From cc99aab607cd5310ae0dd9468475e1d084fa9eda Mon Sep 17 00:00:00 2001
From: Anas Khan <83116240+anxkhn@users.noreply.github.com>
Date: Mon, 6 Jul 2026 19:41:57 +0530
Subject: [PATCH] fix: correct class name in SyntheticDataKit.chunk_data guard
 message (#6901)

---
 tests/test_synthetic_chunk_data.py | 26 ++++++++++++++++++++++++++
 unsloth/dataprep/synthetic.py      |  2 +-
 2 files changed, 27 insertions(+), 1 deletion(-)

diff --git a/tests/test_synthetic_chunk_data.py b/tests/test_synthetic_chunk_data.py
index b9167d214..abc2c0144 100644
--- a/tests/test_synthetic_chunk_data.py
+++ b/tests/test_synthetic_chunk_data.py
@@ -104,10 +104,36 @@ def test_chunk_data_rejects_overlap_not_smaller_than_chunk():
         os.unlink(path)
 
 
+def test_chunk_data_uninitialized_error_names_real_class():
+    # Without max_seq_length the guard tells the user which method to call first.
+    # The message must name the real class (SyntheticDataKit) so copying it works;
+    # a misspelling would raise NameError when the user follows it verbatim.
+    kit = SyntheticDataKit.__new__(SyntheticDataKit)
+    kit.tokenizer = _MockTokenizer()  # max_seq_length intentionally unset
+    with tempfile.NamedTemporaryFile("w", suffix = ".txt", delete = False) as f:
+        f.write("word " * 50)
+        path = f.name
+    try:
+        try:
+            kit.chunk_data(filename = path)
+            raise AssertionError("expected RuntimeError when max_seq_length is unset")
+        except RuntimeError as e:
+            msg = str(e)
+            assert (
+                "SyntheticDataKit.from_pretrained" in msg
+            ), f"error must name SyntheticDataKit.from_pretrained, got: {msg}"
+            assert (
+                "SynthetidDataKit" not in msg
+            ), f"error must not misspell the class name, got: {msg}"
+    finally:
+        os.unlink(path)
+
+
 if __name__ == "__main__":
     test_chunk_data_keeps_single_chunk_document()
     test_chunk_data_still_splits_long_document()
     test_chunk_data_empty_document_yields_no_chunks()
     test_chunk_data_short_document_is_not_split_into_fragments()
     test_chunk_data_rejects_overlap_not_smaller_than_chunk()
+    test_chunk_data_uninitialized_error_names_real_class()
     print("OK")
diff --git a/unsloth/dataprep/synthetic.py b/unsloth/dataprep/synthetic.py
index 10690810f..6f025343f 100644
--- a/unsloth/dataprep/synthetic.py
+++ b/unsloth/dataprep/synthetic.py
@@ -391,7 +391,7 @@ class SyntheticDataKit:
         assert os.path.exists(filename)
         assert hasattr(self, "tokenizer")
         if not hasattr(self, "max_seq_length"):
-            raise RuntimeError("Please use SynthetidDataKit.from_pretrained(...) first!")
+            raise RuntimeError("Please use SyntheticDataKit.from_pretrained(...) first!")
         if not hasattr(self, "overlap") or not hasattr(self, "max_generation_tokens"):
             raise RuntimeError("Please use prepare_qa_generation first!")