Update README.md

README.md

@@ -63,8 +63,8 @@ sae_acts = sae_model.get_acts(esm_layer_acts) # (L+2, SAE_DIM)
 sae_acts
 ```
 
-## Note on the default checkpoint on [https://interprot.com](interprot.com)
+## Note on the default checkpoint on [interprot.com](https://interprot.com)
 
 In Novermber 2024, we shared an earlier version of our layer 24 SAE on [X](https://x.com/liambai21/status/1852765669080879108?s=46) and got a lot of amazing community support in identifying SAE features; therefore, we have kept it as the default on [interprot.com](interprot.com). Since then, we retrained the layer 24 SAE with slightly different hyperparameters and on more sequences (1M vs. the original 100K). The new SAE is named `esm2_plm1280_l24_sae4096.safetensors` whereas the original is named `esm2_plm1280_l24_sae4096_100k.safetensors`.
 
-We recommend using `esm2_plm1280_l24_sae4096.safetensors`, but if you'd like to reproduce the default SAE on [https://interprot.com](interprot.com), you can use `esm2_plm1280_l24_sae4096_100k.safetensors`. All other layer SAEs are trained with the same configrations as `esm2_plm1280_l24_sae4096.safetensors`.
+We recommend using `esm2_plm1280_l24_sae4096.safetensors`, but if you'd like to reproduce the default SAE on [interprot.com](https://interprot.com), you can use `esm2_plm1280_l24_sae4096_100k.safetensors`. All other layer SAEs are trained with the same configrations as `esm2_plm1280_l24_sae4096.safetensors`.