Upload app (Copy).py

app (Copy).py

@@ -173,6 +173,15 @@ def generate_response(model, human_prompt, tokenizer = None):
 
     messages.pop()
     return english, japanese
+    
+from llama_cpp import Llama
+
+llm = Llama.from_pretrained(
+	repo_id="google/gemma-3-4b-it-qat-q4_0-gguf",
+    filename='gemma-3-4b-it-q4_0.gguf',
+	local_dir='/kaggle/working/model',
+    n_ctx = 2048
+)
 
 if __name__ == '__main__':
     device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')