This message was deleted.

Hi, you can use the

custom_objects

field in the save model.

bentoml.pytorch.save_model(
    "demo_mnist",   # model name in the local model store
    trained_model,  # model instance being saved
    custom_objects={    # save additional user-defined python objects
        "tokenizer": tokenizer_object,
    }
)

And get the custom object with

model.custom_objects[]

.

import bentoml
bento_model: bentoml.Model = bentoml.models.get("iris_clf:latest")
print(bento_model.custom_objects)

Read more here. https://docs.bentoml.org/en/latest/concepts/model.html

Thanks! Can

custom_objects

field be used to put the binary fine into the model as is? As far as I understand, I'd have to first read the file content into python as e.g. bytes object and then it will be saved as pickle, so the actual file in the model folder will be different as pickle adds its header I guess.

If it is a file you’d like to include, you can add it through the

includes

field in the

bentofile.yaml

. It will get packaged into the bento. You can load the file using Python.

Nice trick with

includes

, thank you!