Files
qwen3-4b-merge-practice-merged/merge_recipe.yaml
ModelHub XC 59c5549db5 初始化项目,由ModelHub XC社区提供模型
Model: MoshiQAQ/qwen3-4b-merge-practice-merged
Source: Original Platform
2026-10-28 04:04:18 +08:00

18 lines
381 B
YAML

base_model:
id: Qwen/Qwen3-4B-Instruct-2507
revision: cdbee75f17c01a7cc42f958dc650907174af0554
method: linear
experts:
math: /workspace/ckpts/math/final
science: /workspace/ckpts/science/final
code: /workspace/ckpts/code/final
weights:
math: 1.0
science: 1.0
code: 1.0
density: 1.0
scale: 1.0
seed: 0
n_target_tensors: 252
mean_rel_delta_norm: 0.006694779499426543