relora
Official code for ReLoRA from the paper Stack More Layers Differently: High-Rank Training Through Low-Rank Updates
File Explorer
Download Latest Version (.zip)- launch.json
- llama_100m.json
- llama_130m.json
- llama_1b.json
- llama_20m.json
- llama_250m.json
- llama_250m_50K.json
- llama_250m_old.json
- llama_350m.json
- llama_35m.json
- llama_3b.json
- llama_40m.json
- llama_60m.json
- llama_71m.json
- llama_7b.json
- llama_9m.json
- pile_megatron_dataset.yaml
- pythia_tokenizer.json
- 01_peft_pretraining.ipynb
- 02_quick_debugs.ipynb
- 03_scaling_laws_plotting.ipynb
- 04_plot_lr.ipynb
- 05_check_ranks.ipynb
- 06_svd.ipynb
- 07_plotting.ipynb
- 08_ranks_before_and_after.ipynb
- 09_bar_plot.ipynb
- 10_chunking.ipynb
- 11_test_pythia.ipynb
- 12_test_relora_init.ipynb
- 13_zero_optimizer_resets.ipynb
- 14_check_pretokenization.ipynb
- 15_debug_dataloading.ipynb
- 16_quantized.ipynb
- arguments.py
- blendable_dataset.py
- data_utils.py
- dataloader.py
- dataset.py
- helpers.cpp
- indexed_dataset.py
- Makefile
- neox_args.py
- samplers.py
- args_utils.py
- dataloader.py
- modeling_llama.py
- modeling_pythia.py
- relora.py
- training_utils.py
- 1B_v1.0.yaml
- .gitignore
- CITATION.cff
- LICENSE
- pretokenize.py
- README.dev.md
- README.md
- requirements.txt
- run_glue.py
- setup.py
- torchrun_main.py
# Installation Guide
1. Get the code
git clone https://github.com/Guitaricet/relora
Downloads the entire project code from GitHub to your computer.
cd relora
Moves into the project folder you just downloaded.
2. Official Install Script
Easy RecommendedPrerequisites
- Python 3 Python is required to use pip.
pip install flash-attn
Installs the package published on PyPI directly β no need to clone the source.
After installing, open a new terminal and run the program's version command (e.g. --version) to confirm it worked.
Pulled directly from this repo's README.
3. Python
EasyPrerequisites
pip install -e .
Installs the Python libraries listed in requirements.txt (or similar).
pip install flash-attn
Installs the package published on PyPI directly β no need to clone the source.
If it runs without errors and prints output in the terminal, it worked.
Pulled directly from this repo's README.
4. Make
MediumPrerequisites
- Git Needed to download the project code from GitHub.
- Make Usually pre-installed on Linux/macOS. On Windows, install separately (e.g. via MSYS2 or WSL).
β οΈ This is a large repository, so this method may point to an internal sub-package rather than the actual core product. Check the full README as well.
cd peft_pretraining/megatron_dataset
This project's files live in a subfolder, so move into it first.
make
Compiles the code based on the generated build configuration to produce an executable.
If it finishes without errors, it worked. Try running the generated executable directly.
// repository documentation
Was this content helpful?
(0 ratings)