How to use transformers-community/sink_cache with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("transformers-community/sink_cache", dtype="auto")
