I've experience the same error. The issue is that the model uri is not correct.

The model is logged with:

mlflow.transformers.log_model( ... , artifact_path="gpt2", ...)

The artifact_path is the last part of the model uri. If you don't specify it, it's "model" by default. Since you're specifying it explicitly here, the correct model uri is:

runs:/<run_id>/gpt2