Update README.md

README.md

@@ -1,3 +1,13 @@
 ---
 license: apache-2.0
 ---
+
+Install Zyphra's fork of mamba-ssm: https://github.com/Zyphra/mamba
+
+Then one should be able to load any iteration using the following snippet (say iteration 10,000):
+```
+from mamba_ssm.models.mixer_seq_simple import MambaLMHeadModel
+model = MambaLMHeadModel.from_pretrained("Zyphra/Mamba-370M", iteration=10_000, device="cuda")
+```
+
+If iteration is not specified, then the model from the root of the repository is loaded, which is the final iteration (610,351).