README.md

ZamAI-Facebook-XLM-Pashto

Metadata

• license: MIT
• datasets: tasal9/ZamAI-Pashto-Dataset-Cleaned
• language: ps (Pashto)
• metrics: accuracy
• base_model: facebook/xlm-roberta-base
• pipeline_tag: fill-mask
• library_name: transformers / adapter-transformers

Overview

This repository contains helper scripts to download and persist the base model facebook/xlm-roberta-base locally (into ./base_model/) and to run a small fill-mask inference example. Large model files should be handled using Git LFS; .gitattributes at the repo root already includes common model file patterns.

Quick start

1. Create and activate a virtualenv:

python -m venv .venv
source .venv/bin/activate

2. Install dependencies:

pip install -r requirements.txt

3. Download and save the base model into ./base_model/:

python download_base_model.py

4. Run the example inference (will load from ./base_model/ if present):

python inference.py

Files

• download_base_model.py — downloads the Hugging Face model and saves to ./base_model/.
• inference.py — small script to run a fill-mask example.
• requirements.txt — Python dependencies.
• .gitignore — common ignores.

If you want me to download the base model into the repository now (it will download ~0.5–1.2 GB depending on files), tell me and I'll run the script and save the model into ./base_model/

license: mit datasets:

• tasal9/ZamAI-Pashto-Dataset-Cleaned language:
• ps metrics:
• accuracy base_model:
• FacebookAI/xlm-roberta-base pipeline_tag: fill-mask library_name: adapter-transformers

---