metadata
license: llama3.1
Merged as below:
slices:
sources:
- model: athirdpath/Llama-3.1-Instruct_NSFW-pretrained_e1-plus_reddit
- layer_range: [0, 23]
sources:
- model: athirdpath/Llama-3.1-Techne-RP-8b-v1
- layer_range: [9, 31]
merge_method: passthrough
dtype: float16
tokenizer_source: athirdpath/Llama-3.1-Techne-RP-8b-v1
Then pretrained for 1 epoch on the Iambe dataset as an 11b model