Where does the `EvaluationLogger` log the `eval_attributes` metadata?

[DevonPeroutky]

I'm instantiating a EvaluationLogger via:

eval_logger = EvaluationLogger(
    name=eval_name, model=model_name, dataset=dataset_name, eval_attributes={"key1": "value1", ...}
)

However, I don't see anything I pass to the eval_attributes anywhere in the weave UI. Where can I access this metadata? Or is it only exposed via API requests currently.

On the same lines, is there any way to pass metadata when logging a prediction? Something like:

result_logger = eval_logger.log_prediction(inputs={"prompt": sample.task}, output=sample.completion, metadata=????)

The primary usecase of the EvaluationLogger seems to be that the client can run predictions with their own harness and then publish the results to Weave, however, without being able to pass any metadata or set the latency/cost/etc of individual predictions, all the predictions are calculated by the weave library and are effectively meaningless because all the predictions happened already.