Liger
Collection
6 items β’ Updated β’ 3
How to use linear-moe-hub/Liger-GSA-8B with Transformers:
# Use a pipeline as a high-level helper
from transformers import pipeline
pipe = pipeline("text-generation", model="linear-moe-hub/Liger-GSA-8B") # Load model directly
from transformers import AutoModelForCausalLM
model = AutoModelForCausalLM.from_pretrained("linear-moe-hub/Liger-GSA-8B", dtype="auto")How to use linear-moe-hub/Liger-GSA-8B with vLLM:
# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "linear-moe-hub/Liger-GSA-8B"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
-H "Content-Type: application/json" \
--data '{
"model": "linear-moe-hub/Liger-GSA-8B",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'docker model run hf.co/linear-moe-hub/Liger-GSA-8B
How to use linear-moe-hub/Liger-GSA-8B with SGLang:
# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
--model-path "linear-moe-hub/Liger-GSA-8B" \
--host 0.0.0.0 \
--port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
-H "Content-Type: application/json" \
--data '{
"model": "linear-moe-hub/Liger-GSA-8B",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'docker run --gpus all \
--shm-size 32g \
-p 30000:30000 \
-v ~/.cache/huggingface:/root/.cache/huggingface \
--env "HF_TOKEN=<secret>" \
--ipc=host \
lmsysorg/sglang:latest \
python3 -m sglang.launch_server \
--model-path "linear-moe-hub/Liger-GSA-8B" \
--host 0.0.0.0 \
--port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/completions" \
-H "Content-Type: application/json" \
--data '{
"model": "linear-moe-hub/Liger-GSA-8B",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'How to use linear-moe-hub/Liger-GSA-8B with Docker Model Runner:
docker model run hf.co/linear-moe-hub/Liger-GSA-8B
[π GitHub] [π Liger] [π arXiv]
We introduce Liger-GSA-8B, a gated linear recurrent model linearized from Transformer-based LLM.
Our Liger framework is compatible with various linear recurrent models with gating structures:
| Model Name | Base Model | Linear Structure | HF Link |
|---|---|---|---|
| Liger-GLA-8B | Llama-3-8B | GLA | π€ link |
| Liger-GSA-8B | Llama-3-8B | GSA | π€ link |
If you find this repo useful, please cite and star our work:
@article{lan2025liger,
title={Liger: Linearizing Large Language Models to Gated Recurrent Structures},
author={Lan, Disen and Sun, Weigao and Hu, Jiaxi and Du, Jusen and Cheng, Yu},
journal={arXiv preprint arXiv:2503.01496},
year={2025}
}