Allow local models to be loaded

app.py

@@ -47,12 +47,18 @@ def generate_text(prompt, example):
 
 parser = argparse.ArgumentParser(formatter_class=argparse.ArgumentDefaultsHelpFormatter)
 parser.add_argument("--model-filename", help="Path to the model file", default="ggml-model-Q8_0.gguf")
+parser.add_argument("--model-local", help="Path to the model file")
 parser.add_argument("--gpu", help="How many GPU layers to use", default=0, type=int)
 parser.add_argument("--ctx", help="How many context tokens to use", default=1024, type=int)
 parser.add_argument("--temp", help="temperature", default=0.8, type=float)
 args = parser.parse_args()
 
-model_path = download_model()
+model_path = None
+if args.model_local:
+    model_path = args.model_local
+else:
+    model_path = download_model()
+print(f"Loading model from {model_path}")
 
 # TODO n_gpu_layers doesn't seem to be working
 llm = Llama(model_path=model_path, n_gpu_layers=args.gpu, n_ctx=args.ctx)