Create config.json

config.json

@@ -0,0 +1,7 @@
+{
+  "model_type": "llama",
+  "quantization_config": {
+    "load_in_4bit": true,
+    "bnb_4bit_compute_dtype": "float16"
+  }
+}