QwQ-DeepSeek-R1-SkyT1-Flash-Light-32B

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

Merge Details

Merge Method

This model was merged using the sce merge method using Qwen/Qwen2.5-32B as a base.

Models Merged

The following models were included in the merge:

• NovaSky-AI/Sky-T1-32B-Flash
• deepseek-ai/DeepSeek-R1-Distill-Qwen-32B
• Qwen/QwQ-32B

Configuration

The following YAML configuration was used to produce this model:

merge_method: sce
base_model: Qwen/Qwen2.5-32B  # Pivot model (zero weight)
dtype: bfloat16
parameters:
  select_topk: 1.0
models:
  # Pivot model (explicitly zero-weighted)
  - model: Qwen/Qwen2.5-32B
    parameters:
      weight: 0.0  # Zero contribution to merged weights
      # sparsity: 0.0  # Optional: enforce sparsity if needed
  # Target models with assigned weights
  - model: Qwen/QwQ-32B
    parameters:
      weight: 0.75  # Dominant contributor
  - model: deepseek-ai/DeepSeek-R1-Distill-Qwen-32B
    parameters:
      weight: 0.10  # Minimal contribution
  - model: NovaSky-AI/Sky-T1-32B-Flash
    parameters:
      weight: 0.15   # Moderate contribution