You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
swarms/demos/autotemp.py

102 lines
3.9 KiB

import re
from concurrent.futures import ThreadPoolExecutor, as_completed
from swarms.models import OpenAIChat
class AutoTempAgent:
"""
AutoTemp is a tool for automatically selecting the best temperature setting for a given task.
Flow:
1. Generate outputs at a range of temperature settings.
2. Evaluate each output using the default temperature setting.
3. Select the best output based on the evaluation score.
4. Return the best output.
Args:
temperature (float, optional): The default temperature setting to use. Defaults to 0.5.
api_key (str, optional): Your OpenAI API key. Defaults to None.
alt_temps ([type], optional): A list of alternative temperature settings to try. Defaults to None.
auto_select (bool, optional): If True, the best temperature setting will be automatically selected. Defaults to True.
max_workers (int, optional): The maximum number of workers to use when generating outputs. Defaults to 6.
Returns:
[type]: [description]
Examples:
>>> from swarms.demos.autotemp import AutoTemp
>>> autotemp = AutoTemp()
>>> autotemp.run("Generate a 10,000 word blog on mental clarity and the benefits of meditation.", "0.4,0.6,0.8,1.0,1.2,1.4")
Best AutoTemp Output (Temp 0.4 | Score: 100.0):
Generate a 10,000 word blog on mental clarity and the benefits of meditation.
"""
def __init__(
self,
temperature: float = 0.5,
api_key: str = None,
alt_temps=None,
auto_select=True,
max_workers=6,
):
self.alt_temps = alt_temps if alt_temps else [0.4, 0.6, 0.8, 1.0, 1.2, 1.4]
self.auto_select = auto_select
self.max_workers = max_workers
self.temperature = temperature
self.alt_temps = alt_temps
self.llm = OpenAIChat(
openai_api_key=api_key,
temperature=temperature,
)
def evaluate_output(self, output: str):
"""Evaluate the output using the default temperature setting."""
eval_prompt = f"""
Evaluate the following output which was generated at a temperature setting of {self.temperature}.
Provide a precise score from 0.0 to 100.0, considering the criteria of relevance, clarity, utility, pride, and delight.
Output to evaluate:
---
{output}
---
"""
score_text = self.llm(prompt=eval_prompt)
score_match = re.search(r"\b\d+(\.\d)?\b", score_text)
return round(float(score_match.group()), 1) if score_match else 0.0
def run(self, task: str, temperature_string):
"""Run the AutoTemp agent."""
temperature_list = [
float(temp.strip()) for temp in temperature_string.split(",")
]
outputs = {}
scores = {}
with ThreadPoolExecutor(max_workers=self.max_workers) as executor:
future_to_temp = {
executor.submit(self.llm.generate, task, temp): temp
for temp in temperature_list
}
for future in as_completed(future_to_temp):
temp = future_to_temp[future]
output_text = future.result()
outputs[temp] = output_text
scores[temp] = self.evaluate_output(output_text, temp)
if not scores:
return "No valid outputs generated.", None
sorted_scores = sorted(scores.items(), key=lambda item: item[1], reverse=True)
best_temp, best_score = sorted_scores[0]
best_output = outputs[best_temp]
return (
f"Best AutoTemp Output (Temp {best_temp} | Score: {best_score}):\n{best_output}"
if self.auto_select
else "\n".join(
f"Temp {temp} | Score: {score}:\n{outputs[temp]}"
for temp, score in sorted_scores
)
)