Contextual-Obedient-MoE-3x8B-Llama3-RAG / generation_config.json

Llama 3 Context-Obedient Models in a 3 x MoE configuration. The models experts are split into Understanding and summarizing input, following the format provided, and outputting only context relevant answers as the three experts involved

2ed3e35 verified 30 days ago

raw history blame contribute delete

No virus

127 Bytes

	{
	"_from_model_config": true,
	"bos_token_id": 128000,
	"eos_token_id": 128003,
	"transformers_version": "4.40.2"
	}