https://github.com/wschella/lass

https://github.com/wschella/lass

Last synced: 6 days ago
JSON representation

• Host: GitHub
• URL: https://github.com/wschella/lass
• Owner: wschella
• Created: 2022-04-07T13:08:32.000Z (over 2 years ago)
• Default Branch: master
• Last Pushed: 2024-07-30T13:06:01.000Z (4 months ago)
• Last Synced: 2024-07-31T15:43:52.032Z (4 months ago)
• Language: Jupyter Notebook
• Size: 82.8 MB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Language assessors

This repo contains the code for the experiments regarding building [assessor models](https://www.aaai.org/AAAI22Papers/SMT-00432-Hernandez-OralloJ.pdf) from the enormous pile of evaluation data generated by [BIG-bench](https://github.com/google/BIG-bench).

## Setup

Due to the use of Git LFS & Git submodules, mixing both Conda and Poetry, and the usual Python install experience (especially in ML), the process is a bit long.

1. Make sure [Git LFS](https://git-lfs.com/) is installed.

2. Clone this repo. Use `git clone --recursive [email protected]:wschella/lass.git` to also immediately clone the BIG-bench submodule.

3. Pull in BIG-bench data: `cd lass; cd bigbench; git lfs install; git lfs pull; cd -`.

4. Initialize the Conda environment.yml with `conda env create --prefix .venv -f environment.yml` (or however you prefer).

5. `poetry install` to install all extra dependencies.

6. Decompress all instance data with `./scripts/decompress.sh` (UNIX only, requires `parallel` and `lrzip` to be installed.)