utils to support METIS with adj_wgt

csrc/cpu/metis_cpu.cpp

@@ -6,9 +6,8 @@
 
 #include "utils.h"
 
-torch::Tensor partition_cpu(torch::Tensor rowptr, torch::Tensor col,
-                            int64_t num_parts, torch::Tensor adjwgt,
-                            bool recursive) {
+torch::Tensor partition_cpu(torch::Tensor rowptr, torch::Tensor col,int64_t num_parts,
+                            torch::Tensor edge_wgt,  bool recursive) {
 #ifdef WITH_METIS
   CHECK_CPU(rowptr);
   CHECK_CPU(col);
@@ -18,7 +17,7 @@ torch::Tensor partition_cpu(torch::Tensor rowptr, torch::Tensor col,
 
   auto *xadj = rowptr.data_ptr<int64_t>();
   auto *adjncy = col.data_ptr<int64_t>();
-  auto *adjwgt = adjwgt.data_ptr<int64_t>();
+  auto *adjwgt = edge_wgt.data_ptr<int64_t>();
   int64_t ncon = 1;
   int64_t objval = -1;
   auto part_data = part.data_ptr<int64_t>();

csrc/metis.cpp

@@ -7,8 +7,8 @@
 PyMODINIT_FUNC PyInit__metis(void) { return NULL; }
 #endif
 
-torch::Tensor partition(torch::Tensor rowptr, torch::Tensor col,
-                        int64_t num_parts, bool recursive) {
+torch::Tensor partition(torch::Tensor rowptr, torch::Tensor col,int64_t num_parts,
+                        torch::Tensor edge_wgt,bool recursive) {
   if (rowptr.device().is_cuda()) {
 #ifdef WITH_CUDA
     AT_ERROR("No CUDA version supported");
@@ -16,7 +16,7 @@ torch::Tensor partition(torch::Tensor rowptr, torch::Tensor col,
     AT_ERROR("Not compiled with CUDA support");
 #endif
   } else {
-    return partition_cpu(rowptr, col, num_parts, recursive);
+    return partition_cpu(rowptr, col,num_parts, edge_wgt, recursive);
   }
 }
 

setup.py

@@ -80,7 +80,7 @@ tests_require = ['pytest', 'pytest-cov']
 
 setup(
     name='torch_sparse',
-    version='0.6.0',
+    version='0.6.1',
     author='Matthias Fey',
     author_email='matthias.fey@tu-dortmund.de',
     url='https://github.com/rusty1s/pytorch_sparse',

torch_sparse/metis.py

@@ -3,15 +3,28 @@ from typing import Tuple
 import torch
 from torch_sparse.tensor import SparseTensor
 from torch_sparse.permute import permute
+from torch_sparse.utils import cartesian1d
+
+
+def metis_wgt(x):
+    t1, t2 = cartesian1d(x, x)
+    diff = t1 - t2
+    diff = diff[diff != 0]
+    res = diff.abs().min()
+    bod = x.max() - x.min()
+    scale = (res / bod).item()
+    tick, arange = scale.as_integer_ratio()
+    x_ratio = (x - x.min()) / bod
+    return (x_ratio * arange + tick).long(), tick, arange
 
 
 def partition(
         src: SparseTensor, num_parts: int, recursive: bool = False
 ) -> Tuple[SparseTensor, torch.Tensor, torch.Tensor]:
-
     rowptr, col = src.storage.rowptr().cpu(), src.storage.col().cpu()
-    adjwgt=src.storage.value().cpu()
-    cluster = torch.ops.torch_sparse.partition(rowptr, col, num_parts,adjwgt,
+    edge_wgt = src.storage.value().cpu()
+    edge_wgt = metis_wgt(edge_wgt)[0]
+    cluster = torch.ops.torch_sparse.partition(rowptr, col, num_parts, edge_wgt,
                                                recursive)
     cluster = cluster.to(src.device())
 

torch_sparse/utils.py

 from typing import Any
+import torch
 
 try:
     from typing_extensions import Final  # noqa
@@ -8,3 +9,9 @@ except ImportError:
 
 def is_scalar(other: Any) -> bool:
     return isinstance(other, int) or isinstance(other, float)
+
+
+def cartesian1d(x, y):
+    a1, a2 = torch.meshgrid([x, y])
+    coos = torch.stack([a1, a2]).T.reshape(-1, 2)
+    return coos.split(1, dim=1)