Model deployment is for the results of your fine-tuning runs. If you want to use foundation models — e.g. vanilla Stable Diffusion instead of your own fine-tuned version — you do not need to deploy the model yourself.
Instead, import foundation models from the
baseten.models subpackage to directly invoke already-deployed models.
Please refer to the individual model docs for more advanced usage of models. At a high level, you can interact with models like this:
from baseten.models import StableDiffusionPipeline sd = StableDiffusionPipeline() image, url = sd("dog running on grass") # To see the image image.save("my-image.png")
More: Stable Diffusion model reference.
from baseten.models import Whisper whisper = Whisper() transcribed = whisper("https://cdn.baseten.co/docs/production/Gettysburg.mp3")
More: Whisper model reference.
from baseten.models import FlanT5 flan = FlanT5() generated = flan("Could Albert Einstein have attended the super bowl?")
More: FLAN-T5 model reference.