Reranking Model for Locker.ai

Reranking Model for Locker.ai is a model for more accurate searches that takes into account the validity of declarations through domain-specific reevaluation.

Core Contributors 🛠️

shio	ituki
#llama-model-composer	#bert-model-composer

Setup with Dev Containers 📦

You can easily launch the development environment of Reranking Model for Locker.ai with Dev Containers.
Here is the step-by-step guide.

Attention

• You need to install Docker and VSCode before.

1. clone git repository

git clone "https://github.com/ashitano-dcon/lockerai-reranking" && cd "./lockerai-reranking/"

2. launch dev containers

Launch containers using the VSCode extension Dev Containers.

3. pin python version

rye pin $(cat "./.python-version")

4. install dependencies

rye sync

5. activate virtual environment

source "./.venv/bin/activate"

6. install FlashAttention-2

uv pip install flash-attn --no-build-isolation

7. train model

rye run python -m llama.train

Setup locally 🖥️

If you want to build an environment more quickly without Docker, you can follow these steps to build your environment locally.

Attention

• You need to install rye before.
• [Optional] You should install project recommended VSCode extensions that specified in .devcontainer/devcontainer.json before.

1. clone git repository

git clone "https://github.com/ashitano-dcon/lockerai-reranking" && cd "./lockerai-reranking/"

2. pin python version

rye pin $(cat "./.python-version")

3. install dependencies

rye sync

4. activate virtual environment

source "./.venv/bin/activate"

5. install FlashAttention-2

uv pip install flash-attn --no-build-isolation

6. train model

rye run python -m llama.train