Llama3-NextGen-9b / mergekit_config.yml
powermove72's picture
Upload folder using huggingface_hub
5696e02 verified
slices:
- sources:
- model: gradientai/Llama-3-8B-Instruct-Gradient-1048k
layer_range: [0, 16]
- sources:
- model: failspy/Llama-3-8B-Instruct-MopeyMule
layer_range: [8, 32]
merge_method: passthrough
tokenizer_source: union
dtype: float16