alpha-nerd-nomyo
434b6d4cca
Mixture of Experts: - prefix any ollama model with "moe-" on api/chat and the original user request gets passed to the selected model 3 times with temp=1 to get response variants. Each variant is then revisited and finally scored to find the best repsonse among them all and finally returned to the user. Runs longer, uses more tokens for expected better quality response.
49 lines
1.4 KiB
Python
49 lines
1.4 KiB
Python
from pydantic import BaseModel
|
|
|
|
class feedback(BaseModel):
|
|
query_id: int
|
|
content: str
|
|
|
|
def moe(query: str, query_id: int, response: str) -> str:
|
|
moe_prompt = f"""
|
|
User query: {query}
|
|
query_id: {query_id}
|
|
|
|
The following is an assistant response to the original user query. Analyse the response, then criticize the it by discussing both strengths and weaknesses. Do not add additional commentary.
|
|
|
|
<assistant_response>
|
|
{response}
|
|
</assistant_response>
|
|
|
|
Respond in the format:
|
|
original_response
|
|
---
|
|
Response Analysis:
|
|
your analysis
|
|
"""
|
|
return moe_prompt
|
|
|
|
def moe_select_candidate(query: str, candidates: list[str]) -> str:
|
|
if not candidates:
|
|
raise ValueError("No candidates supplied")
|
|
|
|
candidate_sections = ""
|
|
for i, cand in enumerate(candidates[:3], start=0):
|
|
candidate_sections += f"""
|
|
<candidate_{i}>
|
|
{cand.message.content}
|
|
</candidate_{i}>
|
|
"""
|
|
|
|
# Strict instruction: "Respond **only** with the final answer."
|
|
select_prompt = f"""
|
|
From the following responses for the user query: {query}
|
|
|
|
{candidate_sections}
|
|
|
|
Choose the best candidate and output the final answer in the language of the query.
|
|
**Do NOT** mention candidate numbers, strengths, weaknesses, or any other commentary.
|
|
Just give the final answer—nothing else.
|
|
"""
|
|
return select_prompt.strip()
|
|
|