⚡ LitGPT

Lightning.ai • Install • Get started • Use LLMs • Finetune LLMs • Pretrain LLMs • Models • Features • Templates • YAML configs • Design principles •

 

⚡ Pretrain, finetune, deploy over 20 LLMs. Uses state-of-the-art techniques like flash attention, 4-bit, LoRA, and more.

Install LitGPT

Install LitGPT with all dependencies (including CLI, quantization, tokenizers for all models, etc.):

pip install 'litgpt[all]'

Advanced install options

Install from source:

git clone https://github.com/Lightning-AI/litgpt
cd litgpt
pip install -e '.[all]'

 

Get started

LitGPT is a command-line tool to use, pretrain, finetune and deploy LLMs.

 

Use an LLM

Here's an example showing how to use the Mistral 7B LLM.

# 1) Download a pretrained model
litgpt download --repo_id mistralai/Mistral-7B-Instruct-v0.2

# 2) Chat with the model
litgpt chat \
  --checkpoint_dir checkpoints/mistralai/Mistral-7B-Instruct-v0.2

>> Prompt: What do Llamas eat?

For more information, refer to the download and inference tutorials.

 

Finetune an LLM

Finetune a model to specialize it on your own custom dataset. Here's an example that finetunes phi-2:

# 1) Download a pretrained model
litgpt download --repo_id microsoft/phi-2

# 2) Finetune the model
litgpt finetune lora \
  --checkpoint_dir checkpoints/microsoft/phi-2 \
  --data Alpaca2k \
  --out_dir out/phi-2-lora

# 3) Chat with the model
litgpt chat \
  --checkpoint_dir out/phi-2-lora/final

 

Pretrain an LLM

Train an LLM from scratch on your own data via pretraining:

# 1) Download a pretrained model
litgpt download --repo_id microsoft/phi-2

# 2) Finetune the model
litgpt pretrain \
  --initial_checkpoint_dir checkpoints/microsoft/phi-2 \
  --data Alpaca2k \
  --out_dir out/custom-phi-2

# 3) Chat with the model
litgpt chat \
  --checkpoint_dir out/phi-2-lora/final

Finetune an LLM on your own data:

 

Note

Full guide and docs are here: Zero to LitGPT: Getting Started with Pretraining, Finetuning, and Using LLMs if you are looking to get started with using LitGPT.

 

Choose from 20 LLMs

✅  View the full list.

Model	Model size	Reference
Code Llama by Meta AI	7B, 13B, 34B, 70B	Rozière et al. 2023
Dolly by Databricks	3B, 7B, 12B	Conover et al. 2023
Falcon by TII UAE	7B, 40B, 180B	TII 2023
FreeWilly2 (Stable Beluga 2) by Stability AI	70B	Stability AI 2023
Function Calling Llama 2 by Trelis	7B	Trelis et al. 2023
Gemma by Google	2B, 7B	Google Team, Google Deepmind
Llama 2 by Meta AI	7B, 13B, 70B	Touvron et al. 2023
LongChat by LMSYS	7B, 13B	LongChat Team 2023
Mistral and Mixtral by Mistral AI	7B	Mistral website
Nous-Hermes by NousResearch	7B, 13B, 70B	Org page
OpenLLaMA by OpenLM Research	3B, 7B, 13B	Geng & Liu 2023
Phi by Microsoft Research	1.3B, 2.7B	Li et al. 2023
Platypus by Lee at el.	7B, 13B, 70B	Lee, Hunter, and Ruiz 2023
Pythia by EleutherAI	{14,31,70,160,410}M, {1,1.4,2.8,6.9,12}B	Biderman et al. 2023
RedPajama-INCITE by Together	3B, 7B	Together 2023
StableCode by Stability AI	3B	Stability AI 2023
StableLM by Stability AI	3B, 7B	Stability AI 2023
StableLM Zephyr by Stability AI	3B	Stability AI 2023
TinyLlama by Zhang et al.	1.1B	Zhang et al. 2023
Vicuna by LMSYS	7B, 13B, 33B	Li et al. 2023

 

State-of-the-art features

✅  State-of-the-art optimizations: Flash Attention v2, multi-GPU support via fully-sharded data parallelism, optional CPU offloading, and TPU and XLA support.

✅  Pretrain, finetune, and deploy

✅  Various precision settings: FP32, FP16, BF16, and FP16/FP32 mixed.

✅  Configuration files for great out-of-the-box performance.

✅  Efficient finetuning: LoRA, QLoRA, Adapter, and Adapter v2.

✅  Quantization: 4-bit floats, 8-bit integers, and double quantization.

✅  Exporting to other popular model weight formats.

✅  Many popular datasets for pretraining and finetuning, and support for custom datasets.

✅  Readable and easy-to-modify code to experiment with the latest research ideas.

 
 

Project templates

The following Lightning Studio templates provide LitGPT tutorials and projects in reproducible environments with multi-GPU and multi-node support:

Prepare the TinyLlama 1T token dataset	Pretrain LLMs - TinyLlama 1.1B
Continued Pretraining with TinyLlama 1.1B	Instruction finetuning - TinyLlama 1.1B LLM

 
 

Use optimized configurations

LitGPT comes with out-of-the-box, highly performant settings via our YAML configs.

litgpt finetune lora \
  --config https://raw.githubusercontent.com/Lightning-AI/litgpt/main/config_hub/finetune/llama-2-7b/lora.yaml

Override any parameter in the CLI:

litgpt finetune lora \
  --config https://raw.githubusercontent.com/Lightning-AI/litgpt/main/config_hub/finetune/llama-2-7b/lora.yaml \
  --lora_r 4

Browse the available configuration files here.

 

Tip

Run large models on smaller consumer devices: We support 4-bit quantization (as in QLoRA), (bnb.nf4, bnb.nf4-dq, bnb.fp4, bnb.fp4-dq) and 8-bit quantization (bnb.int8) for inference by following this guide.

 
 

Customize configs

LitGPT supports rich and customizable config files to tailor the LLM training to your dataset and hardware needs. Shown below is a configuration file for LoRA finetuning:

# The path to the base model's checkpoint directory to load for finetuning. (type: <class 'Path'>, default: checkpoints/stabilityai/stablelm-base-alpha-3b)
checkpoint_dir: checkpoints/meta-llama/Llama-2-7b-hf

# Directory in which to save checkpoints and logs. (type: <class 'Path'>, default: out/lora)
out_dir: out/finetune/qlora-llama2-7b

# The precision to use for finetuning. Possible choices: "bf16-true", "bf16-mixed", "32-true". (type: Optional[str], default: null)
precision: bf16-true

# If set, quantize the model with this algorithm. See ``tutorials/quantize.md`` for more information. (type: Optional[Literal['nf4', 'nf4-dq', 'fp4', 'fp4-dq', 'int8-training']], default: null)
quantize: bnb.nf4

# How many devices/GPUs to use. (type: Union[int, str], default: 1)
devices: 1

# The LoRA rank. (type: int, default: 8)
lora_r: 32

# The LoRA alpha. (type: int, default: 16)
lora_alpha: 16

# The LoRA dropout value. (type: float, default: 0.05)
lora_dropout: 0.05

# Whether to apply LoRA to the query weights in attention. (type: bool, default: True)
lora_query: true

# Whether to apply LoRA to the key weights in attention. (type: bool, default: False)
lora_key: false

# Whether to apply LoRA to the value weights in attention. (type: bool, default: True)
lora_value: true

# Whether to apply LoRA to the output projection in the attention block. (type: bool, default: False)
lora_projection: false

# Whether to apply LoRA to the weights of the MLP in the attention block. (type: bool, default: False)
lora_mlp: false

# Whether to apply LoRA to output head in GPT. (type: bool, default: False)
lora_head: false

# Data-related arguments. If not provided, the default is ``litgpt.data.Alpaca``.
data:
  class_path: litgpt.data.Alpaca2k
  init_args:
    mask_prompt: false
    val_split_fraction: 0.05
    prompt_style: alpaca
    ignore_index: -100
    seed: 42
    num_workers: 4
    download_dir: data/alpaca2k

# Training-related arguments. See ``litgpt.args.TrainArgs`` for details
train:

  # Number of optimizer steps between saving checkpoints (type: Optional[int], default: 1000)
  save_interval: 200

  # Number of iterations between logging calls (type: int, default: 1)
  log_interval: 1

  # Number of samples between optimizer steps across data-parallel ranks (type: int, default: 128)
  global_batch_size: 8

  # Number of samples per data-parallel rank (type: int, default: 4)
  micro_batch_size: 2

  # Number of iterations with learning rate warmup active (type: int, default: 100)
  lr_warmup_steps: 10

  # Number of epochs to train on (type: Optional[int], default: 5)
  epochs: 4

  # Total number of tokens to train on (type: Optional[int], default: null)
  max_tokens:

  # Limits the number of optimizer steps to run (type: Optional[int], default: null)
  max_steps:

  # Limits the length of samples (type: Optional[int], default: null)
  max_seq_length: 512

  # Whether to tie the embedding weights with the language modeling head weights (type: Optional[bool], default: null)
  tie_embeddings:

  #   (type: float, default: 0.0003)
  learning_rate: 0.0002

  #   (type: float, default: 0.02)
  weight_decay: 0.0

  #   (type: float, default: 0.9)
  beta1: 0.9

  #   (type: float, default: 0.95)
  beta2: 0.95

  #   (type: Optional[float], default: null)
  max_norm:

  #   (type: float, default: 6e-05)
  min_lr: 6.0e-05

# Evaluation-related arguments. See ``litgpt.args.EvalArgs`` for details
eval:

  # Number of optimizer steps between evaluation calls (type: int, default: 100)
  interval: 100

  # Number of tokens to generate (type: Optional[int], default: 100)
  max_new_tokens: 100

  # Number of iterations (type: int, default: 100)
  max_iters: 100

# The name of the logger to send metrics to. (type: Literal['wandb', 'tensorboard', 'csv'], default: csv)
logger_name: csv

# The random seed to use for reproducibility. (type: int, default: 1337)
seed: 1337

 

LitGPT design principles

This repository follows the main principle of openness through clarity.

LitGPT is:

• Simple: Single-file implementation without boilerplate.
• Correct: Numerically equivalent to the original model.
• Optimized: Runs fast on consumer hardware or at scale.
• Open-source: No strings attached.

Avoiding code duplication is not a goal. Readability and hackability are.

 

Get involved!

We appreciate your feedback and contributions. If you have feature requests, questions, or want to contribute code or config files, please don't hesitate to use the GitHub Issue tracker.

We welcome all individual contributors, regardless of their level of experience or hardware. Your contributions are valuable, and we are excited to see what you can accomplish in this collaborative and supportive environment.

 

Tip

Unsure about contributing? Check out our How to Contribute to LitGPT guide.

If you have general questions about building with LitGPT, please join our Discord.

 

Tutorials, how-to guides, and docs

Note

We recommend starting with the Zero to LitGPT: Getting Started with Pretraining, Finetuning, and Using LLMs if you are looking to get started with using LitGPT.

Tutorials and in-depth feature documentation can be found below:

• Finetuning, incl. LoRA, QLoRA, and Adapters (tutorials/finetune.md)
• Pretraining (tutorials/pretrain.md)
• Model evaluation (tutorials/evaluation.md)
• Supported and custom datasets (tutorials/prepare_dataset.md)
• Quantization (tutorials/quantize.md)
• Tips for dealing with out-of-memory (OOM) errors (tutorials/oom.md)

 

XLA

Lightning AI has partnered with Google to add first-class support for Cloud TPUs in Lightning's frameworks and LitGPT, helping democratize AI for millions of developers and researchers worldwide.

Using TPUs with Lightning is as straightforward as changing one line of code.

We provide scripts fully optimized for TPUs in the XLA directory.

 

Acknowledgements

This implementation extends on Lit-LLaMA and nanoGPT, and it's powered by Lightning Fabric ⚡.

• @karpathy for nanoGPT
• @EleutherAI for GPT-NeoX and the Evaluation Harness
• @TimDettmers for bitsandbytes
• @Microsoft for LoRA
• @tridao for Flash Attention 2

 

Community showcase

Check out the projects below that use and build on LitGPT. If you have a project you'd like to add to this section, please don't hesitate to open a pull request.

 

🏆 NeurIPS 2023 Large Language Model Efficiency Challenge: 1 LLM + 1 GPU + 1 Day

The LitGPT repository was the official starter kit for the NeurIPS 2023 LLM Efficiency Challenge, which is a competition focused on finetuning an existing non-instruction tuned LLM for 24 hours on a single GPU.

 

🦙 TinyLlama: An Open-Source Small Language Model

LitGPT powered the TinyLlama project and TinyLlama: An Open-Source Small Language Model research paper.

 

Citation

If you use LitGPT in your research, please cite the following work:

@misc{litgpt-2023,
  author       = {Lightning AI},
  title        = {LitGPT},
  howpublished = {\url{https://github.com/Lightning-AI/litgpt}},
  year         = {2023},
}

 

License

LitGPT is released under the Apache 2.0 license.