LightXML

LightXML: Transformer with dynamic negative sampling for High-Performance Extreme Multi-label Text Classification

Requirements

pip install -r requirements.txt

Please also install apex as follows

git clone https://github.com/NVIDIA/apex
cd apex
pip install -v --disable-pip-version-check --no-cache-dir --global-option="--cpp_ext" --global-option="--cuda_ext" ./

Datasets

LightXML uses the same dataset with AttentionXML and X-Transform.

please download them from the following links.

• EUR-Lex
• Wiki10-31K
• AmazonCat-13K
• Amazon-670K
• Wiki-500K

Experiments

train and eval

./run.sh [eurlex4k|wiki31k|amazon13k|amazon670k|wiki500k]