Unsloth-Approach / mergekit_moe_config.yml
Treza12's picture
Upload folder using huggingface_hub
61f4a08 verified
raw
history blame contribute delete
725 Bytes
base_model: BioMistral/BioMistral-7B
gate_mode: approach # one of "hidden", "cheap_embed", or "random"
dtype: bfloat16 # output dtype (float32, float16, or bfloat16)
## (optional)
# experts_per_token: 2
experts:
- source_model: Treza12/BioMistral-Unsloth-Full-Class-0
positive_prompts:
- "This is a prompt that is demonstrative of what expert_model_1 excels at Screening"
- source_model: Treza12/BioMistral-Unsloth-Full-Class-1
positive_prompts:
- "This is a prompt that is demonstrative of what expert_model_1 excels at Diagnosis"
- source_model: Treza12/BioMistral-Unsloth-Full-Class-2
positive_prompts:
- "This is a prompt that is demonstrative of what expert_model_1 excels at Staging"