tree-prompt

Name: tree-prompt
Author: csinva

23installs/mo · pypi42stars4forks

How to install tree-prompt?

csinva/tree-prompt

$git clone https://github.com/csinva/tree-prompt

Installs into the current project.

Summary

Reference: ``r @misc{ch2022augmenting, title={Tree Prompting: Efficient Task Adaptation without Fine-Tuning}, year={2023}, archivePrefix={arXiv}, primaryClass={cs.AI} }

Files · 1

View on GitHub

README.md

1	<h1 align="center"> Tree Prompting </h1>
2	<p align="center"> Tree Prompting: Efficient Task Adaptation without Fine-Tuning, code for the <a href="">Tree-prompt paper</a>.
3	</p>
4
5	<p align="center">
6	<img src="https://img.shields.io/badge/license-mit-blue.svg">
7	<img src="https://img.shields.io/badge/python-3.6+-blue">
8	<img src="https://img.shields.io/pypi/v/imodelsx?color=green">
9	</p>
10
11	<p align="center"> Tree Prompting uses training examples to learn a tree of prompts to make a classificationg, yielding higher accuracy and better efficiency that baseline ensembles.
12	</p>
13
14	### Quickstart
15
16	For a simple scikit-learn interface to use Tree-Prompt, use the <a href="https://github.com/csinva/imodelsX">imodelsX package</a>. Installation: `pip install imodelsx`
17
18	```python
19	from imodelsx import TreePromptClassifier
20	import datasets
21	import numpy as np
22	from sklearn.tree import plot_tree
23	import matplotlib.pyplot as plt
24
25	# set up data
26	rng = np.random.default_rng(seed=42)
27	dset_train = datasets.load_dataset('rotten_tomatoes')['train']
28	dset_train = dset_train.select(rng.choice(
29	len(dset_train), size=100, replace=False))
30	dset_val = datasets.load_dataset('rotten_tomatoes')['validation']
31	dset_val = dset_val.select(rng.choice(
32	len(dset_val), size=100, replace=False))
33
34	# set up arguments
35	prompts = [
36	"This movie is",
37	" Positive or Negative? The movie was",
38	" The sentiment of the movie was",
39	" The plot of the movie was really",
40	" The acting in the movie was",
41	]
42	verbalizer = {0: " Negative.", 1: " Positive."}
43	checkpoint = "gpt2"
44
45	# fit model
46	m = TreePromptClassifier(
47	checkpoint=checkpoint,
48	prompts=prompts,
49	verbalizer=verbalizer,
50	cache_prompt_features_dir=None, # 'cache_prompt_features_dir/gp2',
51	)
52	m.fit(dset_train["text"], dset_train["label"])
53
54
55	# compute accuracy
56	preds = m.predict(dset_val['text'])
57	print('\nTree-Prompt acc (val) ->',
58	np.mean(preds == dset_val['label'])) # -> 0.7
59
60	# compare to accuracy for individual prompts
61	for i, prompt in enumerate(prompts):
62	print(i, prompt, '->', m.prompt_accs_[i]) # -> 0.65, 0.5, 0.5, 0.56, 0.51
63
64	# visualize decision tree
65	plot_tree(
66	m.clf_,
67	fontsize=10,
68	feature_names=m.feature_names_,
69	class_names=list(verbalizer.values()),
70	filled=True,
71	)
72	plt.show()
73	```
74
75	Reference:
76	```r
77	@misc{ch2022augmenting,
78	title={Tree Prompting: Efficient Task Adaptation without Fine-Tuning},
79	year={2023},
80	archivePrefix={arXiv},
81	primaryClass={cs.AI}
82	}
83	```
84
85
86	# Reproducing experiments
87
88	### Organization
89	- `tprompt`: contains main code for modeling (e.g. model architecture)
90	- `experiments`: code for runnning experiments (e.g. loading data, training models, evaluating models)
91	- `scripts`: scripts for running experiments (e.g. python scripts that launch jobs in `experiments` folder with different hyperparams)
92	- `notebooks`: jupyter notebooks for analyzing results and making figures
93	- `tests`: unit tests
94
95	### Setup
96	- clone and run `pip install -e .`, resulting in a package named `tprompt` that can be imported
97	- see `setup.py` for dependencies, not all are required
98	- example run: run `python scripts/01_train_basic_models.py` (which calls `experiments/01_train_model.py` then view the results in `notebooks/01_model_results.ipynb`

People who liked this also viewed

6 picks

Type

model-alignmentModel Alignment is a python library from the PAIR team that enable users to create model prompts through user feedback instead of manual prompt writing and editing. The technique makes use of constitutional principles to align prompts to users' desired values.PromptsOct 202521239
prompt-decoratorsA standardized framework for enhancing how LLMs process and respond to prompts through composable decorators, featuring an official open standard specification and Python reference implementation. Claude Code plugin and MCP server integration.PromptsJul 202617742
claudifyAI-powered Spotify playlist generator using Claude API. Create personalized playlists from text prompts or transform your liked songs into curated collections. Seamlessly blends AI creativity with music discovery.PromptsJul 20241729
system-prompts-and-models-of-ai-toolsFULL Augment Code, Claude Code, Cluely, CodeBuddy, Comet, Cursor, Devin AI, Junie, Kiro, Leap.new, Lovable, Manus, NotionAI, Orchids.app, Perplexity, Poke, Qoder, Replit, Same.dev, Trae, Traycer AI, VSCode Agent, Warp.dev, Windsurf, Xcode, Z.ai Code, Dia & v0. (And other Open Sourced) System Prompts, Internal Tools & AI ModelsPromptsJul 2026—142k
CL4R1T4SLEAKED SYSTEM PROMPTS FOR CHATGPT, CLAUDE, GEMINI, GROK, PERPLEXITY, CURSOR, LOVABLE, REPLIT, AND MORE! - AI SYSTEMS TRANSPARENCY FOR ALL! 👐PromptsJul 2026—46k
ChatGPT-Shortcut🚀💪Maximize your efficiency and productivity. The ultimate hub to manage, customize, and share prompts. (English/中文/Español/العربية). 让生产力加倍的 AI 快捷指令。更高效地管理提示词，在分享社区中发现适用于不同场景的灵感。PromptsJul 2026—8.6k

tree-prompt

23installs/mo · pypi42stars4forks

View on GitHub

How to install tree-prompt?

csinva/tree-prompt

$git clone https://github.com/csinva/tree-prompt

Installs into the current project.

Summary

Reference: ``r @misc{ch2022augmenting, title={Tree Prompting: Efficient Task Adaptation without Fine-Tuning}, year={2023}, archivePrefix={arXiv}, primaryClass={cs.AI} }

Files · 1

View on GitHub

README.md

1	<h1 align="center"> Tree Prompting </h1>
2	<p align="center"> Tree Prompting: Efficient Task Adaptation without Fine-Tuning, code for the <a href="">Tree-prompt paper</a>.
3	</p>
4
5	<p align="center">
6	<img src="https://img.shields.io/badge/license-mit-blue.svg">
7	<img src="https://img.shields.io/badge/python-3.6+-blue">
8	<img src="https://img.shields.io/pypi/v/imodelsx?color=green">
9	</p>
10
11	<p align="center"> Tree Prompting uses training examples to learn a tree of prompts to make a classificationg, yielding higher accuracy and better efficiency that baseline ensembles.
12	</p>
13
14	### Quickstart
15
16	For a simple scikit-learn interface to use Tree-Prompt, use the <a href="https://github.com/csinva/imodelsX">imodelsX package</a>. Installation: `pip install imodelsx`
17
18	```python
19	from imodelsx import TreePromptClassifier
20	import datasets
21	import numpy as np
22	from sklearn.tree import plot_tree
23	import matplotlib.pyplot as plt
24
25	# set up data
26	rng = np.random.default_rng(seed=42)
27	dset_train = datasets.load_dataset('rotten_tomatoes')['train']
28	dset_train = dset_train.select(rng.choice(
29	len(dset_train), size=100, replace=False))
30	dset_val = datasets.load_dataset('rotten_tomatoes')['validation']
31	dset_val = dset_val.select(rng.choice(
32	len(dset_val), size=100, replace=False))
33
34	# set up arguments
35	prompts = [
36	"This movie is",
37	" Positive or Negative? The movie was",
38	" The sentiment of the movie was",
39	" The plot of the movie was really",
40	" The acting in the movie was",
41	]
42	verbalizer = {0: " Negative.", 1: " Positive."}
43	checkpoint = "gpt2"
44
45	# fit model
46	m = TreePromptClassifier(
47	checkpoint=checkpoint,
48	prompts=prompts,
49	verbalizer=verbalizer,
50	cache_prompt_features_dir=None, # 'cache_prompt_features_dir/gp2',
51	)
52	m.fit(dset_train["text"], dset_train["label"])
53
54
55	# compute accuracy
56	preds = m.predict(dset_val['text'])
57	print('\nTree-Prompt acc (val) ->',
58	np.mean(preds == dset_val['label'])) # -> 0.7
59
60	# compare to accuracy for individual prompts
61	for i, prompt in enumerate(prompts):
62	print(i, prompt, '->', m.prompt_accs_[i]) # -> 0.65, 0.5, 0.5, 0.56, 0.51
63
64	# visualize decision tree
65	plot_tree(
66	m.clf_,
67	fontsize=10,
68	feature_names=m.feature_names_,
69	class_names=list(verbalizer.values()),
70	filled=True,
71	)
72	plt.show()
73	```
74
75	Reference:
76	```r
77	@misc{ch2022augmenting,
78	title={Tree Prompting: Efficient Task Adaptation without Fine-Tuning},
79	year={2023},
80	archivePrefix={arXiv},
81	primaryClass={cs.AI}
82	}
83	```
84
85
86	# Reproducing experiments
87
88	### Organization
89	- `tprompt`: contains main code for modeling (e.g. model architecture)
90	- `experiments`: code for runnning experiments (e.g. loading data, training models, evaluating models)
91	- `scripts`: scripts for running experiments (e.g. python scripts that launch jobs in `experiments` folder with different hyperparams)
92	- `notebooks`: jupyter notebooks for analyzing results and making figures
93	- `tests`: unit tests
94
95	### Setup
96	- clone and run `pip install -e .`, resulting in a package named `tprompt` that can be imported
97	- see `setup.py` for dependencies, not all are required
98	- example run: run `python scripts/01_train_basic_models.py` (which calls `experiments/01_train_model.py` then view the results in `notebooks/01_model_results.ipynb`

People who liked this also viewed

6 picks

Type