Qwen2-Math-2B-Instruct / mergekit_config.yml
chargoddard's picture
Upload folder using huggingface_hub
9500c57 verified
raw
history blame contribute delete
426 Bytes
slices:
- sources:
- layer_range: [0, 12]
model: Qwen/Qwen2-Math-1.5B-Instruct
- sources:
- layer_range: [6, 18]
model: Qwen/Qwen2-Math-1.5B-Instruct
parameters:
scale:
- filter: o_proj
value: 0.0
- filter: down_proj
value: 0.0
- value: 1.0
- sources:
- layer_range: [12, 24]
model: Qwen/Qwen2-Math-1.5B-Instruct
merge_method: passthrough
dtype: bfloat16