thesreedath/slm-125m-legal-demo
0
Legal SLM (125M) - inference API
FastAPI service for a 125M decoder-only Transformer pretrained from scratch on ~2.4B tokens of US caselaw, SEC filings, and general English. Base (pretrained-only) model: a text completer, not a chatbot. Runs on CPU.
POST /generate with JSON {prompt, max_new_tokens, temperature, top_p, repetition_penalty} returns {prompt, completion, generated}. CORS is open so a browser frontend can call it.