Update README.md

README.md

@@ -1,3 +1,10 @@
+---
+license: llama2
+pipeline_tag: text-generation
+tags:
+- code
+- automated program repair
+---
 # CodeLlama-70B_for_NTR
 
 We fine-tuned [CodeLlama-70B](https://huggingface.co/codellama/CodeLlama-70b-hf) on [Transfer_dataset](https://drive.google.com/drive/folders/1F1BPfTxHDGX-OCBthudCbu_6Qvcg_fbP?usp=drive_link) under the [NTR](https://sites.google.com/view/neuraltemplaterepair) framework for APR research.
@@ -108,11 +115,11 @@ generated_ids = model.generate(
 for generated_id in generated_ids:
     generated_text = tokenizer.decode(generated_id, skip_special_tokens=False)
     patch = generated_text.split(E_INST)[1]
-    patch = text.replace(tokenizer.eos_token,'')
+    patch = patch.replace(tokenizer.eos_token,'')
     print(patch)
 
 
 ```
 
 ## Model Details
-*Note: Use of this model is governed by the Meta license. Meta developed and publicly released the Code Llama family of large language models (LLMs).
+*Note: Use of this model is governed by the Meta license. Meta developed and publicly released the Code Llama family of large language models (LLMs).