Possible conflicting keywords in sample_posterior_predictive

[davidsuculum]

In PyMC discourse, I asked this this, and I reproduce here my code for convenience:

X = np.array([32, 38, 94, 83, 99, 78])
y = np.array([1702, 1514, 683, 269, 900, 86])
y = (y - np.mean(y)) / np.std(y)
X = (X - np.mean(X))/np.std(X)

with pm.Model() as d_model:
    # Specify the covariance function.
    vert = 1*pm.HalfNormal("vert", sigma=1)
    l = pm.HalfNormal(name='l', sigma=1)
    cov_func = vert**2 * pm.gp.cov.Matern32(1, ls=l)

    # Specify the GP. The default mean function is zero.
    #gp = pm.gp.Marginal(mean_init, cov_func=cov_func)
    gp = pm.gp.Marginal(cov_func=cov_func)

    # Place a GP prior over the function f and do the noise:
    sigma = pm.HalfNormal("sigma", sigma=1)
    y_ = gp.marginal_likelihood("y_", X=X, y=y.flatten(), noise=sigma)

    # MCMC:
    trace = pm.sample(10000, chains=3, tune=1000, target_accept=0.99)

X_new = np.linspace(-1.5, 1.5, 500)
X_new = X_new.reshape(-1,1)
with d_model:
    f_p = gp.conditional("f_p", X_new)
    pred_samples = pm.sample_posterior_predictive(trace, samples=10, keep_size=False) 

And I obtain the error

ValueError: conflicting sizes for dimension 'chain': length 1 on the data but length 3 on coordinate 'chain'

In the PyMC discourse link it was suggested to not use samples and keep_size, which effectively solved the issue. It was also suggested that samples in sample_posterior_predicitive is going to be deprecated. Anyway, here it is, as an issue in github, as it was suggested to me.

Versions and main components

• PyMC/PyMC3 Version: PyMC 4.0.1
• Aesara/Theano Version: aesara=2.7.3
• Python Version: 3.7.4
• Operating system: Linux CentOS 7
• How did you install PyMC/PyMC3: (conda/pip) admin installation

[ricardoV94]

This should be solved by #6029

[michaelosthege]

Closed by #6029