples from each (hdf5,json) pair."""
+        for idx, (h5_path, json_path) in enumerate(file_pairs):
+            with h5py.File(h5_path, "r") as h5f:
                 inp = np.stack([h5f[k][:5] for k in ["dfun", "temperature", "velx", "vely"]])
                 out = np.stack([h5f[k][5:10] for k in ["dfun", "temperature", "velx", "vely"]])
-
-            # Load metadata JSON
-            meta_path = path.replace(".hdf5", ".json")
-            if os.path.isfile(meta_path):
-                with open(meta_path, "r") as jf:
-                    p = json.load(jf)
-                fluid_params = [
-                    p["inv_reynolds"],
-                    p["cpgas"],
-                    p["mugas"],
-                    p["rhogas"],
-                    p["thcogas"],
-                    p.get("stefan", 0.0),
-                    p["prandtl"],
-                    p["heater"]["nucWaitTime"],
-                    p["heater"].get("wallTemp", 0.0),
-                ]
-            else:
-                fluid_params = []
-
+            with open(json_path, "r") as jf:
+                p = json.load(jf)
+            fluid_params = [
+                p["inv_reynolds"],
+                p["cpgas"],
+                p["mugas"],
+                p["rhogas"],
+                p["thcogas"],
+                p["stefan"],
+                p["prandtl"],
+                p["heater"]["nucWaitTime"],
+                p["heater"]["wallTemp"],
+            ]
             yield idx, {
                 "input": inp.astype(np.float32).tolist(),
                 "output": out.astype(np.float32).tolist(),
                 "fluid_params": fluid_params,
-                "filename": os.path.basename(path),
+                "filename": os.path.basename(h5_path),
             }