Add eos_token_id to generation config

If the model is used without `pipeline`, the model shows unexpected behaviour and generates extra unneeded tokens. The `eos_token_id` has to be passed explicitly to fix it since the generation config does not have it.

generation_config.json

@@ -1,5 +1,6 @@
 {
   "_from_model_config": true,
   "transformers_version": "4.28.1",
+  "eos_token_id": 0,
   "use_cache": false
 }