DeciMamba: Exploring the Length Extrapolation Potential of Mamba
Paper • 2406.14528 • Published
How to use assafbk/mamba-130m-pg19 with Transformers:
# Use a pipeline as a high-level helper
from transformers import pipeline
pipe = pipeline("text-generation", model="assafbk/mamba-130m-pg19") # Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("assafbk/mamba-130m-pg19", dtype="auto")How to use assafbk/mamba-130m-pg19 with vLLM:
# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "assafbk/mamba-130m-pg19"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
-H "Content-Type: application/json" \
--data '{
"model": "assafbk/mamba-130m-pg19",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'docker model run hf.co/assafbk/mamba-130m-pg19
How to use assafbk/mamba-130m-pg19 with SGLang:
# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
--model-path "assafbk/mamba-130m-pg19" \
--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": "assafbk/mamba-130m-pg19",
"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 "assafbk/mamba-130m-pg19" \
--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": "assafbk/mamba-130m-pg19",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'How to use assafbk/mamba-130m-pg19 with Docker Model Runner:
docker model run hf.co/assafbk/mamba-130m-pg19
The official checkpoint of Mamba-130m, finetuned for Language Modeling over the PG-19 dataset as presented in DeciMamba: Exploring the Length Extrapolation Potential of Mamba.
See our Github Repo for evalution and training scripts.
Bibtex:
@misc{benkish2024decimambaexploringlengthextrapolation,
title={DeciMamba: Exploring the Length Extrapolation Potential of Mamba},
author={Assaf Ben-Kish and Itamar Zimerman and Shady Abu-Hussein and Nadav Cohen and Amir Globerson and Lior Wolf and Raja Giryes},
year={2024},
eprint={2406.14528},
archivePrefix={arXiv},
primaryClass={cs.LG},
url={https://arxiv.org/abs/2406.14528},
}