Views
No views yet
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.deepseek-ai/deepseek-llm-7b-base1.0 each):1merge_method: multislerp
2base_model: deepseek-ai/deepseek-llm-7b-base
3tokenizer_source: base
4dtype: float32
5out_dtype: bfloat16
6parameters:
7 normalize_weights: true
8 eps: 1.0e-8
9models:
10 - model: deepseek-ai/deepseek-math-7b-instruct
11 parameters: {weight: 1.0}
12 - model: deepseek-ai/deepseek-coder-7b-instruct-v1.5
13 parameters: {weight: 1.0}