README.md

---
title: Aidc Llm Laos
emoji: 📉
colorFrom: indigo
colorTo: green
sdk: gradio
sdk_version: 5.42.0
app_file: app.py
pinned: false
license: other
short_description: aidc-llm--laos
---

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

# AIDC LLM Laos 10k (Gemma-3 4B IT) — Chat Space

This Space runs a Gradio chat UI for:
`Phonepadith/aidc-llm-laos-10k-gemma-3-4b-it`

## Deploy
- Space type: Python
- Hardware: Prefer GPU (T4/A10/A100). CPU works but is slow.
- Add files:
  - app.py
  - requirements.txt
- (Optional) Set Space secret `USE_4BIT=1` to enable 4-bit loading (GPU + bitsandbytes).

## Notes
- The app uses the tokenizer's chat template via `apply_chat_template`.
- Adjust `max_new_tokens` to control response length.
- If OOM on small GPUs, try:
  - Reduce `max_new_tokens`
  - Set `USE_4BIT=1`
  - Enable `low_cpu_mem_usage=True` (can be added in the loader)