GeneticPromptLab

Name: GeneticPromptLab
Author: AmanPriyanshu

34stars4forks

How to install GeneticPromptLab?

amanpriyanshu/geneticpromptlab

$git clone https://github.com/AmanPriyanshu/GeneticPromptLab

Installs into the current project.

Summary

GeneticPromptLab is a Python library designed to harness the power of genetic algorithms for automated prompt engineering for large language models (LLMs). By iteratively applying selection, crossover, and mutation processes, GeneticPromptLab enhances the quality and diversity of

Files · 1

View on GitHub

README.md

1	# GeneticPromptLab
2	GeneticPromptLab uses genetic algorithms for automated prompt engineering (for LLMs), enhancing quality and diversity through iterative selection, crossover, and mutation, while efficiently exploring minimal yet diverse samples from the training set.
3
4	## Overview
5
6	GeneticPromptLab is a Python library designed to harness the power of genetic algorithms for automated prompt engineering for large language models (LLMs). By iteratively applying selection, crossover, and mutation processes, GeneticPromptLab enhances the quality and diversity of prompts, leading to improved performance in automated question-answering and classification tasks.
7
8	### Pip Install
9	`pip install genetic-prompt-lab`
10
11	This library specifically leverages Sentence Transformers for embedding generation and k-means clustering to sample minimal yet diverse data from training sets. This strategic sampling ensures efficient exploration and optimization of prompts over multiple generations.
12
13	```mermaid
14	graph TD;
15	A[Initialize OpenAI Client] --> B[Load Dataset Configuration]
16	B --> C[Setup QuestionsAnswersOptimizer]
17	C --> D[Run Genetic Algorithm]
18
19	subgraph Genetic Algorithm Process
20	D --> GA[Start Genetic Algorithm]
21	GA --> Init[Generate Initial Prompts]
22	Init --> FE1
23
24	subgraph Fitness Evaluation
25	FE1[Current Population]
26	FE1 --> FE2[Evaluate Individual Fitness]
27	FE2 --> FE3[Calculate Average Fitness]
28	end
29
30	FE1 --> STP[Select Top Prompts]
31	STP --> Cross[Crossover]
32	subgraph Crossover Mechanism
33	Cross --> Cross1[Select Parent Prompts]
34	Cross1 --> Cross2[Apply Crossover Logic]
35	Cross2 --> Cross3[Generate Offspring Prompts]
36	end
37
38	Cross3 --> Mut[Mutation]
39	subgraph Mutation Mechanism
40	Mut --> Mut1[Select Prompts for Mutation]
41	Mut1 --> Mut2[Apply Mutation Logic]
42	Mut2 --> Mut3[Generate Mutated Prompts]
43	end
44
45	Mut3 --> Check[Check Convergence]
46	Check -- No --> FE1
47	Check -- Yes --> Finish[End Genetic Algorithm]
48	end
49
50	Finish --> Out[Output]
51
52	style D fill:#f9f,stroke:#333,stroke-width:4px
53	style Genetic Algorithm Process fill:#ccf,stroke:#333,stroke-width:2px
54	style Fitness Evaluation fill:#fdd,stroke:#333,stroke-width:2px
55	style Crossover Mechanism fill:#dfd,stroke:#333,stroke-width:2px
56	style Mutation Mechanism fill:#ddf,stroke:#333,stroke-width:2px
57	```
58
59	## Features
60
61	- Genetic Algorithm Implementation: Complete genetic algorithm cycle including initialization, fitness evaluation, selection, crossover, and mutation for prompt engineering.
62	- Integration with Sentence Transformers: Utilizes Sentence Transformers for generating embeddings, enabling effective clustering and sampling.
63	- Diverse Sampling: Uses k-means clustering to select a representative subset of data, ensuring diverse genetic material for algorithm initialization and evolution.
64	- Dynamic Mutation and Crossover: Incorporates custom mutation and crossover operations tailored to prompt characteristics and task requirements.
65	- Multi-Dataset Support: Pre-configured to run experiments on the TREC and AG News datasets, demonstrating its adaptability to different types of text classification tasks.
66
67	## Installation
68
69	To install GeneticPromptLab, clone this repository and install the required packages:
70
71	```bash
72	git clone https://github.com/username/GeneticPromptLab.git
73	cd GeneticPromptLab
74	pip install -r requirements.txt
75	```
76
77	## Benchmarks
78
79	### Performance on TREC Dataset
80
81	The TREC dataset comprises 4,500 English questions categorized into various labels. Below is the performance improvement graph showing the evolution of prompt effectiveness over 10 generations.
82	#### Training Progression exploring only 10% of the training data
83	![Performance on TREC Dataset](https://raw.githubusercontent.com/AmanPriyanshu/GeneticPromptLab/main/images/

People who liked this also viewed

6 picks

Type

GeneticPromptLab

34stars4forks

View on GitHub

How to install GeneticPromptLab?

amanpriyanshu/geneticpromptlab

$git clone https://github.com/AmanPriyanshu/GeneticPromptLab

Installs into the current project.

Summary

Files · 1

View on GitHub

README.md

1	# GeneticPromptLab
2	GeneticPromptLab uses genetic algorithms for automated prompt engineering (for LLMs), enhancing quality and diversity through iterative selection, crossover, and mutation, while efficiently exploring minimal yet diverse samples from the training set.
3
4	## Overview
5
6	GeneticPromptLab is a Python library designed to harness the power of genetic algorithms for automated prompt engineering for large language models (LLMs). By iteratively applying selection, crossover, and mutation processes, GeneticPromptLab enhances the quality and diversity of prompts, leading to improved performance in automated question-answering and classification tasks.
7
8	### Pip Install
9	`pip install genetic-prompt-lab`
10
11	This library specifically leverages Sentence Transformers for embedding generation and k-means clustering to sample minimal yet diverse data from training sets. This strategic sampling ensures efficient exploration and optimization of prompts over multiple generations.
12
13	```mermaid
14	graph TD;
15	A[Initialize OpenAI Client] --> B[Load Dataset Configuration]
16	B --> C[Setup QuestionsAnswersOptimizer]
17	C --> D[Run Genetic Algorithm]
18
19	subgraph Genetic Algorithm Process
20	D --> GA[Start Genetic Algorithm]
21	GA --> Init[Generate Initial Prompts]
22	Init --> FE1
23
24	subgraph Fitness Evaluation
25	FE1[Current Population]
26	FE1 --> FE2[Evaluate Individual Fitness]
27	FE2 --> FE3[Calculate Average Fitness]
28	end
29
30	FE1 --> STP[Select Top Prompts]
31	STP --> Cross[Crossover]
32	subgraph Crossover Mechanism
33	Cross --> Cross1[Select Parent Prompts]
34	Cross1 --> Cross2[Apply Crossover Logic]
35	Cross2 --> Cross3[Generate Offspring Prompts]
36	end
37
38	Cross3 --> Mut[Mutation]
39	subgraph Mutation Mechanism
40	Mut --> Mut1[Select Prompts for Mutation]
41	Mut1 --> Mut2[Apply Mutation Logic]
42	Mut2 --> Mut3[Generate Mutated Prompts]
43	end
44
45	Mut3 --> Check[Check Convergence]
46	Check -- No --> FE1
47	Check -- Yes --> Finish[End Genetic Algorithm]
48	end
49
50	Finish --> Out[Output]
51
52	style D fill:#f9f,stroke:#333,stroke-width:4px
53	style Genetic Algorithm Process fill:#ccf,stroke:#333,stroke-width:2px
54	style Fitness Evaluation fill:#fdd,stroke:#333,stroke-width:2px
55	style Crossover Mechanism fill:#dfd,stroke:#333,stroke-width:2px
56	style Mutation Mechanism fill:#ddf,stroke:#333,stroke-width:2px
57	```
58
59	## Features
60
61	- Genetic Algorithm Implementation: Complete genetic algorithm cycle including initialization, fitness evaluation, selection, crossover, and mutation for prompt engineering.
62	- Integration with Sentence Transformers: Utilizes Sentence Transformers for generating embeddings, enabling effective clustering and sampling.
63	- Diverse Sampling: Uses k-means clustering to select a representative subset of data, ensuring diverse genetic material for algorithm initialization and evolution.
64	- Dynamic Mutation and Crossover: Incorporates custom mutation and crossover operations tailored to prompt characteristics and task requirements.
65	- Multi-Dataset Support: Pre-configured to run experiments on the TREC and AG News datasets, demonstrating its adaptability to different types of text classification tasks.
66
67	## Installation
68
69	To install GeneticPromptLab, clone this repository and install the required packages:
70
71	```bash
72	git clone https://github.com/username/GeneticPromptLab.git
73	cd GeneticPromptLab
74	pip install -r requirements.txt
75	```
76
77	## Benchmarks
78
79	### Performance on TREC Dataset
80
81	The TREC dataset comprises 4,500 English questions categorized into various labels. Below is the performance improvement graph showing the evolution of prompt effectiveness over 10 generations.
82	#### Training Progression exploring only 10% of the training data
83	![Performance on TREC Dataset](https://raw.githubusercontent.com/AmanPriyanshu/GeneticPromptLab/main/images/

People who liked this also viewed

6 picks

Type