Update balanced_positive_negative_sampler.py (#93)

use cuda version torch.randperm to avoid copy from gpu to cpu and a fatal bug in multi-thread cpu version

maskrcnn_benchmark/modeling/balanced_positive_negative_sampler.py

@@ -46,8 +46,8 @@ def __call__(self, matched_idxs):
             num_neg = min(negative.numel(), num_neg)
 
             # randomly select positive and negative examples
-            perm1 = torch.randperm(positive.numel())[:num_pos]
-            perm2 = torch.randperm(negative.numel())[:num_neg]
+            perm1 = torch.randperm(positive.numel(), device=positive.device)[:num_pos]
+            perm2 = torch.randperm(negative.numel(), device=negative.device)[:num_neg]
 
             pos_idx_per_image = positive[perm1]
             neg_idx_per_image = negative[perm2]