Update app.py

app.py

@@ -15,17 +15,18 @@ tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)
 # Load the Lora model
 model = PeftModel.from_pretrained(model, peft_model_id)
 
+prompt = f"Adapt my Cover Letter ```\n{my_cover_letter}\n```\nto this Job Posting\n```\n{job_posting}\n```"
 
 def make_inference(my_cover_letter, job_posting):
     batch = tokenizer(
-        f"Adapt my Cover Letter ```\n{my_cover_letter}\n```\nto this Job Posting\n```\n{job_posting}\n```",
+        prompt,
         return_tensors="pt",
     )
 
     with torch.cuda.amp.autocast():
         output_tokens = model.generate(**batch, max_new_tokens=300)
 
-    return tokenizer.decode(output_tokens[0], skip_special_tokens=True)
+    return tokenizer.decode(output_tokens[0], skip_special_tokens=True).replace(prompt, '')
 
 
 if __name__ == "__main__":