LLMcompe-Team-Watanabe/Qwen3-32B-merge-progressive

qwen32b_progressive_output

This is a merge of pre-trained language models created using mergekit.

Merge Details

Merge Method

This model was merged using the DARE TIES merge method using Qwen/Qwen3-32B as a base.

Models Merged

The following models were included in the merge:

• LLMcompe-Team-Watanabe/Qwen3-32B-sft-HIS-Chem-Engineering-45k-1ep-lr5e6-4096
• LLMcompe-Team-Watanabe/Qwen3-32B-textbookreasoning-UGPhysics-AoPsInstruct-openmathreasoning-sft
• LLMcompe-Team-Watanabe/Qwen3-32B-openmathreasoning-sft
• LLMcompe-Team-Watanabe/Qwen3-32B-textbookreasoning-UGPhysics-AoPsInstruct-sft

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: Qwen/Qwen3-32B
  - model: LLMcompe-Team-Watanabe/Qwen3-32B-openmathreasoning-sft
    parameters:
      density: 0.70  # 高密度で数学を強化
      weight: 0.30
  - model: LLMcompe-Team-Watanabe/Qwen3-32B-textbookreasoning-UGPhysics-AoPsInstruct-sft
    parameters:
      density: 0.60  # 中密度で物理・生物を追加
      weight: 0.25
  - model: LLMcompe-Team-Watanabe/Qwen3-32B-sft-HIS-Chem-Engineering-45k-1ep-lr5e6-4096
    parameters:
      density: 0.45  # 低密度で化学・工学を補完
      weight: 0.20
  - model: LLMcompe-Team-Watanabe/Qwen3-32B-textbookreasoning-UGPhysics-AoPsInstruct-openmathreasoning-sft
    parameters:
      density: 0.40
      weight: 0.10
  - model: Qwen/Qwen3-32B
    parameters:
      density: 0.53
      weight: 0.15
merge_method: dare_ties
base_model: Qwen/Qwen3-32B
parameters:
  int8_mask: true
  normalize: false
dtype: bfloat16