avoid gpu memory overflow

app.py

@@ -89,6 +89,8 @@ def evaluate(
     **kwargs,
 ):
     content = process_webpage(url=url)
+    # avoid GPU memory overflow
+    torch.cuda.empty_cache()
     prompt = generate_prompt(instruction, content)
     inputs = tokenizer(prompt, return_tensors="pt")
     input_ids = inputs["input_ids"].to(device)
@@ -109,6 +111,8 @@ def evaluate(
         )
     s = generation_output.sequences[0]
     output = tokenizer.decode(s)
+    # avoid GPU memory overflow
+    torch.cuda.empty_cache()
     return output.split("### Response:")[1].strip()
 
 

scrape_website.py

@@ -1,6 +1,7 @@
 import requests
 from bs4 import BeautifulSoup
 
+
 def process_webpage(url:str):
     # A set to keep track of visited pages
     visited_pages = set()