Marmoka-es-Llama-3.1-8B-Ins…/mergekit_config.yml

models:
 - model: meta-llama/Llama-3.1-8B-Instruct
 - model: /data/abarrena/mn5/marmoka-es0-mn5ep3_v1.1
   parameters:
     density: 0.8
     weight: 0.2
 - model: /data/abarrena/mn5/marmoka-es1-mn5ep3_v1.1
   parameters:
     density: 0.8
     weight: 0.6
 - model: /data/abarrena/mn5/marmoka-es2-mn5ep3_v1.1
   parameters:
     density: 0.8
     weight: 0.2
merge_method: dare_ties
base_model: meta-llama/Llama-3.1-8B-Instruct
dtype: bfloat16
初始化项目，由ModelHub XC社区提供模型 Model: HiTZ/Marmoka-es-Llama-3.1-8B-Instruct Source: Original Platform 2026-10-22 13:58:16 +08:00			`models:`
			`- model: meta-llama/Llama-3.1-8B-Instruct`
			`- model: /data/abarrena/mn5/marmoka-es0-mn5ep3_v1.1`
			`parameters:`
			`density: 0.8`
			`weight: 0.2`
			`- model: /data/abarrena/mn5/marmoka-es1-mn5ep3_v1.1`
			`parameters:`
			`density: 0.8`
			`weight: 0.6`
			`- model: /data/abarrena/mn5/marmoka-es2-mn5ep3_v1.1`
			`parameters:`
			`density: 0.8`
			`weight: 0.2`
			`merge_method: dare_ties`
			`base_model: meta-llama/Llama-3.1-8B-Instruct`
			`dtype: bfloat16`