hi,

I am trying to score a pymc model on different datasets. Here is how I am generating the trace -

```
with model_factory(df):
trace = pm.sampling.jax.sample_numpyro_nuts(
draws=1000,
tune=1000,
chains=4,
random_seed=1111,
target_accept=0.99
)
```

I need to sum up all the values of obs variable after taking mean. Here is how I am feeding the test dataset. Is this the right way to get predictions on test datasets

```
with model_factory(df):
pm.set_data({"input_data": test_df})
trace_new = pm.sample_posterior_predictive(trace)
trace_new.posterior_predictive['obs'].mean(dim=["chain", "draw"]).sum().values
```