modified inference preprocessing

src/mini_transformer/inference.py

@@ -115,6 +115,23 @@ def _get_or_create_session(
     return transformer, tokenizer
 
 
+def _preprocess_text(text: str) -> str:
+    """Lowercase and ensure sentence ends with punctuation.
+
+    Simple preprocessing used before tokenization during inference:
+    - Lowercase the entire input.
+    - If the last non-space character is not one of '.!?', append a period.
+    """
+    s = text.strip()
+    if not s:
+        return s
+    s = s.lower()
+    last = s[-1]
+    if last not in ".!?":
+        s += "."
+    return s
+
+
 def run_inference(cfg: DictConfig) -> list[str]:
     """Run inference using a composed Hydra configuration."""
 
@@ -151,7 +168,7 @@ def run_inference(cfg: DictConfig) -> list[str]:
         device=device,
     )
 
-    text_input = scfg.input_text
+    text_input = _preprocess_text(scfg.input_text)
     if not text_input:
         raise SystemExit('Pass text like: input_text="hello world"')