Small library which aim is to check your dataset for being solved by simple heuristics. Multilingual!
Project description
check_your_heuristic
#Quick start
Installation
pip install check_your_heuristic
Configurations
To check your dataset fill the config, using unix-like paths.
Example config:
train_dataset_dir: "dataset/dir/train.jsonl"
valid_dataset_dir: "dataset/dir/val.jsonl"
column_name1: "premise"
column_name2: "hypothesis"
target_name: "label"
output_dir: ""
Output_dir parameter is optional. If it isn't specified, all the plots will be saved into output_check_your_heuristic/
folder in the directory where you've ran the command
Other config variations can be found here
CLI Use
Our library offers four build-in commands for checking your datasets depending on the dataset structure you have.
- Base case or two text columns + one target column (for example, CommitmentBank from SuperGLUE or TERRa from Russian SuperFLUE)
run-base-case --path_to_config config.yaml
- When you have some long text and some questions and answers for it (for example, MultiRC from SuperGLUE or MuSeRC from Russian SuperFLUE)
run-multirc-case --path_to_config config.yaml
- When you have passage, questions and some NERs (or entities) that serve as answers (for example, ReCoRD from SuperGLUE or RuCoS from Russian SuperFLUE)
run-record-case --path_to_config config.yaml
- Case when you have two cases and need to compare some words in them (for example, Words in Context (WiC) from SuperGLUE or RUSSE from Russian SuperFLUE)
run-wordincontext-case --path_to_config config.yaml
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
check_your_heuristic-0.0.1.tar.gz
(11.9 kB
view hashes)
Close
Hashes for check_your_heuristic-0.0.1.tar.gz
Algorithm | Hash digest | |
---|---|---|
SHA256 | 7c37b0e71fc2ef7f8713790ed646c4a79250cd5e451195eca8abfa1c87a2db2e |
|
MD5 | ccbe4a4b0c9592a07e5102641d66c093 |
|
BLAKE2b-256 | f02f1d8f0fc7dd2d14f5e316f72df48eb8146779109d5f59283f03e7b2a03615 |