PREWL

Prompt Engineering Wrapper for LLM (PREWL)：一个库，用于通过 NLU 的快速工程快速构建基于 LLM 的应用程序原型。

用法

import prewl, json

# Load configuration for backend (e.g., GPT-3 credentials)
prewl.configure("config.json")

# Load the example prompts
examples =  prewl.load_promps("prompts.json")

PATTERN = """
Text: {text}
Sentiment: {sentiment}
"""

# Prompts objects
prompts = prewl.load_prompts(PATTERN, examples, output='sentiment')

# Build the backend-driven model that will be used
model = prewl.train(prompts) # Model object


# Use the model to build a prompt for the LLM, fetch the completion, and parse it
new_input = "This movie was off the hook!"
resp = model.infer(new_input)


print("\n New input: ", new_input)
print("Prediction: ", resp)
print()

更多示例可以在examples/目录中找到。

贡献

快来了...

要求

设置虚拟环境

python -m venv .env
source .env/bin/activate

安装割炬

pip install torch --extra-index-url https://download.pytorch.org/whl/cu113

引用这项工作

快来了...