René's URL Explorer Experiment

{"@context":"https://schema.org","@type":"DiscussionForumPosting","headline":"Inconsistent DirectSolver interface.","articleBody":"Hi everyone! I've noticed a small inconsistency in the sparse solver interface, namely when it comes to using it with a 2D tensor where the last dimension is 1.\n\nGiven a 3x3 lhs sparse matrix and a 3x2 rhs dense matrix:\n```\na = torch.sparse_coo_tensor([[0, 1, 2], [0, 1, 2]], [1.0, 1.0, 1.0], device=\"cuda\").to_sparse_csr()\nb = torch.randn([3, 2], device=\"cuda\")\n```\nwe can set up a solver for with:\n```\nnvsparse.DirectSolver(a, b.mT.contiguous().mT)\n```\nThis works fine. However, for a 3x1 rhs:\n```\nb = torch.randn([3,1], device=\"cuda\")\n```\nsetting up the solver in the same way breaks with:\n\n```\nTypeError(\"The RHS tensor([[1.],\\n        [1.],\\n        [1.]], device='cuda:0') must be a matrix or vector with col-major layout, and for implicitly-batched RHS (N-D \u003e= 3), each matrix sample must have col-major layout (the second dimension from the end must have unit stride.\")\n```\n\nIt would be great not to have to special-case this.\n\nBest,\nAna\n","author":{"url":"https://github.com/anadodik","@type":"Person","name":"anadodik"},"datePublished":"2026-01-15T16:04:38.000Z","interactionStatistic":{"@type":"InteractionCounter","interactionType":"https://schema.org/CommentAction","userInteractionCount":3},"url":"https://github.com/53/nvmath-python/issues/53"}