Find the most efficient way for a specific large language model to learn a specific task
- CoreScheduler: A High-Performance Scheduling Framework for Large-Scale Model Training in C++.
- Tiny-DeepSpeed: A minimalistic re-implementation of the DeepSpeed library.
- Tiny-Megatron: A minimalistic re-implementation of the Megatron library.
git clone https://github.com/kaustpradalab/LLM-efficient-learning.git
cd LLM-efficient-learning
# To use the newest version of all repos.
git submodule update --init --recursive
# or use the specific repo
git submodule update --init path/to/specific/repo
cd path/to/specific/repo