Qwriter2

24 downloads 0 likes on Hugging Face updated

Model card

Qwriter

Cursed merge attempt number 3

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: THU-KEG/LongWriter-Zero-32B
    parameters:
      weight: 0.5
  - model: Qwen/QwQ-32B
    parameters:
      weight: 0.5
merge_method: della
base_model: Qwen/Qwen2.5-32B
dtype: bfloat16
tokenizer:
  source: union
  tokens:
    # Use embedding from a specific model
    <|im_start|>:
      source: "Qwen/QwQ-32B"
    <|im_end|>:
      source: "Qwen/QwQ-32B"