diff --git a/MSCOCO.py b/MSCOCO.py
index 7b8d034..e20184c 100644
--- a/MSCOCO.py
+++ b/MSCOCO.py
@@ -836,7 +836,7 @@ class MsCocoConfig(ds.BuilderConfig):
     TASKS: Tuple[str, ...] = (
         "captions",
         "instances",
-        "person_keypoints",
+        "person-keypoints",
     )
 
     def __init__(
@@ -894,7 +894,7 @@ def get_processor(self) -> MsCocoProcessor:
             return CaptionsProcessor()
         elif self.task == "instances":
             return InstancesProcessor()
-        elif self.task == "person_keypoints":
+        elif self.task == "person-keypoints":
             return PersonKeypointsProcessor()
         else:
             raise ValueError(f"Invalid task: {self.task}")
@@ -924,7 +924,7 @@ def dataset_configs(year: int, version: ds.Version) -> List[MsCocoConfig]:
         ),
         MsCocoConfig(
             year=year,
-            coco_task="person_keypoints",
+            coco_task="person-keypoints",
             version=version,
         ),
         # MsCocoConfig(
@@ -934,7 +934,7 @@ def dataset_configs(year: int, version: ds.Version) -> List[MsCocoConfig]:
         # ),
         # MsCocoConfig(
         #     year=year,
-        #     coco_task=("captions", "person_keypoints"),
+        #     coco_task=("captions", "person-keypoints"),
         #     version=version,
         # ),
     ]
diff --git a/tests/MSCOCO_caption_test.py b/tests/MSCOCO_caption_test.py
new file mode 100644
index 0000000..438c7f7
--- /dev/null
+++ b/tests/MSCOCO_caption_test.py
@@ -0,0 +1,47 @@
+import os
+
+import datasets as ds
+import pytest
+
+
+@pytest.mark.skipif(
+    condition=bool(os.environ.get("CI", False)),
+    reason=(
+        "Because this loading script downloads a large dataset, "
+        "we will skip running it on CI."
+    ),
+)
+@pytest.mark.parametrize(
+    argnames=(
+        "dataset_year",
+        "coco_task",
+        "expected_num_train",
+        "expected_num_validation",
+    ),
+    argvalues=(
+        (2014, "captions", 82783, 40504),
+        (2017, "captions", 118287, 5000),
+    ),
+)
+def test_load_caption_dataset(
+    dataset_path: str,
+    dataset_year: int,
+    coco_task: str,
+    expected_num_train: int,
+    expected_num_validation: int,
+    repo_id: str,
+):
+    dataset = ds.load_dataset(
+        path=dataset_path,
+        year=dataset_year,
+        coco_task=coco_task,
+    )
+    assert isinstance(dataset, ds.DatasetDict)
+
+    assert dataset["train"].num_rows == expected_num_train
+    assert dataset["validation"].num_rows == expected_num_validation
+
+    dataset.push_to_hub(
+        repo_id=repo_id,
+        config_name=f"year={dataset_year}_task={coco_task}",
+    )
diff --git a/tests/MSCOCO_instances_test.py b/tests/MSCOCO_instances_test.py
new file mode 100644
index 0000000..e0ccf74
--- /dev/null
+++ b/tests/MSCOCO_instances_test.py
@@ -0,0 +1,56 @@
+import os
+
+import datasets as ds
+import pytest
+
+
+@pytest.mark.skipif(
+    condition=bool(os.environ.get("CI", False)),
+    reason=(
+        "Because this loading script downloads a large dataset, "
+        "we will skip running it on CI."
+    ),
+)
+@pytest.mark.parametrize(
+    argnames="decode_rle,",
+    argvalues=(
+        True,
+        False,
+    ),
+)
+@pytest.mark.parametrize(
+    argnames=(
+        "dataset_year",
+        "coco_task",
+        "expected_num_train",
+        "expected_num_validation",
+    ),
+    argvalues=(
+        (2014, "instances", 82081, 40137),
+        (2017, "instances", 117266, 4952),
+    ),
+)
+def test_load_instances_dataset(
+    dataset_path: str,
+    dataset_year: int,
+    coco_task: str,
+    decode_rle: bool,
+    expected_num_train: int,
+    expected_num_validation: int,
+    repo_id: str,
+):
+    dataset = ds.load_dataset(
+        path=dataset_path,
+        year=dataset_year,
+        coco_task=coco_task,
+        decode_rle=decode_rle,
+    )
+    assert isinstance(dataset, ds.DatasetDict)
+
+    assert dataset["train"].num_rows == expected_num_train
+    assert dataset["validation"].num_rows == expected_num_validation
+
+    dataset.push_to_hub(
+        repo_id=repo_id,
+        config_name=f"year={dataset_year}_task={coco_task}_decode-rle={decode_rle}",
+    )
diff --git a/tests/MSCOCO_keypoints_test.py b/tests/MSCOCO_keypoints_test.py
new file mode 100644
index 0000000..338a9e2
--- /dev/null
+++ b/tests/MSCOCO_keypoints_test.py
@@ -0,0 +1,56 @@
+import os
+
+import datasets as ds
+import pytest
+
+
+@pytest.mark.skipif(
+    condition=bool(os.environ.get("CI", False)),
+    reason=(
+        "Because this loading script downloads a large dataset, "
+        "we will skip running it on CI."
+    ),
+)
+@pytest.mark.parametrize(
+    argnames="decode_rle,",
+    argvalues=(
+        True,
+        False,
+    ),
+)
+@pytest.mark.parametrize(
+    argnames=(
+        "dataset_year",
+        "coco_task",
+        "expected_num_train",
+        "expected_num_validation",
+    ),
+    argvalues=(
+        (2014, "person-keypoints", 45174, 21634),
+        (2017, "person-keypoints", 64115, 2693),
+    ),
+)
+def test_load_keypoints_dataset(
+    dataset_path: str,
+    dataset_year: int,
+    coco_task: str,
+    decode_rle: bool,
+    expected_num_train: int,
+    expected_num_validation: int,
+    repo_id: str,
+):
+    dataset = ds.load_dataset(
+        path=dataset_path,
+        year=dataset_year,
+        coco_task=coco_task,
+        decode_rle=decode_rle,
+    )
+    assert isinstance(dataset, ds.DatasetDict)
+
+    assert dataset["train"].num_rows == expected_num_train
+    assert dataset["validation"].num_rows == expected_num_validation
+
+    dataset.push_to_hub(
+        repo_id=repo_id,
+        config_name=f"year={dataset_year}_task={coco_task}_decode-rle={decode_rle}",
+    )
diff --git a/tests/MSCOCO_test.py b/tests/MSCOCO_test.py
index 56fd48d..79922ae 100644
--- a/tests/MSCOCO_test.py
+++ b/tests/MSCOCO_test.py
@@ -1,64 +1,6 @@
-import os
-
-import datasets as ds
-import pytest
-
 from MSCOCO import CATEGORIES, SUPER_CATEGORIES
 
 
-@pytest.fixture
-def dataset_path() -> str:
-    return "MSCOCO.py"
-
-
-@pytest.mark.skipif(
-    condition=bool(os.environ.get("CI", False)),
-    reason=(
-        "Because this loading script downloads a large dataset, "
-        "we will skip running it on CI."
-    ),
-)
-@pytest.mark.parametrize(
-    argnames="decode_rle,",
-    argvalues=(
-        True,
-        False,
-    ),
-)
-@pytest.mark.parametrize(
-    argnames=(
-        "dataset_year",
-        "coco_task",
-        "expected_num_train",
-        "expected_num_validation",
-    ),
-    argvalues=(
-        (2014, "captions", 82783, 40504),
-        (2017, "captions", 118287, 5000),
-        (2014, "instances", 82081, 40137),
-        (2017, "instances", 117266, 4952),
-        (2014, "person_keypoints", 45174, 21634),
-        (2017, "person_keypoints", 64115, 2693),
-    ),
-)
-def test_load_dataset(
-    dataset_path: str,
-    dataset_year: int,
-    coco_task: str,
-    decode_rle: bool,
-    expected_num_train: int,
-    expected_num_validation: int,
-):
-    dataset = ds.load_dataset(
-        path=dataset_path,
-        year=dataset_year,
-        coco_task=coco_task,
-        decode_rle=decode_rle,
-    )
-    assert dataset["train"].num_rows == expected_num_train
-    assert dataset["validation"].num_rows == expected_num_validation
-
-
 def test_consts():
     assert len(CATEGORIES) == 80
     assert len(SUPER_CATEGORIES) == 12
diff --git a/tests/conftest.py b/tests/conftest.py
new file mode 100644
index 0000000..f28111b
--- /dev/null
+++ b/tests/conftest.py
@@ -0,0 +1,21 @@
+import pytest
+
+
+@pytest.fixture
+def org_name() -> str:
+    return "shunk031"
+
+
+@pytest.fixture
+def dataset_name() -> str:
+    return "MSCOCO"
+
+
+@pytest.fixture
+def dataset_path(dataset_name: str) -> str:
+    return f"{dataset_name}.py"
+
+
+@pytest.fixture
+def repo_id(org_name: str, dataset_name: str) -> str:
+    return f"{org_name}/{dataset_name}"