suryatmodulus/LFM2.5-Encoder-350M-Prompt-Router

1---2language:3- en4- de5- es6- fr7- it8- nl9- pl10- pt11- ar12- hi13- ja14- ru15- tr16- vi17- zh18tags:19- liquid20- lfm221- lfm2.522- bidirectional23- masked-lm24- encoder25library_name: transformers26license: other27license_name: lfm1.028license_link: LICENSE29pipeline_tag: text-classification30base_model:31    - LiquidAI/LFM2.5-Encoder-350M32---33 34<div align="center">35  <img 36    src="https://cdn-uploads.huggingface.co/production/uploads/61b8e2ba285851687028d395/2b08LKpev0DNEk6DlnWkY.png" 37    alt="Liquid AI" 38    style="width: 100%; max-width: 100%; height: auto; display: inline-block; margin-bottom: 0.5em; margin-top: 0.5em;"39  />40  <div style="display: flex; justify-content: center; gap: 0.5em; margin-bottom: 1em;">41    <a href="https://playground.liquid.ai/"><strong>Try LFM</strong></a> • 42    <a href="https://docs.liquid.ai/lfm/getting-started/welcome"><strong>Docs</strong></a> • 43    <a href="https://leap.liquid.ai/"><strong>LEAP</strong></a> • 44    <a href="https://discord.com/invite/liquid-ai"><strong>Discord</strong></a>45  </div>46</div>47 48# LFM2.5-Encoder-350-Prompt-Router49 50A full fine-tune of [LFM2.5-Encoder-350M](https://huggingface.co/LiquidAI/LFM2.5-Encoder-350M) with a zero-shot routing head that scores a prompt against user-defined routing lanes in a single encoder pass.51 52Find more details about our encoders in our [blog post](https://www.liquid.ai/blog/lfm2-5-encoders).53 54> [!NOTE]55> 💻 **Demos**: Try this fine-tuned model running in a CPU-only Hugging Face space:56> **[Zero-shot prompt routing](https://huggingface.co/spaces/LiquidAI/prompt-routing)** — define your own routing lanes as free text. The model scores the whole prompt against every lane in one pass.57 58## Usage59 60> ⚠️ Loads custom code via `trust_remote_code=True` (the model wraps a `trust_remote_code` encoder).61 62Install the required packages:63 64```bash65pip install torch transformers66```67 68Run zero-shot prompt routing:69 70```python71from transformers import AutoModel, AutoTokenizer72 73model_id = "LiquidAI/LFM2.5-Encoder-350-Prompt-Router"74 75tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)76model = AutoModel.from_pretrained(model_id, trust_remote_code=True).eval()77 78routes = ["Coding", "Sales", "Creative writing", "General knowledge"]79prompt = "Can you help me debug a failing Python unit test?"80 81print(model.route(prompt, routes, tokenizer=tokenizer))82```83 84## 📬 Contact85 86- Got questions or want to connect? [Join our Discord community](https://discord.com/invite/liquid-ai)87- If you are interested in custom solutions with edge deployment, please contact [our sales team](https://www.liquid.ai/contact).88 89## Citation90 91```bibtex92@article{liquidAI2026Encoders,93  author = {Liquid AI},94  title = {LFM2.5-Encoders: Fast at Long Context, Even on CPU},95  journal = {Liquid AI Blog},96  year = {2026},97  note = {www.liquid.ai/blog/lfm2-5-encoders},98}99```100