tree-prompt's Introduction

🌳 Tree Prompting 🌳

Tree Prompting: Efficient Task Adaptation without Fine-Tuning, code for the Tree-prompt paper.

Tree Prompting uses training examples to learn a tree of prompts to make a classification, yielding higher accuracy and better efficiency that baseline ensembles.

Quickstart

Installation: pip install treeprompt (or clone this repo and pip install -e .)

from treeprompt.treeprompt import TreePromptClassifier
import datasets
import numpy as np
from sklearn.tree import plot_tree
import matplotlib.pyplot as plt

# set up data
rng = np.random.default_rng(seed=42)
dset_train = datasets.load_dataset('rotten_tomatoes')['train']
dset_train = dset_train.select(rng.choice(
    len(dset_train), size=100, replace=False))
dset_val = datasets.load_dataset('rotten_tomatoes')['validation']
dset_val = dset_val.select(rng.choice(
    len(dset_val), size=100, replace=False))

# set up arguments
prompts = [
    "This movie is",
    " Positive or Negative? The movie was",
    " The sentiment of the movie was",
    " The plot of the movie was really",
    " The acting in the movie was",
]
verbalizer = {0: " Negative.", 1: " Positive."}
checkpoint = "gpt2"

# fit model
m = TreePromptClassifier(
    checkpoint=checkpoint,
    prompts=prompts,
    verbalizer=verbalizer,
    cache_prompt_features_dir=None,  # 'cache_prompt_features_dir/gp2',
)
m.fit(dset_train["text"], dset_train["label"])


# compute accuracy
preds = m.predict(dset_val['text'])
print('\nTree-Prompt acc (val) ->',
      np.mean(preds == dset_val['label']))  # -> 0.7

# compare to accuracy for individual prompts
for i, prompt in enumerate(prompts):
    print(i, prompt, '->', m.prompt_accs_[i])  # -> 0.65, 0.5, 0.5, 0.56, 0.51

# visualize decision tree
plot_tree(
    m.clf_,
    fontsize=10,
    feature_names=m.feature_names_,
    class_names=list(verbalizer.values()),
    filled=True,
)
plt.show()

Reference:

@misc{morris2023tree,
    title={Tree Prompting: Efficient Task Adaptation without Fine-Tuning},
    author={John X. Morris and Chandan Singh and Alexander M. Rush and Jianfeng Gao and Yuntian Deng},
    year={2023},
    eprint={2310.14034},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}