remove sample configs from auto import

fast_llm/data/auto.py

@@ -17,7 +17,3 @@
 from fast_llm.data.preparator.dataset_discovery.config import DatasetDiscoveryConfig  # isort: skip
 from fast_llm.data.preparator.gpt_memmap.config import GPTMemmapDatasetPreparatorConfig  # isort: skip
 from fast_llm.data.sample.abstract import NullReaderConfig  # isort: skip
-from fast_llm.data.sample.language_model import LanguageModelReaderConfig  # isort: skip
-from fast_llm.data.sample.patch import PatchReaderConfig  # isort: skip
-from fast_llm.data.sample.range import RangeReaderConfig  # isort: skip
-from fast_llm.data.sample.token import TokenReaderConfig  # isort: skip

fast_llm/data/dataset/memmap.py

@@ -28,6 +28,12 @@ def read_reader_config(path: pathlib.Path | str) -> MemmapIndexDatasetReaderConf
         """
         Read the MemmapIndexDatasetReaderConfig from a memmap file.
         """
+        # Import reader configs to register them in the dynamic class registry
+        from fast_llm.data.sample.language_model import LanguageModelReaderConfig  # noqa: F401
+        from fast_llm.data.sample.patch import PatchReaderConfig  # noqa: F401
+        from fast_llm.data.sample.range import RangeReaderConfig  # noqa: F401
+        from fast_llm.data.sample.token import TokenReaderConfig  # noqa: F401
+
         path = pathlib.Path(path) if isinstance(path, str) else path
         with path.open("rb") as stream:
             # Verify file type.

fast_llm/data/preparator/dataset_discovery/__init__.py

@@ -1,4 +0,0 @@
-from fast_llm.data.preparator.dataset_discovery.config import DatasetDiscoveryConfig
-from fast_llm.data.preparator.dataset_discovery.prepare import DatasetDiscoveryPreparator
-
-__all__ = ["DatasetDiscoveryConfig", "DatasetDiscoveryPreparator"]