def get_inference_data3(self, data, eight_schools_params):
"""Read with observed Tensor var_names and dims."""
import tensorflow as tf
if int(tf.__version__[0]) > 1:
import tensorflow.compat.v1 as tf # pylint: disable=import-error
tf.disable_v2_behavior()
inference_data = from_tfp(
data.obj,
var_names=["mu", "tau", "eta"],
model_fn=lambda: data.model(
eight_schools_params["J"], eight_schools_params["sigma"].astype(np.float32)
),
posterior_predictive_samples=100,
posterior_predictive_size=3,
observed=tf.convert_to_tensor(
np.vstack(
(
eight_schools_params["y"],
eight_schools_params["y"],
eight_schools_params["y"],
)
).astype(np.float32),
np.float32,
),
coords={"school": np.arange(eight_schools_params["J"])},
dims={"eta": ["school"], "obs": ["size_dim", "school"]},
)
return inference_data
def get_inference_data4(self, data, eight_schools_params):
"""Test setter."""
inference_data = from_tfp(
data.obj + [np.ones_like(data.obj[0]).astype(np.float32)],
var_names=["mu", "tau", "eta", "avg_effect"],
model_fn=lambda: data.model(
eight_schools_params["J"], eight_schools_params["sigma"].astype(np.float32)
),
observed=eight_schools_params["y"].astype(np.float32),
)
return inference_data
def get_inference_data(self, data, eight_schools_params):
"""Normal read with observed and var_names."""
inference_data = from_tfp(
data.obj,
var_names=["mu", "tau", "eta"],
model_fn=lambda: data.model(
eight_schools_params["J"], eight_schools_params["sigma"].astype(np.float32)
),
observed=eight_schools_params["y"].astype(np.float32),
)
return inference_data
def get_inference_data2(self, data):
"""Fit only."""
inference_data = from_tfp(data.obj)
return inference_data
