How to use flashrt/fp8-kv-attention with Transformers:
# Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("flashrt/fp8-kv-attention", device_map="auto")
