Llama-3-DARE-8B / mergekit_config.yml
mlabonne's picture
Upload folder using huggingface_hub
d60cf08 verified
raw
history blame
303 Bytes
models:
- model: meta-llama/Meta-Llama-3-8B
# No parameters necessary for base model
- model: meta-llama/Meta-Llama-3-8B-Instruct
parameters:
density: 0.53
weight: 0.6
merge_method: dare_ties
base_model: meta-llama/Meta-Llama-3-8B
parameters:
int8_mask: true
dtype: bfloat16
random_seed: 0