Weblinx

Description

This dataset is a Minari formatted version of WebLINX. WebLINX is a benchmark for evaluating web navigation agents, and provides a dataset featuring complex goal-oriented tasks within interactive web environments.

Dataset Specs

Total Steps

24417

Total Episodes

969

Dataset Observation Space

Dict('action_history': Text(1, 16384, charset=0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz), 'candidates': Text(1, 16384, charset=0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz), 'clean_html': Text(1, 16384, charset=0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz), 'utterances': Text(1, 8192, charset=0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz), 'viewport': Text(1, 1024, charset=0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz))

Dataset Action Space

Text(1, 1024, charset=0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz)

Algorithm

Not provided

Author

Zdeněk Kasner, Siva Reddy, Xing Han Lù

Email

xing.han.lu@mail.mcgill.ca

Code Permalink

https://github.com/Farama-Foundation/minari-dataset-generation-scripts

Minari Version

0.5.2 (supported)

Download

minari download webagents/weblinx-v0

This dataset doesn’t contain an env_spec, neither an eval_env_spec attribute. Any call to minari.MinariDataset.recover_environment() will throw an error.