Skip to main content
Until Repoch Middleware is available, you can integrate Repoch into your own training pipeline by pushing trained models to the Repoch server.

​
Push Model

After training, push your model to Repoch so it can be pulled later for evaluation.
1

Import and create client

from repoch.client_api import RepochClient

client = RepochClient()
2

Upload model files

upload = client.models.upload_model(
    model_name="my_policy_v1",
    file_paths=["outputs/model.pt", "outputs/config.json"],
    tags=["act", "pick-and-place"],
    model_format="PyTorch",
)
3

Wait for upload

Uploads run in the background by default. Wait for completion if needed:
result = upload.wait()

​
Next Steps

Evaluate Model

Test your trained model on real hardware