Add LithoBench_PDE.py

LithoBench_PDE.py

@@ -0,0 +1,133 @@
+"""
+LithoBench-PDE: PyTorch Dataset for computational lithography PDE benchmark.
+
+Usage (local):
+    from LithoBench_PDE import LithoBenchPDE
+    ds = LithoBenchPDE("./LithoBench_PDE", categories=["Contact_I"])
+    sample = ds[0]  # dict with M, E, h, m, R, T, sample_id, category
+
+Usage (from HuggingFace Hub):
+    from LithoBench_PDE import LithoBenchPDE
+    ds = LithoBenchPDE.from_hub("AISDL-SNU/LithoBench-PDE", categories=["Contact_I"])
+    sample = ds[0]
+"""
+
+import os
+import glob
+from typing import Optional
+
+import torch
+from torch.utils.data import Dataset
+
+
+CATEGORIES = [
+    "Contact_I",
+    "Contact_T",
+    "Metal_I",
+    "Metal_T",
+]
+
+TASKS = {
+    "maxwell": ("M", "E"),
+    "reaction_diffusion": ("h", "m"),
+    "eikonal": ("R", "T"),
+}
+
+
+class LithoBenchPDE(Dataset):
+    """PyTorch Dataset for LithoBench-PDE.
+
+    Args:
+        root: Path to the root directory containing category subdirectories.
+        categories: List of categories to include. None = all categories.
+        task: If set, one of 'maxwell', 'reaction_diffusion', 'eikonal'.
+              Returns only (input, target) for that task instead of all fields.
+    """
+
+    def __init__(
+        self,
+        root: str,
+        categories: Optional[list[str]] = None,
+        task: Optional[str] = None,
+    ):
+        self.root = root
+        self.task = task
+
+        if task is not None and task not in TASKS:
+            raise ValueError(f"Unknown task '{task}'. Choose from {list(TASKS.keys())}")
+
+        cats = categories or CATEGORIES
+        self.files = []
+        for cat in cats:
+            cat_dir = os.path.join(root, cat)
+            if not os.path.isdir(cat_dir):
+                raise FileNotFoundError(f"Category directory not found: {cat_dir}")
+            pt_files = sorted(glob.glob(os.path.join(cat_dir, "*.pt")))
+            for f in pt_files:
+                self.files.append((cat, f))
+
+    @classmethod
+    def from_hub(
+        cls,
+        repo_id: str = "AISDL-SNU/LithoBench-PDE",
+        categories: Optional[list[str]] = None,
+        task: Optional[str] = None,
+        cache_dir: Optional[str] = None,
+    ):
+        """Download from HuggingFace Hub and create dataset.
+
+        Args:
+            repo_id: HuggingFace dataset repository ID.
+            categories: List of categories to download. None = all.
+            task: Optional PDE task filter.
+            cache_dir: Local cache directory for downloaded files.
+        """
+        from huggingface_hub import snapshot_download
+
+        cats = categories or CATEGORIES
+        patterns = [f"LithoBench_PDE/{cat}/*.pt" for cat in cats]
+
+        local_dir = snapshot_download(
+            repo_id,
+            repo_type="dataset",
+            allow_patterns=patterns,
+            cache_dir=cache_dir,
+        )
+        root = os.path.join(local_dir, "LithoBench_PDE")
+        return cls(root, categories=categories, task=task)
+
+    def __len__(self):
+        return len(self.files)
+
+    def __getitem__(self, idx):
+        category, filepath = self.files[idx]
+        sample_id = os.path.splitext(os.path.basename(filepath))[0]
+        data = torch.load(filepath, map_location="cpu", weights_only=False)
+
+        if self.task is not None:
+            inp_key, tgt_key = TASKS[self.task]
+            return {
+                "input": data[inp_key],
+                "target": data[tgt_key],
+                "sample_id": sample_id,
+                "category": category,
+            }
+
+        return {
+            "M": data["M"],
+            "E": data["E"],
+            "h": data["h"],
+            "m": data["m"],
+            "R": data["R"],
+            "T": data["T"],
+            "sample_id": sample_id,
+            "category": category,
+        }
+
+    def __repr__(self):
+        cats = sorted(set(cat for cat, _ in self.files))
+        task_str = f", task='{self.task}'" if self.task else ""
+        return (
+            f"LithoBenchPDE(samples={len(self)}, "
+            f"categories={cats}{task_str})"
+        )