Update README.md

README.md

@@ -41,6 +41,12 @@ It is available in the following sizes:
 
 You can use these models directly with the `transformers` library. Since ModernBERT is a Masked Language Model (MLM), you can use the `fill-mask` pipeline or load it via `AutoModelForMaskedLM`.
 
+**⚠️ We strongly suggest using ModernBERT with Flash Attention 2, as it is by far the best performing variant of the model, and is a 1:1 match of our research implementation. To do so, install Flash Attention as follows, then use the model as normal:**
+
+```bash
+pip install flash-attn
+```
+
 Using `AutoModelForMaskedLM`:
 
 ```python