Hey, you could use a template on brev.dev to spin up a gpu box with the model and Jupyter notebook. Alternatively, the falcon 7b model should be small enough for colab