Document ESM-C embedding requirement

README.md

@@ -35,6 +35,24 @@ Single protein-glycan pair or batch CSV. Glycans should be WURCS strings. Protei
 
 Affinose expects per-residue ESM-C 300M embeddings with shape `[L, 960]`. Do not use mean-pooled protein embeddings as model input.
 
+ESM-C is a separate EvolutionaryScale protein model. The ESM-C weights are not included in this repository. Users should install the `esm` package and let it download ESM-C 300M from Hugging Face/EvolutionaryScale into their own runtime cache.
+
+```python
+from esm.models.esmc import ESMC
+from esm.sdk.api import ESMProtein, LogitsConfig
+
+model = ESMC.from_pretrained("esmc_300m").to("cuda")  # or "cpu"
+protein = ESMProtein(sequence="MKTAYIAKQRQISFVKSHFSRQLEERLGLIEVQ")
+protein_tensor = model.encode(protein)
+out = model.logits(
+    protein_tensor,
+    LogitsConfig(sequence=True, return_embeddings=True),
+)
+protein_embeddings = out.embeddings  # per-residue ESM-C 300M embeddings
+```
+
+If Hugging Face requests authentication for ESM-C, users should authenticate with their own Hugging Face account/token and accept any required EvolutionaryScale terms. Bertose/Affinose tokens are not required once these repositories are public.
+
 ## Output
 
 A scalar protein-glycan score from the trained Affinose head.
@@ -42,3 +60,8 @@ A scalar protein-glycan score from the trained Affinose head.
 ## Draft Notes
 
 The public model card still needs final license, citation, intended-use, and limitation wording.
+
+## References
+
+- EvolutionaryScale ESM package: https://github.com/evolutionaryscale/esm
+- ESM-C 300M Hugging Face model: https://huggingface.co/EvolutionaryScale/esmc-300m-2024-12