| base_model: tokyotech-llm/Swallow-7b-instruct-v0.1 #tokyotech-llm/Swallow-7b-instruct-v0.1 | |
| gate_mode: hidden # one of "hidden", "cheap_embed", or "random" | |
| dtype: bfloat16 # output dtype (float32, float16, or bfloat16) | |
| ## (optional) | |
| experts_per_token: 2 | |
| experts: | |
| - source_model: tokyotech-llm/Swallow-7b-instruct-v0.1 #okyotech-llm/Swallow-7b-instruct-v0.1 | |
| positive_prompts: | |
| - "日本語でのマルチターンの質問回答の専門家です。" | |
| - source_model: elyza/ELYZA-japanese-Llama-2-7b-fast-instruct #https://huggingface.co/elyza/ELYZA-japanese-Llama-2-7b-fast-instruct | |
| positive_prompts: | |
| - "日本語でのQ&Aの専門家です。" | |
| - source_model: elyza/ELYZA-japanese-CodeLlama-7b-instruct #https://huggingface.co/elyza/ELYZA-japanese-CodeLlama-7b-instruct | |
| positive_prompts: | |
| - "コーディングの専門家です。" | |
| - source_model: meta-math/MetaMath-7B-V1.0 #https://huggingface.co/meta-math/MetaMath-7B-V1.0 | |
| positive_prompts: | |
| - "Math(数学)の専門家です。" | |