bug fix

app.py

@@ -8,13 +8,14 @@ os.environ["TOKENIZERS_PARALLELISM"] = "false"  # Hindari beban CPU
 torch.set_num_threads(2)  # Sesuai limit Spaces gratis
 
 # Load model & tokenizer
-# model_name_or_path = "TheBloke/Llama-2-7B-Chat-GPTQ"
 print("PyTorch Version",torch.__version__)  # Versi PyTorch
 print("Is GPU Available",torch.cuda.is_available())  # Apakah GPU terdeteksi?
 print("CPU cores:", psutil.cpu_count())
 print("RAM (GB):", psutil.virtual_memory().total / (1024**3))
 
-model_name_or_path = "TheBloke/Llama-2-7B-Chat-GGUF"
+# model_name_or_path = "TheBloke/Llama-2-7B-Chat-GPTQ"
+model_name_or_path = "meta-llama/Llama-2-7b-chat-hf"
+# model_name_or_path = "TheBloke/Llama-2-7B-Chat-GGUF"
 # model_name_or_path = "TheBloke/Mistral-7B-v0.1-GPTQ"
 # model_name_or_path = "unsloth/DeepSeek-R1-0528-GGUF" # 3x lebih cepat dari Mistral-7B
 # tokenizer = AutoTokenizer.from_pretrained(model_name_or_path, use_fast=True)