feat: update algorithm integrations and point cloud overlays

alg/平整度、平面度、轮廓测量v1.5/evaluation.py

@@ -74,15 +74,23 @@ class Evaluator:
     ) -> tuple[np.ndarray, DistanceStats]:
         device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
         dtype = torch.float64
+
+        # vertices/triangles 上 GPU（mesh 通常几百MB，可以接受）
         vertices_t = torch.as_tensor(np.asarray(vertices, dtype=np.float64), dtype=dtype, device=device)
         faces_t = torch.as_tensor(np.asarray(faces, dtype=np.int64), dtype=torch.long, device=device)
         triangles = vertices_t[faces_t]
-        points_t = torch.as_tensor(points, dtype=dtype, device=device)
+        del vertices_t, faces_t  # 释放原始 vertices/faces，只保留 triangles
+
+        # points 保留在 CPU，按 point_chunk 逐批传上 GPU，避免 10 亿点全量上 GPU OOM
+        points_np = np.asarray(points, dtype=np.float64)
+        n_points = points_np.shape[0]
 
         distances = []
         with torch.no_grad():
-            for start in range(0, points_t.shape[0], point_chunk_size):
-                chunk = points_t[start:start + point_chunk_size]
+            for start in range(0, n_points, point_chunk_size):
+                chunk = torch.as_tensor(
+                    points_np[start:start + point_chunk_size], dtype=dtype, device=device
+                )
                 if triangle_chunk_size is None or triangle_chunk_size <= 0:
                     dist2 = self._point_to_triangle_distance_squared(chunk, triangles)
                     min_dist2 = dist2.min(dim=1).values
@@ -90,7 +98,7 @@ class Evaluator:
                     min_dist2 = torch.full(
                         (chunk.shape[0],),
                         torch.inf,
-                        dtype=points_t.dtype,
+                        dtype=dtype,
                         device=device,
                     )
                     for triangle_start in range(0, triangles.shape[0], triangle_chunk_size):