Update README.md

README.md

@@ -58,6 +58,7 @@ pip install git+https://github.com/illuin-tech/colpali
 ```python
 import torch
 from PIL import Image
+from transformers.utils.import_utils import is_flash_attn_2_available
 
 from colpali_engine.models import ColQwen2_5, ColQwen2_5_Processor
 
@@ -65,6 +66,7 @@ model = ColQwen2_5.from_pretrained(
         "vidore/colqwen2.5-v0.2",
         torch_dtype=torch.bfloat16,
         device_map="cuda:0",  # or "mps" if on Apple Silicon
+        attn_implementation="flash_attention_2" if is_flash_attn_2_available() else None,
     ).eval()
 processor = ColQwen2_5_Processor.from_pretrained("vidore/colqwen2.5-v0.1")