kajuma/QwQ-32B-Kumo

QwQ-32B Kumo

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

Configuration

The following YAML configuration was used to produce this model:

merge_method: slerp
base_model: Qwen/QwQ-32B
models:
  - model: Qwen/QwQ-32B
  - model: NovaSky-AI/Sky-T1-32B-Flash
parameters:
  t: 0.4
dtype: bfloat16
name: merge_model_1
---
merge_method: breadcrumbs_ties
base_model: Qwen/QwQ-32B
tokenizer_source: Qwen/QwQ-32B
name: merge_model_2
models:
  - model: Qwen/QwQ-32B
    parameters:
      weight: 1.0
  - model: FuseAI/FuseO1-DeepSeekR1-Qwen2.5-Coder-32B-Preview
    parameters:
      weight: 0.75
dtype: bfloat16
---
merge_method: task_arithmetic
base_model: Qwen/Qwen2.5-32B
tokenizer_source: Qwen/QwQ-32B
name: merge_model_3
models:
  - model: rinna/deepseek-r1-distill-qwen2.5-bakeneko-32b
    parameters:
      weight: 1.0
  - model: cyberagent/DeepSeek-R1-Distill-Qwen-32B-Japanese
    parameters:
      weight: 0.9
tokenizer_source: base
dtype: bfloat16
---
merge_method: slerp
base_model: Qwen/QwQ-32B
models:
  - model: Qwen/QwQ-32B
  - model: TeamDelta/ABEJA-Qwen2.5-32B-base-jp-v0.1
parameters:
  t: 0.5
tokenizer_source: base
dtype: bfloat16
name: merge_model_4
---
merge_method: model_stock

base_model: Qwen/QwQ-32B

models:
  - model: Qwen/QwQ-32B
  - model: merge_model_1
  - model: merge_model_2
  - model: merge_model_3
  - model: merge_model_4
  
dtype: bfloat16

pad_to_multiple_of: 512
tokenizer_source: base

name: QwQ-32B-Kumo