Update README.md

README.md

@@ -30,16 +30,21 @@ Llama3-2-3B-IT-Byte performs competitively although it has been trained only on
 from transformers import AutoTokenizer, AutoModelForCausalLM
 
 tokenizer = AutoTokenizer.from_pretrained("benjamin/Llama3-2-3B-IT-Byte")
-print("Vocab Size:", len(tokenizer)) # 256 bytes + some special tokens
+print("Vocab Size:", len(tokenizer))  # 256 bytes + some special tokens
 
 device = "cuda"
-model = AutoModelForCausalLM.from_pretrained("benjamin/Llama3-2-3B-IT-Byte", trust_remote_code=True)
+model = AutoModelForCausalLM.from_pretrained(
+    "benjamin/Llama3-2-3B-IT-Byte", trust_remote_code=True
+)
 model = model.to(device)
 
-tokens = tokenizer.apply_chat_template([{"role": "user", "content": "Hello, how are you doing?"}], return_tensors="pt")
+tokens = tokenizer.apply_chat_template(
+    [{"role": "user", "content": "Hello, how are you doing?"}], return_tensors="pt"
+)
 eot_id = tokenizer.convert_tokens_to_ids("<|eot_id|>")
 out = model.generate(tokens.to(model.device), eos_token_id=eot_id)
 print(tokenizer.decode(out[0]))
+
 ```
 
 ## Training