Views
No views yet
1# Task Arithmetic - Grid Search
2# Weights: 0.8, 1.0
3base_model: deepseek-ai/deepseek-coder-7b-base-v1.5
4models:
5 - model: deepseek-ai/deepseek-math-7b-instruct
6 parameters:
7 weight: 0.8
8 - model: deepseek-ai/deepseek-coder-7b-instruct-v1.5
9 parameters:
10 weight: 1.0
11merge_method: task_arithmetic
12parameters:
13 normalize: false
14 lambda: 1.0
15dtype: float16
16tokenizer:
17 source: union
18