How to use from
SGLangUse Docker images
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 "rpchinhara/qwen-3.5-4b-cs6013" \
--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": "rpchinhara/qwen-3.5-4b-cs6013",
"prompt": "Once upon a time,",
"max_tokens": 512,
"temperature": 0.5
}'Quick Links
Qwen3.5-4B Model Compression
Model Description
This repository contains a compressed version of Qwen3.5-4B, developed as part of the CS6013: Efficient AI course project. The objective is to reduce model memory while preserving performance on mathematical reasoning tasks using model compression techniques.
- Base model: Qwen/Qwen3.5-4B
- Developed by: Rudrapratap Chinhara
- Course: CS6013 – Efficient AI
- Affliation: IIT Bombay
- Language: English
- License: CC BY-NC 4.0
Compression Method
This model was obtained by applying:
- Quantization
- Pruning (if applicable)
- Other compression techniques (if applicable)
Evaluation
| Metric | Value |
|---|---|
| Model Size | XX GB |
| Compression Ratio | X.XX× |
| Accuracy | XX.XX% |
Citation
If you use this repository, please cite the original Qwen model:
@misc{qwen3.5,
title = {{Qwen3.5}: Towards Native Multimodal Agents},
author = {{Qwen Team}},
month = {February},
year = {2026},
url = {https://qwen.ai/blog?id=qwen3.5}
}
Acknowledgements
This work is based on the Qwen3.5-4B model released by the Qwen Team and was developed for the CS6013 Efficient AI course project.
Install from pip and serve model
# Install SGLang from pip: pip install sglang# Start the SGLang server: python3 -m sglang.launch_server \ --model-path "rpchinhara/qwen-3.5-4b-cs6013" \ --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": "rpchinhara/qwen-3.5-4b-cs6013", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'