Which data format is loaded when I use the high level ‘pipeline’ function on a model which has both pickle and safetensors files? What is the default behavior and how do I guarantee the safetensors will be loaded? Couldn’t find a definitive answer on the docs.
For example:
t5-base has both safetensors and pickle files in the repo.
summarizer = pipeline(“summarization”, model=“t5-base”)