test_that("out of bound error message", { x <- torch_tensor(1) funs <- list( torch_sum, torch_mean, torch_median, torch_max, torch_min, torch_cummax, torch_cummin, torch_mode, torch_size, torch_softmax, torch_std, torch_squeeze, torch_var ) for (f in funs) { expect_error( f(x, dim = 2), regex = "Dimension out of range (expected to be in range of [-1, 1], but got 2)", fixed = TRUE ) } for (f in funs) { expect_error( f(x, dim = -2), regex = "Dimension out of range (expected to be in range of [-1, 1], but got -2)", fixed = TRUE ) } x <- torch_randn(size = rep(1, 10)) for (f in funs) { expect_error( f(x, dim = 11), regex = "Dimension out of range (expected to be in range of [-10, 10], but got 11)", fixed = TRUE ) } }) test_that("more than 1 dim", { x <- torch_randn(2, 2, 2) expect_error( torch_sum(x, dim = c(1, 4)), regex = "Dimension out of range (expected to be in range of [-3, 3], but got 4)", fixed = TRUE ) }) test_that("dim1 & dim2", { x <- torch_randn(2, 2) expect_error( torch_transpose(x, 3, 1), regex = "Dimension out of range (expected to be in range of [-2, 2], but got 3)", fixed = TRUE ) expect_error( torch_transpose(x, 2, 3), regex = "Dimension out of range (expected to be in range of [-2, 2], but got 3)", fixed = TRUE ) expect_error( torch_diag_embed(x, dim1 = 4, dim2 = 1), "Dimension out of range (expected to be in range of [-3, 3], but got 4)", fixed = TRUE ) expect_error( torch_diag_embed(x, dim1 = 3, dim2 = 4), "Dimension out of range (expected to be in range of [-3, 3], but got 4)", fixed = TRUE ) }) test_that("dimension x does not have size y", { a <- torch_randn(c(4, 3)) b <- torch_randn(c(4, 3)) expect_error( torch_cross(a, b, dim = 1), regex = "inputs dimension 1 must have length 3", fixed = TRUE ) }) test_that("indices error message", { x <- torch_randn(4, 4, 2) e <- torch_max_pool2d_with_indices(x, kernel_size = c(2, 2)) i <- as_array(torch_max(e[[2]])$to(torch_int())) indices <- torch_ones_like(e[[2]]) indices[1, ..] <- 35L expect_error( torch_max_unpool2d(e[[1]], indices = indices, output_size = c(4, 4)), paste0("Found an invalid max index: ", 35), fixed = TRUE ) }) test_that("torch_flatten dim error", { x <- torch_randn(2, 2) expect_error( torch_flatten(x, start_dim = 3), "Dimension out of range (expected to be in range of [-2, 2], but got 3)", fixed = TRUE ) }) test_that("index argument", { x <- torch_tensor(c(1, 2, 3)) expect_equal(as_array(torch_select(x, 1, 1)), 1) expect_error( torch_select(x, 1, 4), regex = "index 4 out of range for tensor of size [3] at dimension 1", fixed = TRUE ) }) test_that("torch_nll_loss out of bound", { x <- torch_randn(1, 5) expect_error( torch_nll_loss(x, torch_tensor(0, dtype = torch_long())), regex = "Indexing starts at 1 but found a 0.", fixed = TRUE ) expect_error( torch_nll_loss(x, torch_tensor(6, dtype = torch_long())), regex = "Target 6 is out of bounds.", fixed = TRUE ) }) test_that("tensordot error message", { a <- torch_arange(start = 1, end = 60)$reshape(c(3, 4, 5)) b <- torch_arange(start = 1, end = 24)$reshape(c(4, 3, 2)) expect_error( torch_tensordot(a, b, list(c(2, 1), c(1, 3))), regex = "contracted dimensions need to match, but first has size 3 in dim 1 and second has size 2 in dim 3", fixed = TRUE ) }) test_that("embedding returns a better error message", { e <- nn_embedding(10, 3) x <- torch_tensor(c(0, 1, 2, 3), dtype = torch_long()) expect_error( e(x), regex = "Indexing starts at 1 but found a 0." ) }) test_that("movedim", { x <- torch_randn(3, 2, 1) expect_error( torch_movedim(x, 0, 1), regex = "Dimension is 1-based, but found 0.", class = "value_error" ) expect_error( torch_movedim(x, 4, 3), regexp = "Dimension out of range (expected to be in range of [-3, 3], but got 4)", fixed = TRUE ) }) test_that("subsetting empty tensors", { x <- torch_tensor(integer()) expect_error( x[1], regexp = "index 1 is out of bounds for dimension 1 with size 0", fixed = TRUE ) expect_error( x[-1], regexp = "index -1 is out of bounds for dimension 1 with size 0", fixed = TRUE ) }) test_that("translate", { x <- torch_tensor(c(2), requires_grad = TRUE) x$register_hook(function(grad) { torch_tensor(1)$sum(dim = 4) }) y <- 2 * x expect_error( y$backward(), regexp = "Dimension out of range (expected to be in range of [-1, 1], but got 4)", fixed = TRUE ) }) test_that("cat", { expect_error( torch_cat(list(torch_randn(8, 2, 7), torch_randn(8, 3, 7)), dim = 1), regex = "Sizes of tensors must match except in dimension 1. Expected size 2 but got size 3 for tensor number 2 in the list.", fixed = TRUE ) })