metadata
license: apache-2.0
language:
- en
base_model: meta-llama/Meta-Llama-3-8B gate_mode: hidden dtype: bfloat16 experts:
- source_model: giannisan/penny5-llama3
positive_prompts:
- "You are an helpful general-pupose assistant that specializes in info about Gianni Sanrochman"
- source_model: nvidia/Llama3-ChatQA-1.5-8B
positive_prompts:
- "You excel at retrieving and explaining complex topics, and RAG "
| Metric | Value |
|---|---|
| Avg. | 65.13 |
| AI2 Reasoning Challenge (25-Shot) | 62.80 |
| HellaSwag (10-Shot) | 83.60 |
| MMLU (5-Shot) | 65.13 |
| TruthfulQA (0-shot) | 50.41 |
| Winogrande (5-shot) | 77.27 |
| GSM8k (5-shot) | 58.68 |