Method
Multi-SLERP (multislerp) performs barycentric spherical interpolation on a hypersphere for more than two models: it projects the models into the tangent space at their weighted Euclidean mean, interpolates, and projects back. Here it is run in task-vector space — each source's delta from the shared base model is computed, the deltas are spherically averaged with equal weight (normalize_weights: true, eps: 1e-8), and the result is added back to the base. Merging was done with mergekit.
All variants share an identical vocabulary (102400); no tokenizer reconciliation was needed.
Sources
Base model (task-vector reference): deepseek-ai/deepseek-llm-7b-base
Merged variants (equal weight 1.0 each):
mergekit config
merge_method: multislerp
base_model: deepseek-ai/deepseek-llm-7b-base
tokenizer_source: base
dtype: float32
out_dtype: bfloat16
parameters:
normalize_weights: true
eps: 1.0e-8
models:
- model: deepseek-ai/deepseek-math-7b-instruct
parameters: {weight: 1.0}
- model: deepseek-ai/deepseek-coder-7b-instruct-v1.5
parameters: {weight: 1.0}