Open Source Implementation of Evol-Instruct-Code as described in the WizardCoder Paper. Code for the intruction generation can be found on Github as Evol-Teacher.

A decontaminated version of evol-codealpaca-v1. Decontamination is done in the same way as StarCoder (bigcode decontamination process).

Evol-Instruct-Python-26k

Filtered version of the nickrosh/Evol-Instruct-Code-80k-v1 dataset that only keeps Python code (26,588 samples). You can find a smaller version of it here mlabonne/Evol-Instruct-Python-1k. Here is the distribution of the number of tokens in each row (instruction + output) using Llama's tokenizer:

The dataset is used in the research related to MultilingualSIFT.

Evol-Instruct-Code-80k is a dataset for evaluating the performance of code generation models.

The dataset used in the paper for in-context learning task

The dataset is created by (1) translating English questions of Evol-instruct-70k into Chinese and (2) requesting GPT4 to generate Chinese responses. For more details, please refer to:

Repository: https://github.com/FreedomIntelligence/AceGPT https://github.com/FreedomIntelligence/LLMZoo

Paper: AceGPT, Localizing Large Language Models in Arabic Phoenix: Democratizing ChatGPT across Languages

  BibTeX entry and citation info

@article{huang2023acegpt, title={AceGPT, Localizing… See the full description on the dataset page: https://huggingface.co/datasets/FreedomIntelligence/Evol-Instruct-Chinese-GPT4.

Dataset

This dataset was created by CoverLover

Released under MIT

Contents

artificial-citizen/Evol-Instruct-Code dataset hosted on Hugging Face and contributed by the HF Datasets community

Alignment-Lab-AI/clusteredleaves-evol-instruct dataset hosted on Hugging Face and contributed by the HF Datasets community

Wizard-LM-Chinese is a data set that translates instructions on MSRA's Wizard-LM data set and then calls GPT to obtain the answer. Wizard-LM contains many instructions that are more difficult than Alpaca. In Chinese translation problems, a small amount of instruction injection may cause translation failure. Chinese answers are obtained by asking questions based on Chinese questions.

plaguss/distilabel-sample-evol-instruct dataset hosted on Hugging Face and contributed by the HF Datasets community

Wizardcoder: Empowering code large language models with evol-instruct

AI Research Instructions and Outputs

Driving Innovation in Machine Learning and AI Exploration

By Huggingface Hub [source]

About this dataset

This dataset contains 80,000 unique pairs of instructions and outputs to be used for Machine Learning and AI research. Instructions such as 'run', 'walk', 'jump', and 'dance' have outputs that represent the results of executing each instruction. It provides a groundbreaking collection of knowledge that can be leveraged in ways such as training AI agents, building intelligent natural language applications, exploring autonomous navigation possibilities, developing dialogues between bots and humans, replicating robotic tasks and research into sophisticated AI models able to understand instructions in various domains like engineering, medicine, finance or law. This dataset has the potential to revolutionize how we approach Artificial Intelligence by pushing boundaries when it comes to data-driven machine learning strategies. With its powerful combination of detailed information from multiple angles – language comprehension from verbal commands alongside increased contextual understanding – we can pave the way for more comprehensive applications of AI technology with exponentially enhanced accuracy when compared to existing methods

More Datasets

For more datasets, click here.

Featured Notebooks

• 🚨 Your notebook can be here! 🚨!

How to use the dataset

This dataset contains 80,000 pairs of instructions and outputs for Machine Learning and AI research. This data can be used to teach a variety of AI agents, as well as for tasks like autonomous navigation, dialogue, language modelling, natural language processing (NLP), robotics applications and more. The following guide outlines the steps you'll need to take in order to get the most out of this incredible resource.

• Download the dataset from Kaggle - Once downloaded you'll have access to two files: instruction.csv & output.csv.
• Examine the data - Take some time familiarizing yourself with the dataset- The columns will contain instructions/verbs such as 'run', walk', 'jump' etc., along with accompanying output results that have been generated from executing those instructions.
• Transform the data - Utilize feature engineering techniques appropriate for your project/proposed application in order to transform or extract relevant features from this dataset that can be utilized downstream by either supervised algorithms such as neural networks or unsupervised methods such as clustering algorithms.
  4 Train & Test models – Develop predictive models using either supervised or unsupervised techniques according; adjust hyperparameters until desired results are obtained; split into a training set (80%) and validation set (20%) first before running on full dataset so that model performance can be properly assessed against validation/test datasets; additional notes here about repeatability vs randomization etc… 5 Deploy Models – Deploy model onto real world scenarios/environments where appropriate .e.. an autonomous car relying on natural language inputs when driving through town; a domestic robot understanding sentences given by its user etc…

Research Ideas

• Training virtual assistants with specific domain knowledge (e.g. medical, finance, etc).
• Develop autonomous navigation systems that respond to verbal instructions given by a user in natural language format.
• Creating dialogue agents that can answer questions based on a pre-defined set of rules pertaining to the instructions given by the user

Acknowledgements

If you use this dataset in your research, please credit the original authors. Data Source

License

License: CC0 1.0 Universal (CC0 1.0) - Public Domain Dedication No Copyright - You can copy, modify, distribute and perform the work, even for commercial purposes, all without...

Evolved codealpaca

Updates:

2023/08/26 - Filtered results now only contain pure english instruction and removed any mentioned of trained by OAI response

Median sequence length : 471 We employed a methodology similar to that of WizardCoder, with the exception that ours is open-source. We used the gpt-4-0314 and gpt-4-0613 models to augment and answer each response, with the bulk of generation handled by gpt-4-0314. The aim of this dataset is twofold: firstly, to facilitate the… See the full description on the dataset page: https://huggingface.co/datasets/theblackcat102/evol-codealpaca-v1.

Locutusque/Collective-Evol-Instruct-v0.1 dataset hosted on Hugging Face and contributed by the HF Datasets community

zichao22/evol-instruct dataset hosted on Hugging Face and contributed by the HF Datasets community

alvarobartt/evol-instruct dataset hosted on Hugging Face and contributed by the HF Datasets community

banksy235/Codefuse-Evol-Instruct-Clean dataset hosted on Hugging Face and contributed by the HF Datasets community

Data from a study investigating the influence of Nature of Science Instruction on evolution acceptance.