update inference.py to return correct SQL without worrying about prompt

inference.py

@@ -39,6 +39,7 @@ def run_inference(question, prompt_file="prompt.md", metadata_file="metadata.sql
         tokenizer=tokenizer,
         max_new_tokens=300,
         do_sample=False,
+        return_full_text=False, # added return_full_text parameter to prevent splitting issues with prompt
         num_beams=5, # do beam search with 5 beams for high quality results
     )
     generated_query = (
@@ -48,8 +49,8 @@ def run_inference(question, prompt_file="prompt.md", metadata_file="metadata.sql
             eos_token_id=eos_token_id,
             pad_token_id=eos_token_id,
         )[0]["generated_text"]
-        .split("```")[1]
         .split(";")[0]
+        .split("```")[0]
         .strip()
         + ";"
     )