Embed multiple sequences with ESMC

Use this developer guide after completing Run your first ESMC inference. The SDK parallel executor coordinates concurrent requests while respecting the service’s request limits.

Before you begin

Prepare a bounded list of valid amino-acid sequences and a Biohub Platform API key.

Create the client and embedding function

from getpass import getpass

from esm.sdk import esmc_client, parallel_executor
from esm.sdk.api import ESMCInferenceClient, ESMProtein, LogitsConfig, LogitsOutput

token = getpass("Biohub API key: ")
model = esmc_client(
    model="esmc-300m-2024-12",
    url="https://biohub.ai",
    token=token,
)

embedding_config = LogitsConfig(
    sequence=True,
    return_hidden_states=False,
    return_mean_hidden_states=True,
)

def embed_sequence(model: ESMCInferenceClient, sequence: str) -> LogitsOutput:
    protein_tensor = model.encode(ESMProtein(sequence=sequence))
    return model.logits(protein_tensor, embedding_config)

Submit the sequence collection

sequences = [
    "MQIFVKTLTGKTITLEVEPSDTIENVKAKIQDKEGIPPDQQRLIFAGKQLEDGRTLSDYNIQKESTLHLVLRLRGG",
    "MKTIIALSYIFCLVFADYKDDDDK",
]

with parallel_executor() as executor:
    outputs = executor.execute_batch(
        user_func=embed_sequence,
        model=model,
        sequence=sequences,
    )

Verify every output

for index, output in enumerate(outputs, start=1):
    embedding = output.mean_hidden_state.float().squeeze()
    print(index, embedding.shape)

Confirm that the number of outputs matches the number of input sequences and that every representation is non-empty.

For a complete dataset example, open the ESMC embedding notebook.

​Before you begin

Before you begin