Update README.md

README.md

@@ -18,22 +18,18 @@ Load the model weights from HuggingFace:
 ```python
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
+device = 'cuda'
 SCAR = AutoModelForCausalLM.from_pretrained(
     "AIML-TUDA/SCAR",
     trust_remote_code=True,
-    device_map = 'cuda',
+    device_map = device,
 )
-```
-
-The model loaded model is based on LLama3-8B base. So we can use the tokenizer from it:
-
-```python
 tokenizer = AutoTokenizer.from_pretrained(
         "meta-llama/Meta-Llama-3-8B", padding_side="left"
     )
 tokenizer.pad_token = tokenizer.eos_token
 text = "This is text."
-inputs = tokenizer(text, return_tensors="pt", padding=True).to('cuda')
+inputs = tokenizer(text, return_tensors="pt", padding=True).to(device)
 ```
 
 To modify the latent feature $h_0$ (`SCAR.hook.mod_features = 0`) of the SAE do the following: