AUBIN-Learn: instant self-learning memory, fast skills, self-calibration + reports

d7feae2 verified 10 days ago

461 Bytes

	# AUBIN server (GPU). Build: docker build -t aubin .
	# Run: docker run --gpus all -p 8009:8009 -e AUBIN_MODEL=emrevrg/AUBIN-12B aubin
	FROM pytorch/pytorch:2.5.1-cuda12.4-cudnn9-runtime
	WORKDIR /app
	RUN pip install --no-cache-dir -U "transformers>=4.50" accelerate bitsandbytes peft huggingface_hub
	COPY aubin/ /app/aubin/
	ENV AUBIN_MODEL=emrevrg/AUBIN-12B PORT=8009
	EXPOSE 8009
	CMD ["sh", "-c", "python -m aubin.cli serve --model $AUBIN_MODEL --port $PORT"]