Revert "Enable model test for CUDA now that they run fast (#6759)" (#6769)

This reverts commit b217e66e67f78bb761d8320cacd3defaad3b84f2.

iree/test/e2e/models/BUILD

diff --git a/iree/test/e2e/models/BUILD b/iree/test/e2e/models/BUILD
index 1732ac0..976a358 100644
--- a/iree/test/e2e/models/BUILD
+++ b/iree/test/e2e/models/BUILD
@@ -72,20 +72,3 @@
     driver = "vulkan",
     target_backend = "vulkan-spirv",
 )
-
-iree_check_single_backend_test_suite(
-    name = "check_linalg_on_tensors_cuda_cuda",
-    timeout = "long",
-    srcs = CHECK_FRAMEWORK_TESTS,
-    compiler_flags = ["-iree-input-type=mhlo"],
-    driver = "cuda",
-    tags = [
-        # CUDA cuInit fails with sanitizer on.
-        "noasan",
-        "nomsan",
-        "notsan",
-        "noubsan",
-        "requires-gpu-nvidia",
-    ],
-    target_backend = "cuda",
-)

iree/test/e2e/models/CMakeLists.txt

diff --git a/iree/test/e2e/models/CMakeLists.txt b/iree/test/e2e/models/CMakeLists.txt
index 3975477..e8d3983 100644
--- a/iree/test/e2e/models/CMakeLists.txt
+++ b/iree/test/e2e/models/CMakeLists.txt
@@ -57,24 +57,4 @@
     "-iree-input-type=mhlo"
 )
 
-iree_check_single_backend_test_suite(
-  NAME
-    check_linalg_on_tensors_cuda_cuda
-  SRCS
-    "bert_encoder_unrolled_fake_weights.mlir"
-    "mobilenetv3_fake_weights.mlir"
-  TARGET_BACKEND
-    "cuda"
-  DRIVER
-    "cuda"
-  COMPILER_FLAGS
-    "-iree-input-type=mhlo"
-  LABELS
-    "noasan"
-    "nomsan"
-    "notsan"
-    "noubsan"
-    "requires-gpu-nvidia"
-)
-
 ### BAZEL_TO_CMAKE_PRESERVES_ALL_CONTENT_BELOW_THIS_LINE ###