CalvinBackup/AutoTemp
1
0
Fork 0
mirror of https://github.com/elder-plinius/AutoTemp.git synced 2026-09-24 10:30:46 +02:00
Code Issues Packages Projects Releases Wiki Activity

Update autotemp.py

Browse Source
This commit is contained in:
pliny
2023-11-07 19:49:14 -08:00
committed by GitHub
parent e8fbf22ca0
commit 4ddef19df4
1 changed files with 33 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files
autotemp.py
+33 -23
View File
@@ -21,7 +21,7 @@ class AutoTemp:
self.max_workers = max_workers self.max_workers = max_workers
self.model_version = model_version self.model_version = model_version
def generate_with_openai(self, prompt, temperature, retries=3): def generate_with_openai(self, prompt, temperature, top_p, retries=3):
while retries > 0: while retries > 0:
try: try:
response = openai.ChatCompletion.create( response = openai.ChatCompletion.create(
@@ -30,14 +30,15 @@ class AutoTemp:
{"role": "system", "content": "You are a helpful assistant."}, {"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": prompt} {"role": "user", "content": prompt}
], ],
temperature=temperature temperature=temperature,
top_p=top_p
) )
message = response['choices'][0]['message']['content'] message = response['choices'][0]['message']['content']
return message.strip() return message.strip()
except Exception as e: except Exception as e:
retries -= 1 retries -= 1
if retries <= 0: if retries <= 0:
return f"Error generating text at temperature {temperature}: {e}" return f"Error generating text at temperature {temperature} and top-p {top_p}: {e}"
def evaluate_output(self, output, temperature): def evaluate_output(self, output, temperature):
eval_prompt = f""" eval_prompt = f"""
@@ -56,21 +57,20 @@ class AutoTemp:
{output} {output}
--- ---
""" """
score_text = self.generate_with_openai(eval_prompt, 0.5) # Use a neutral temperature for evaluation to get consistent results score_text = self.generate_with_openai(eval_prompt, 0.5, top_p) # Use a neutral temperature for evaluation
score_match = re.search(r'\b\d+(\.\d)?\b', score_text) score_match = re.search(r'\b\d+(\.\d)?\b', score_text)
if score_match: if score_match:
return round(float(score_match.group()), 1) # Round the score to one decimal place return round(float(score_match.group()), 1) # Round the score to one decimal place
else: else:
return 0.0 # Unable to parse score, default to 0.0 return 0.0 # Unable to parse score, default to 0.0
def run(self, prompt, temperature_list=None): def run(self, prompt, temperature_string, top_p):
if temperature_list is not None: temperature_list = [float(temp.strip()) for temp in temperature_string.split(',')]
self.alt_temps = temperature_list
outputs = {} outputs = {}
scores = {} scores = {}
with ThreadPoolExecutor(max_workers=self.max_workers) as executor: with ThreadPoolExecutor(max_workers=self.max_workers) as executor:
future_to_temp = { future_to_temp = {
executor.submit(self.generate_with_openai, prompt, temp): temp for temp in self.alt_temps executor.submit(self.generate_with_openai, prompt, temp, top_p): temp for temp in temperature_list
} }
for future in as_completed(future_to_temp): for future in as_completed(future_to_temp):
temp = future_to_temp[future] temp = future_to_temp[future]
@@ -89,33 +89,43 @@ class AutoTemp:
# If auto_select is enabled, return only the best result # If auto_select is enabled, return only the best result
if self.auto_select: if self.auto_select:
best_temp, best_output, best_score = sorted_outputs[0] best_temp, best_output, best_score = sorted_outputs[0]
return f"Best AutoTemp Output (Temp {best_temp} | Score: {best_score}):\n{best_output}" return f"Best AutoTemp Output (Temp {best_temp} | Top-p {top_p} | Score: {best_score}):\n{best_output}"
else: else:
return "\n".join(f"Temp {temp} | Score: {score}:\n{text}" for temp, text, score in sorted_outputs) return "\n".join(f"Temp {temp} | Top-p {top_p} | Score: {score}:\n{text}" for temp, text, score in sorted_outputs)
# Gradio app logic # Gradio app logic
def run_autotemp(prompt, temperature_string, auto_select): def run_autotemp(prompt, temperature_string, top_p, auto_select):
temperature_list = [float(temp.strip()) for temp in temperature_string.split(',')]
agent = AutoTemp(auto_select=auto_select) agent = AutoTemp(auto_select=auto_select)
output = agent.run(prompt, temperature_list=temperature_list) output = agent.run(prompt, temperature_string, top_p=float(top_p))
return output return output
# Gradio interface setup # Gradio interface setup
def main(): def main():
iface = gr.Interface( iface = gr.Interface(
fn=run_autotemp, fn=run_autotemp,
inputs=["text", "text", "checkbox"], inputs=[
"text",
"text",
gr.Slider(minimum=0.0, maximum=1.0, step=0.1, default=1.0, label="Top-p value"),
"checkbox"
],
outputs="text", outputs="text",
title="AutoTemp: Improved LLM Completions through Temperature Tuning", title="AutoTemp: Enhanced LLM Responses with Temperature and Top-p Tuning",
description="Enter different temperatures separated by commas (e.g., 0.4, 0.6, 0.8, 1.0, 1.2). Toggle 'Auto Select' to either see just the best output or all evaluated outputs.", description="""AutoTemp generates responses at different temperatures, evaluates them, and ranks them based on quality.
Enter temperatures separated by commas for evaluation.
Adjust 'Top-p' to control output diversity: lower for precision, higher for creativity.
Toggle 'Auto Select' to either see the top-rated output or all evaluated outputs.""",
article="""**What's Top-p?** 'Top-p' controls the diversity of AI responses: a low 'top-p' makes output more focused and predictable,
while a high 'top-p' encourages variety and surprise. Pair with temperature to fine-tune AI creativity:
higher temperatures with high 'top-p' for bold ideas, or lower temperatures with low 'top-p' for precise answers.""",
examples=[ examples=[
["Write a short story about AGI learning to love", "0.5, 0.7, 0.9, 1.1", False], ["Write a short story about AGI learning to love", "0.5, 0.7, 0.9, 1.1", 1.0, False],
["Create a dialogue between a chef and an alien creating an innovative new recipe", "0.3, 0.6, 0.9, 1.2", True], ["Create a dialogue between a chef and an alien creating an innovative new recipe", "0.3, 0.6, 0.9, 1.2", 0.9, True],
["Explain quantum computing to a 5-year-old", "0.4, 0.8, 1.2, 1.5", False], ["Explain quantum computing to a 5-year-old", "0.4, 0.8, 1.2, 1.5", 0.8, False],
["Draft an email to a hotel asking for a special arrangement for a marriage proposal", "0.4, 0.7, 1.0, 1.3", True], ["Draft an email to a hotel asking for a special arrangement for a marriage proposal", "0.4, 0.7, 1.0, 1.3", 0.7, True],
["Describe a futuristic city powered by renewable energy", "0.5, 0.75, 1.0, 1.25", False], ["Describe a futuristic city powered by renewable energy", "0.5, 0.75, 1.0, 1.25", 0.6, False],
["Generate a poem about the ocean's depths in the style of Edgar Allan Poe", "0.6, 0.8, 1.0, 1.2", True], ["Generate a poem about the ocean's depths in the style of Edgar Allan Poe", "0.6, 0.8, 1.0, 1.2", 0.5, True],
["What are some innovative startup ideas for improving urban transportation?", "0.45, 0.65, 0.85, 1.05", False] ["What are some innovative startup ideas for improving urban transportation?", "0.45, 0.65, 0.85, 1.05", 0.4, False]
] ]
) )
iface.launch() iface.launch()