Llama-3-8B-Instruct-MoE-3 / mergekit_moe_config.yml
VictorDCh's picture
Upload folder using huggingface_hub
6a23b7d verified
raw
history blame
No virus
1.37 kB
base_model: meta-llama/Meta-Llama-3-8B-Instruct
gate_mode: random
dtype: bfloat16
experts_per_token: 2
experts:
- source_model: meta-llama/Meta-Llama-3-8B-Instruct
positive_prompts:
- "What are the different"
- "what are the distinct"
- "Give me the unique"
- source_model: meta-llama/Meta-Llama-3-8B-Instruct
positive_prompts:
- "When"
- "when"
- "Where"
- "where"
- "Which"
- "which"
- "Who"
- "who"
- "What"
- "what"
- "Whom"
- "whom"
- "Whose"
- "whose"
- source_model: meta-llama/Meta-Llama-3-8B-Instruct
positive_prompts:
- "Larger"
- "larger"
- "Smaller"
- "smaller"
- "Bigger"
- "bigger"
- "Smallest"
- "smallest"
- "Largest"
- "largest"
- "Biggest"
- "biggest"
- "Most"
- "most"
- "Least"
- "least"
- "More"
- "more"
- "Less"
- "less"
- "Number"
- "number"
- "Numbers"
- "numbers"
- "Quantity"
- "quantity"
- "At least"
- "at least"
- "At most"
- "at most"
- "Greater"
- "greater"
- "Fewer"
- "fewer"
- "Than"
- "than"
- "Equal"
- "equal"
- "Same"
- "same"
- "Equal to"
- "equal to"
- source_model: meta-llama/Meta-Llama-3-8B-Instruct
positive_prompts:
- "that also"
- "who have the same"