Update app.py

app.py

@@ -23,15 +23,20 @@ def extract_text_from_docx(docx_file):
 def generate_roast(resume_text):
     tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-neo-1.3B")
     model = AutoModelForCausalLM.from_pretrained("EleutherAI/gpt-neo-1.3B")
+
+    # Define the prompt separately
+    prompt_text = "Roast this resume:\n\n"
     
-    # Tokenize the resume text and truncate it to 2048 tokens
-    inputs = tokenizer(resume_text, return_tensors="pt", truncation=True, max_length=2048)
+    # Tokenize the prompt and resume text, truncating to fit within the 2048 token limit
+    max_tokens = 2048 - len(tokenizer(prompt_text)['input_ids'])  # Reserve space for the prompt
+    inputs = tokenizer(resume_text, return_tensors="pt", truncation=True, max_length=max_tokens)
 
     # Convert the tokenized inputs back to text for the prompt
     truncated_resume_text = tokenizer.decode(inputs["input_ids"][0], skip_special_tokens=True)
 
-    prompt = f"Roast this resume:\n\n{truncated_resume_text}\n\nRoast:"
-    
+    # Create the final prompt by combining the prompt and the truncated resume
+    prompt = f"{prompt_text}{truncated_resume_text}\n\nRoast:"
+
     generator = pipeline('text-generation', model=model, tokenizer=tokenizer)
     
     # Generate roast
@@ -41,6 +46,7 @@ def generate_roast(resume_text):
 
 
 
+
 # Gradio interface function
 def roast_resume(file):
     if file.name.endswith('.pdf'):