CodeLlemur-2B-Instruct / mergekit_config.yml
JoPmt's picture
Upload folder using huggingface_hub
0df50e4 verified
raw
history blame contribute delete
578 Bytes
dtype: bfloat16
merge_method: linear
slices:
- sources:
- layer_range: [0, 32] # Assuming the first half of the model is more general and can be reduced more
model: codellama/CodeLlama-7b-Instruct-hf
parameters:
weight: 0.25 # Reduce the weight of the first half to make room for the second half
- layer_range: [0, 32] # Assuming the second half of the model is more specialized and can be reduced less
model: codellama/CodeLlama-7b-Instruct-hf
parameters:
weight: 0.25 # Maintain the weight of the second half