tests

R/paramset_torchlearner.R

@@ -70,7 +70,7 @@ paramset_torchlearner = function(task_type) {
     collate_fn            = p_uty(tags = c("train", "predict"), default = NULL),
     pin_memory            = p_lgl(default = FALSE, tags = c("train", "predict")),
     drop_last             = p_lgl(tags = "train", default = FALSE),
-    timeout               = p_int(default = -1L, tags = c("train_predict")),
+    timeout               = p_int(default = -1L, tags = c("train", "predict")),
     worker_init_fn        = p_uty(tags = c("train", "predict")),
     worker_globals        = p_uty(tags = c("train", "predict")),
     worker_packages       = p_uty(tags = c("train", "predict"), custom_check = check_character, special_vals = list(NULL))

tests/testthat/helper_learner.R

@@ -29,20 +29,29 @@ LearnerTorchTest1 = R6Class("LearnerTorchTest1",
     .dataloader = function(task, param_vals) {
       ingress_token = TorchIngressToken(task$feature_names, batchgetter_num, c(NA, length(task$feature_names)))
       dataset = task_dataset(
-        task,
+       task,
         feature_ingress_tokens = list(num = ingress_token),
         target_batchgetter = crate(function(data, device) {
           torch_tensor(data = as.integer(data[[1]]), dtype = torch_long(), device = device)
         }),
         device = param_vals$device
       )
-      dl = dataloader(
-        dataset = dataset,
-        batch_size = param_vals$batch_size,
-        shuffle = param_vals$shuffle
+      dl_args = c(
+        "batch_size",
+        "shuffle",
+        "sampler",
+        "batch_sampler",
+        "num_workers",
+        "collate_fn",
+        "pin_memory",
+        "drop_last",
+        "timeout",
+        "worker_init_fn",
+        "worker_globals",
+        "worker_packages"
       )
-      return(dl)
-
+      args = param_vals[names(param_vals) %in% dl_args]
+      invoke(dataloader, dataset = dataset)
     }
   )
 )

tests/testthat/test_paramset_torchlearner.R

@@ -2,10 +2,6 @@ test_that("paramset works", {
   test_ps = function(param_set) {
     expect_r6(param_set, "ParamSet")
     expect_true(all(map_lgl(param_set$tags, function(tags) "train" %in% tags || "predict" %in% tags)))
-    # all parameters are required and have initial values set
-    expect_true(all(map_lgl(param_set$tags, function(tags) "required" %in% tags)))
-    # only parameters batch_size and epochs don't have initial values
-    expect_true(length(param_set$values) == param_set$length - 2)
   }
   param_set_regr = paramset_torchlearner("regr")