Skip to content

Latest commit

 

History

History

scripts

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 
 
 
 
 

Openvalidators wandb data collector

This folder contain a set of scripts to automate the data collection of the openvalidators project in wandb.

Requirements

Ensure that the requirements of the project are properly installed on your machine.

pip install -e ../setup.py # Install openvalidators from root folder

The current implementation requires you to be logged in wandb. You can do so by running the following command:

wandb login

You will also need to install make to profit from the predefined targets in the Makefile. To install make on Ubuntu, run the following command:

sudo apt install make

If you are on Mac, you can install it by running:

brew install make

Note: All the data collection scripts are designed to be run from the scripts directory, so ensure that you are in the right place before starting:

 cd scripts # From root folder

Usage

This repository provides a convenient way to collect data from the WandB Openvalidators platform using the data_collector.py tool. The repository includes a Makefile that simplifies the execution of commands for data extraction by providing predefined targets.

The data_collector.py script is designed to extract data from WandB runs based on a given set of parameters. It supports various options to specify the type of data to collect and the configuration for exporting the data.

The repository's Makefile includes the following targets to facilitate data collection:

openvalidators_dataset

make openvalidators_dataset

This command downloads all the runs from the latest version of the project and exports them to a CSV file named openvalidators_dataset.csv. It utilizes the following options under the hood of the Makefile:

  • --download_all: Downloads all the runs.
  • --export_path: Specifies the path and filename for the exported CSV file.

run_id_dataset

make run_id_dataset RUN_ID=<run_id>

This command downloads a specific run from the project and exports it to a CSV file named $(RUN_ID)_openvalidators_dataset.csv. It utilizes the following options:

  • --export_path: Specifies the path and filename for the exported CSV file.
  • --wandb_run_id: Specifies the ID of the run to download.

mining_dataset

make mining_dataset

This command downloads all the runs from the latest version of the project with a mining dataset and exports them to a CSV file named openvalidators_dataset.csv. It utilizes the following options:

  • --download_all: Downloads all the runs.
  • --export_path: Specifies the path and filename for the exported CSV file.
  • --export_mining_dataset: Enables the export of mining dataset.

scored_mining_dataset

make scored_mining_dataset

This command downloads all the runs from the latest version of the project with a scored mining dataset and exports them to a CSV file named openvalidators_dataset.csv. It utilizes the following options:

  • --download_all: Downloads all the runs.
  • --export_path: Specifies the path and filename for the exported CSV file.
  • --export_mining_with_scoring_dataset: Enables the export of mining dataset with scoring.

openai_mining_dataset

make openai_mining_dataset

This command downloads all the runs from the latest version of the project and exports them to jsonl file named openai_mining_dataset_openvalidators.jsonl in the openai fine-tuning format.

Note: Feel completely free to adjust the data_collector.py script and Makefile as necessary to match your project configuration and requirements.


Data collector parameters

The data_collector.py script accepts the following parameters:

  • --download_all: This parameter is a flag that, when set, instructs the script to download all the runs from the latest version of openvalidators. By default, it is set to False. Example usage: --download_all.

  • --include_tags: This parameter allows you to specify a list of wandb tags to filter the data collection. Example usage: --include_tags=0.1.0,nfsw_filter,your-hot-key.

  • --wandb_run_id: This parameter allows you to specify the WandB run ID to download. It is used when you want to extract data from a specific run. By default, it is set to None. Example usage: --wandb_run_id=ugutvtox.

    Note: The run_id can be retrieved at the end of your URL, e.g. :

  • --export_mining_dataset: This parameter is a flag that, when set, enables the export of the mining dataset. It is used when you want to extract data specifically for mining purposes, in the following format:

    {
        "base_prompt" : "best_followup",
        "answer_prompt" : "best_answer",    
        //...
    }

    By default, it is set to False. Example usage: --export_mining_dataset.

  • --export_mining_with_scoring_dataset: This parameter is a flag that, when set, enables the export of the mining dataset with scores, in the following format:

    {
        "base_prompt" : { "best_followup" : "score" },
        "answer_prompt" : { "best_answer" : "score" },
        //...
    }

    It is used when you want to extract data for mining with scoring. By default, it is set to False. Example usage: --export_mining_with_scoring_dataset.

  • --export_path: This parameter allows you to specify the path where the exported dataset will be saved. By default, it is set to "validator_dataset.csv". Example usage: --export_path=your_path.csv.

  • --mining_dataset_output_format: This parameter allows you to specify the output format of the mining dataset. Defaults to json, currently supports json and csv. Example usage: --mining_dataset_output_format=csv.

  • --blacklist_path: This parameter allows you to specify the path to a file containing blacklist phrases. The script will exclude any data that contains these phrases. By default, it is set to blacklist_phrases.txt. Example usage: --blacklist_path=blacklist_phrases.txt.

  • --export_openai_dataset: This parameter is a flag that, when set, enables the export of the mining dataset in the jsonl openai format for fine-tuning:

    {"prompt": "base_prompt", "completion": "best_followup"},
    {"prompt": "answer_prompt", "completion": "best_answer" }
    ...

Make sure to adjust the parameters accordingly when executing the data_collector.py script for your specific data collection needs.