feat: MinHash+LSH 去重 + Strategy/Factory 存储后端

2026-07-04 14:47:38 +08:00
parent 816c02dab0
commit 900cd91798
6 changed files with 344 additions and 13 deletions
@@ -4,16 +4,26 @@ This module provides:
 - FileScanner: File and directory scanning utilities
 - HDF5Handler: Tensor data persistence
 - export_dataset: HuggingFace Dataset to JSONL export
- cache_jsonl: JSONL to HDF5 tokenization and caching
+- cache_jsonl: JSONL to HDF5/binary tokenization and caching
+- dedup_jsonl: MinHash+LSH deduplication for pretraining text
+- writers: BaseWriter / H5Writer / BinWriter / TextWriter (Strategy + Factory)
 """

 from pipeline.io.file_scanner import FileScanner
 from pipeline.io.hdf5_handler import HDF5Handler
 from pipeline.io.export import export_dataset, cache_jsonl
+from pipeline.io.dedup import dedup_jsonl
+from pipeline.io.writers import BaseWriter, H5Writer, BinWriter, TextWriter, create_writer

 __all__ = [
    "FileScanner",
    "HDF5Handler",
    "export_dataset",
    "cache_jsonl",
+    "dedup_jsonl",
+    "BaseWriter",
+    "H5Writer",
+    "BinWriter",
+    "TextWriter",
+    "create_writer",
 ]