Instructions to use liufanfanlff/C3-Context-Cascade-Compression with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use liufanfanlff/C3-Context-Cascade-Compression with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-generation", model="liufanfanlff/C3-Context-Cascade-Compression", trust_remote_code=True)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("liufanfanlff/C3-Context-Cascade-Compression", trust_remote_code=True, dtype="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use liufanfanlff/C3-Context-Cascade-Compression with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "liufanfanlff/C3-Context-Cascade-Compression"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "liufanfanlff/C3-Context-Cascade-Compression",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use Docker

docker model run hf.co/liufanfanlff/C3-Context-Cascade-Compression

SGLang

How to use liufanfanlff/C3-Context-Cascade-Compression with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "liufanfanlff/C3-Context-Cascade-Compression" \
    --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": "liufanfanlff/C3-Context-Cascade-Compression",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Use 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 "liufanfanlff/C3-Context-Cascade-Compression" \
        --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": "liufanfanlff/C3-Context-Cascade-Compression",
		"prompt": "Once upon a time,",
		"max_tokens": 512,
		"temperature": 0.5
	}'

Docker Model Runner
How to use liufanfanlff/C3-Context-Cascade-Compression with Docker Model Runner:
```
docker model run hf.co/liufanfanlff/C3-Context-Cascade-Compression
```

Context Cascade Compression: Exploring the Upper Limits of Text Compression

🌟GitHub | 📜Paper

Fanfan Liu, Haibo Qiu

Usage

from transformers import AutoModel, AutoTokenizer

model_name = 'liufanfanlff/C3-Context-Cascade-Compression'
tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
model = AutoModel.from_pretrained(model_name , trust_remote_code=True, low_cpu_mem_usage=True, device_map='cuda', use_safetensors=True, pad_token_id=tokenizer.eos_token_id)
model = model.eval().cuda()
prompt = 'Repeat the text: '
context = "帝高阳之苗裔兮，朕皇考曰伯庸。摄提贞于孟陬兮，"
#context = "lfflfflfflfflfflfflfflfflff"
outputs = model.chat(tokenizer, context, prompt)
print ("Repeat the text: ",outputs)

viz

Contact

Don't hesitate to contact me by email, liufanfan19@mails.ucas.ac.cn, if you have any questions.

Acknowledgement

DeepSeek-OCR: the idea originated from reconsideration of this work.
GOT-OCR2.0: the code was adapted from GOT-OCR2.0.
Qwen: the LLM base model of C3.

Citation

@article{liu2025context,
  title={Context Cascade Compression: Exploring the Upper Limits of Text Compression},
  author={Liu, Fanfan and Qiu, Haibo},
  journal={arXiv preprint arXiv:2511.15244},
  year={2025}
}

Downloads last month: 299

Safetensors

Model size

3B params

Tensor type

BF16

Paper for liufanfanlff/C3-Context-Cascade-Compression

Context Cascade Compression: Exploring the Upper Limits of Text Compression

Paper • 2511.15244 • Published Nov 19, 2025 • 4