diff --git a/gemma.py b/gemma.py
index 86f6bd1..aeaa6c8 100644
--- a/gemma.py
+++ b/gemma.py
@@ -38,7 +38,7 @@ dataset = create_training_data()
 
 # 5️⃣ Ładowanie modelu Gemma 2 7B
 device = "cuda" if torch.cuda.is_available() else "cpu"
-model_name = "google/gemma-2b"
+model_name = "google/gemma-2-2b"
 model = AutoModelForCausalLM.from_pretrained(model_name, torch_dtype=torch.float16).to(device)
 tokenizer = AutoTokenizer.from_pretrained(model_name)