Skip to content

maxreciprocate/kaggle-lmarena-1st-place

Repository files navigation

WSDM Cup - Multilingual Chatbot Arena

https://www.kaggle.com/competitions/wsdm-cup-multilingual-chatbot-arena

Requirements

Hardware

H100x8 (to speed up teachers training we used H100x32, however it's not required)

Software

python3.12
SLURM

Packages

deepspeed==0.15.4
accelerate==1.2.1
transformers==4.46.3
flash-attn==2.7.1.post4

Reproduction

Preparation

Download the following datasets and put them into data/ without renaming the files (the rest of datasets are available on huggingface):

Training

Structure

All scripts have to be executed from the root directory

├── data
│   ├── train.parquet
│   └── ...
├── ckpts
│   └── ...
├── stage1
│   ├── README.md
│   └── prepare_pretrain_data.py
├── stage2
│   ├── README.md
│   └── prepare_teacher_data.py
├── stage3
│   ├── README.md
│   ├── collect_labels.py
│   ├── merge_students.py
│   ├── pack_student.py
│   ├── prepare_student_data.py
│   └── prepare_synth_data.py
├── packing
│   └── # https://github.com/tascj/kaggle-lmsys-chatbot-arena/tree/main/human_pref
├── format.py
├── label.py
├── label.sh
├── label.slurm
├── launch.slurm
├── models.py
├── readme.md
├── requirements.txt
├── run.sh
└── train.py

About

Solution of Kaggle "Multilingual Chatbot Arena" competition

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors