llm.py corrected. llama-cpp-python version set to 0.3.29

app/models/llm.py

@@ -44,7 +44,7 @@ def load_model() -> Llama:
             return _model
 
         print("⬇️ [load_model] downloading model...", flush=True)
-        model_path = hf_hub_download(repo_id=HF_REPO, filename=HF_FILENAME)
+        model_path = hf_hub_download(repo_id=HF_REPO, filename=HF_FILENAME, force_download=True,)
         print(f"✅ [load_model] model downloaded at {model_path}", flush=True)
 
         _preload_cuda_libs()

requirements.txt

@@ -3,7 +3,7 @@
 
 nvidia-cublas-cu12
 nvidia-cuda-runtime-cu12
-llama-cpp-python==0.3.23
+llama-cpp-python==0.3.29
 
 gradio>=5.0.0
 spaces>=0.43.0