Fix deprecation warnings

pytensor_distributions/asymmetriclaplace.py

@@ -111,7 +111,9 @@ def isf(x, mu, b, kappa):
 
 
 def rvs(mu, b, kappa, size=None, random_state=None):
-    random_samples = pt.random.uniform(-kappa, 1 / kappa, size=size, rng=random_state)
+    random_samples = pt.random.uniform(
+        -kappa, 1 / kappa, size=size, rng=random_state, return_next_rng=True
+    )[1]
     sgn = pt.sign(random_samples)
     return mu - (1 / (1 / b * sgn * pt.power(kappa, sgn))) * pt.log(
         1 - random_samples * sgn * pt.power(kappa, sgn)

pytensor_distributions/bernoulli.py

@@ -1,5 +1,5 @@
 import pytensor.tensor as pt
-from pytensor.tensor.xlogx import xlogx
+from pytensor.tensor.special import xlogy
 
 from pytensor_distributions.helper import cdf_bounds, ppf_bounds_disc
 
@@ -51,11 +51,11 @@ def lmoment4(p):
 
 def entropy(p):
     q = 1 - p
-    return -xlogx(p) - xlogx(q)
+    return -xlogy(p, p) - xlogy(q, q)
 
 
 def rvs(p, size=None, random_state=None):
-    return pt.random.binomial(1, p, size=size, rng=random_state)
+    return pt.random.binomial(1, p, size=size, rng=random_state, return_next_rng=True)[1]
 
 
 def cdf(x, p):

pytensor_distributions/beta.py

@@ -100,7 +100,7 @@ def sf(x, alpha, beta):
 
 
 def rvs(alpha, beta, size=None, random_state=None):
-    return pt.random.beta(alpha, beta, rng=random_state, size=size)
+    return pt.random.beta(alpha, beta, rng=random_state, size=size, return_next_rng=True)[1]
 
 
 def logcdf(x, alpha, beta):

pytensor_distributions/betabinomial.py

@@ -126,7 +126,7 @@ def isf(x, n, alpha, beta):
 
 
 def rvs(n, alpha, beta, size=None, random_state=None):
-    return pt.random.betabinom(n, alpha, beta, size=size, rng=random_state)
+    return pt.random.betabinom(n, alpha, beta, size=size, rng=random_state, return_next_rng=True)[1]
 
 
 def logcdf(x, n, alpha, beta):

pytensor_distributions/betascaled.py

@@ -106,7 +106,7 @@ def sf(x, alpha, beta, lower, upper):
 
 
 def rvs(alpha, beta, lower, upper, size=None, random_state=None):
-    beta_samples = pt.random.beta(alpha, beta, rng=random_state, size=size)
+    beta_samples = pt.random.beta(alpha, beta, rng=random_state, size=size, return_next_rng=True)[1]
     return beta_samples * (upper - lower) + lower
 
 

pytensor_distributions/binomial.py

@@ -77,7 +77,7 @@ def isf(q, n, p):
 
 
 def rvs(n, p, size=None, random_state=None):
-    return pt.random.binomial(n, p, size=size, rng=random_state)
+    return pt.random.binomial(n, p, size=size, rng=random_state, return_next_rng=True)[1]
 
 
 def logpdf(x, n, p):

pytensor_distributions/categorical.py

@@ -1,5 +1,5 @@
 import pytensor.tensor as pt
-from pytensor.tensor.xlogx import xlogx
+from pytensor.tensor.special import xlogy
 
 from pytensor_distributions.helper import cdf_bounds, ppf_bounds_disc
 
@@ -54,7 +54,7 @@ def kurtosis(p):
 
 def entropy(p):
     p = _normalize_p(p)
-    return -pt.sum(xlogx(p), axis=-1)
+    return -pt.sum(xlogy(p, p), axis=-1)
 
 
 def pdf(x, p):
@@ -114,4 +114,4 @@ def isf(q, p):
 
 def rvs(p, size=None, random_state=None):
     p = _normalize_p(p)
-    return pt.random.categorical(p, size=size, rng=random_state)
+    return pt.random.categorical(p, size=size, rng=random_state, return_next_rng=True)[1]

pytensor_distributions/cauchy.py

@@ -82,7 +82,7 @@ def sf(x, alpha, beta):
 
 
 def rvs(alpha, beta, size=None, random_state=None):
-    return pt.random.cauchy(alpha, beta, rng=random_state, size=size)
+    return pt.random.cauchy(alpha, beta, rng=random_state, size=size, return_next_rng=True)[1]
 
 
 def logcdf(x, alpha, beta):

pytensor_distributions/chisquared.py

@@ -68,7 +68,7 @@ def pdf(x, nu):
 
 
 def rvs(nu, size=None, random_state=None):
-    return pt.random.chisquare(nu, rng=random_state, size=size)
+    return pt.random.chisquare(nu, rng=random_state, size=size, return_next_rng=True)[1]
 
 
 def logcdf(x, nu):

pytensor_distributions/dirichlet.py

@@ -73,4 +73,4 @@ def logpdf(x, alpha):
 
 
 def rvs(alpha, size=None, random_state=None):
-    return pt.random.dirichlet(alpha, size=size, rng=random_state)
+    return pt.random.dirichlet(alpha, size=size, rng=random_state, return_next_rng=True)[1]

pytensor_distributions/dirichletmultinomial.py

@@ -75,5 +75,5 @@ def logpdf(x, n, a):
 
 
 def rvs(n, a, size=None, random_state=None):
-    p = pt.random.dirichlet(a, size=size, rng=random_state)
-    return pt.random.multinomial(n, p, rng=random_state)
+    next_rng, p = pt.random.dirichlet(a, size=size, rng=random_state, return_next_rng=True)
+    return pt.random.multinomial(n, p, rng=next_rng, return_next_rng=True)[1]

pytensor_distributions/discreteuniform.py

@@ -81,7 +81,9 @@ def isf(q, lower, upper):
 
 
 def rvs(lower, upper, size=None, random_state=None):
-    return pt.random.integers(lower, upper + 1, size=size, rng=random_state)
+    return pt.random.integers(lower, upper + 1, size=size, rng=random_state, return_next_rng=True)[
+        1
+    ]
 
 
 def logpdf(x, lower, upper):

pytensor_distributions/discreteweibull.py

@@ -76,7 +76,9 @@ def sf(x, q, beta):
 
 
 def rvs(q, beta, size=None, random_state=None):
-    return ppf(pt.random.uniform(0, 1, rng=random_state, size=size), q, beta)
+    return ppf(
+        pt.random.uniform(0, 1, rng=random_state, size=size, return_next_rng=True)[1], q, beta
+    )
 
 
 def logcdf(x, q, beta):

pytensor_distributions/exgaussian.py

@@ -88,8 +88,8 @@ def sf(x, mu, sigma, nu):
 
 
 def rvs(mu, sigma, nu, size=None, random_state=None):
-    next_rng, exp_rvs = pt.random.exponential(nu, rng=random_state, size=size).owner.outputs
-    normal_rvs = pt.random.normal(mu, sigma, rng=next_rng, size=size)
+    next_rng, exp_rvs = pt.random.exponential(nu, rng=random_state, size=size, return_next_rng=True)
+    normal_rvs = pt.random.normal(mu, sigma, rng=next_rng, size=size, return_next_rng=True)[1]
     return normal_rvs + exp_rvs
 
 

pytensor_distributions/exponential.py

@@ -80,7 +80,7 @@ def sf(x, lam):
 
 
 def rvs(lam, size=None, random_state=None):
-    return pt.random.exponential(1.0 / lam, rng=random_state, size=size)
+    return pt.random.exponential(1.0 / lam, rng=random_state, size=size, return_next_rng=True)[1]
 
 
 def logcdf(x, lam):

pytensor_distributions/gamma.py

@@ -76,7 +76,9 @@ def isf(x, alpha, beta):
 
 
 def rvs(alpha, beta, size=None, random_state=None):
-    return pt.random.gamma(shape=alpha, scale=1 / beta, rng=random_state, size=size)
+    return pt.random.gamma(
+        shape=alpha, scale=1 / beta, rng=random_state, size=size, return_next_rng=True
+    )[1]
 
 
 def logcdf(x, alpha, beta):

pytensor_distributions/geometric.py

@@ -78,7 +78,7 @@ def isf(q, p):
 
 
 def rvs(p, size=None, random_state=None):
-    return pt.random.geometric(p, size=size, rng=random_state)
+    return pt.random.geometric(p, size=size, rng=random_state, return_next_rng=True)[1]
 
 
 def logpdf(x, p):

pytensor_distributions/gumbel.py

@@ -86,7 +86,9 @@ def sf(x, mu, beta):
 
 
 def rvs(mu, beta, size=None, random_state=None):
-    return ppf(pt.random.uniform(0, 1, rng=random_state, size=size), mu, beta)
+    return ppf(
+        pt.random.uniform(0, 1, rng=random_state, size=size, return_next_rng=True)[1], mu, beta
+    )
 
 
 def logcdf(x, mu, beta):

pytensor_distributions/halfcauchy.py

@@ -81,7 +81,7 @@ def sf(x, beta):
 
 
 def rvs(beta, size=None, random_state=None):
-    uniform_samples = pt.random.uniform(0, 1, rng=random_state, size=size)
+    uniform_samples = pt.random.uniform(0, 1, rng=random_state, size=size, return_next_rng=True)[1]
     return beta * pt.tan(pt.pi / 2 * uniform_samples)
 
 

pytensor_distributions/halfnormal.py

@@ -79,7 +79,7 @@ def sf(x, sigma):
 
 
 def rvs(sigma, size=None, random_state=None):
-    return pt.abs(pt.random.normal(0, sigma, rng=random_state, size=size))
+    return pt.abs(pt.random.normal(0, sigma, rng=random_state, size=size, return_next_rng=True)[1])
 
 
 def logpdf(x, sigma):

pytensor_distributions/halfstudentt.py

@@ -123,7 +123,7 @@ def sf(x, nu, sigma):
 
 
 def rvs(nu, sigma, size=None, random_state=None):
-    t_samples = pt.random.t(nu, rng=random_state, size=size)
+    t_samples = pt.random.t(nu, rng=random_state, size=size, return_next_rng=True)[1]
     return pt.abs(t_samples * sigma)
 
 

pytensor_distributions/hypergeometric.py

@@ -130,4 +130,6 @@ def isf(q, N, k, n):
 
 
 def rvs(N, k, n, size=None, random_state=None):
-    return pt.random.hypergeometric(k, N - k, n, size=size, rng=random_state)
+    return pt.random.hypergeometric(k, N - k, n, size=size, rng=random_state, return_next_rng=True)[
+        1
+    ]

pytensor_distributions/inversegamma.py

@@ -83,7 +83,7 @@ def sf(x, alpha, beta):
 
 
 def rvs(alpha, beta, size=None, random_state=None):
-    return 1.0 / pt.random.gamma(alpha, beta, rng=random_state, size=size)
+    return 1.0 / pt.random.gamma(alpha, beta, rng=random_state, size=size, return_next_rng=True)[1]
 
 
 def logcdf(x, alpha, beta):

pytensor_distributions/kumaraswamy.py

@@ -103,7 +103,7 @@ def isf(x, a, b):
 
 
 def rvs(a, b, size=None, random_state=None):
-    u = pt.random.uniform(0, 1, size=size, rng=random_state)
+    u = pt.random.uniform(0, 1, size=size, rng=random_state, return_next_rng=True)[1]
     return ppf(u, a, b)
 
 

pytensor_distributions/laplace.py

@@ -85,7 +85,7 @@ def isf(q, mu, b):
 
 
 def rvs(mu, b, size=None, random_state=None):
-    return pt.random.laplace(mu, b, rng=random_state, size=size)
+    return pt.random.laplace(mu, b, rng=random_state, size=size, return_next_rng=True)[1]
 
 
 def logpdf(x, mu, b):

pytensor_distributions/logistic.py

@@ -83,7 +83,7 @@ def sf(x, mu, s):
 
 
 def rvs(mu, s, size=None, random_state=None):
-    return pt.random.logistic(mu, s, rng=random_state, size=size)
+    return pt.random.logistic(mu, s, rng=random_state, size=size, return_next_rng=True)[1]
 
 
 def logcdf(x, mu, s):

pytensor_distributions/logitnormal.py

@@ -204,4 +204,6 @@ def isf(q, mu, sigma):
 
 
 def rvs(mu, sigma, size=None, random_state=None):
-    return pt.sigmoid(pt.random.normal(mu, sigma, rng=random_state, size=size))
+    return pt.sigmoid(
+        pt.random.normal(mu, sigma, rng=random_state, size=size, return_next_rng=True)[1]
+    )

pytensor_distributions/loglogistic.py

@@ -133,5 +133,5 @@ def isf(q, alpha, beta):
 
 
 def rvs(alpha, beta, size=None, random_state=None):
-    u = pt.random.uniform(size=size, rng=random_state)
+    u = pt.random.uniform(size=size, rng=random_state, return_next_rng=True)[1]
     return alpha * (u / (1 - u)) ** (1 / beta)

pytensor_distributions/lognormal.py

@@ -81,7 +81,7 @@ def sf(x, mu, sigma):
 
 
 def rvs(mu, sigma, size=None, random_state=None):
-    return pt.exp(pt.random.normal(mu, sigma, rng=random_state, size=size))
+    return pt.exp(pt.random.normal(mu, sigma, rng=random_state, size=size, return_next_rng=True)[1])
 
 
 def logcdf(x, mu, sigma):

pytensor_distributions/matrixnormal.py

@@ -120,5 +120,5 @@ def rvs(mu, rowcov, colcov, size=None, random_state=None):
     else:
         full_shape = base_shape
 
-    Z = pt.random.normal(0, 1, size=full_shape, rng=random_state)
+    Z = pt.random.normal(0, 1, size=full_shape, rng=random_state, return_next_rng=True)[1]
     return target + L_row @ Z @ pt.swapaxes(L_col, -1, -2)

pytensor_distributions/moyal.py

@@ -80,7 +80,7 @@ def isf(x, mu, sigma):
 
 
 def rvs(mu, sigma, size=None, random_state=None):
-    u = pt.random.uniform(size=size, rng=random_state)
+    u = pt.random.uniform(size=size, rng=random_state, return_next_rng=True)[1]
     return ppf(u, mu, sigma)
 
 

pytensor_distributions/multinomial.py

@@ -55,4 +55,4 @@ def logpdf(x, n, p):
 
 
 def rvs(n, p, size=None, random_state=None):
-    return pt.random.multinomial(n, p, size=size, rng=random_state)
+    return pt.random.multinomial(n, p, size=size, rng=random_state, return_next_rng=True)[1]

pytensor_distributions/mvnormal.py

@@ -85,4 +85,6 @@ def logpdf(x, mu, cov):
 
 def rvs(mu, cov, size=None, random_state=None):
     mu = pt.broadcast_to(mu, cov.shape[:-1])
-    return pt.random.multivariate_normal(mu, cov, size=size, rng=random_state)
+    return pt.random.multivariate_normal(
+        mu, cov, size=size, rng=random_state, return_next_rng=True
+    )[1]

pytensor_distributions/mvstudentt.py

@@ -70,8 +70,12 @@ def logpdf(x, nu, mu, cov):
 
 def rvs(nu, mu, cov, size=None, random_state=None):
     mu = pt.broadcast_to(mu, cov.shape[:-1])
-    z = pt.random.multivariate_normal(pt.zeros_like(mu), cov, size=size, rng=random_state)
-    chi2 = pt.random.chisquare(nu, size=size if size is not None else 1, rng=random_state)
+    z = pt.random.multivariate_normal(
+        pt.zeros_like(mu), cov, size=size, rng=random_state, return_next_rng=True
+    )[1]
+    chi2 = pt.random.chisquare(
+        nu, size=size if size is not None else 1, rng=random_state, return_next_rng=True
+    )[1]
     if size is None:
         chi2 = chi2[None]
     return mu + z / pt.sqrt(chi2[..., None] / nu)

pytensor_distributions/negativebinomial.py

@@ -81,7 +81,7 @@ def isf(q, n, p):
 
 
 def rvs(n, p, size=None, random_state=None):
-    return pt.random.negative_binomial(n, p, rng=random_state, size=size)
+    return pt.random.negative_binomial(n, p, rng=random_state, size=size, return_next_rng=True)[1]
 
 
 def logpdf(x, n, p):

pytensor_distributions/normal.py

@@ -85,7 +85,7 @@ def sf(x, mu, sigma):
 
 
 def rvs(mu, sigma, size=None, random_state=None):
-    return pt.random.normal(mu, sigma, rng=random_state, size=size)
+    return pt.random.normal(mu, sigma, rng=random_state, size=size, return_next_rng=True)[1]
 
 
 def logcdf(x, mu, sigma):

pytensor_distributions/pareto.py

@@ -86,7 +86,7 @@ def isf(x, alpha, m):
 
 
 def rvs(alpha, m, size=None, random_state=None):
-    u = pt.random.uniform(size=size, rng=random_state)
+    u = pt.random.uniform(size=size, rng=random_state, return_next_rng=True)[1]
     return m / (1 - u) ** (1 / alpha)
 
 

pytensor_distributions/poisson.py

@@ -81,7 +81,7 @@ def isf(q, mu):
 
 
 def rvs(mu, size=None, random_state=None):
-    return pt.random.poisson(mu, rng=random_state, size=size)
+    return pt.random.poisson(mu, rng=random_state, size=size, return_next_rng=True)[1]
 
 
 def logpdf(x, mu):

pytensor_distributions/polyagamma.py

@@ -197,7 +197,9 @@ def rvs(h, z, size=None, random_state=None):
     else:
         gamma_size = (K, *size)
 
-    gamma_draws = pt.random.gamma(h, scale=1.0, size=gamma_size, rng=random_state)
+    gamma_draws = pt.random.gamma(
+        h, scale=1.0, size=gamma_size, rng=random_state, return_next_rng=True
+    )[1]
 
     z2_term = z**2 / (4 * pt.pi**2)
     k_bc = k.reshape((-1,) + (1,) * (gamma_draws.ndim - 1))

pytensor_distributions/rice.py

@@ -127,8 +127,8 @@ def isf(q, nu, sigma):
 
 
 def rvs(nu, sigma, size=None, random_state=None):
-    next_rng, x = pt.random.normal(nu, sigma, size=size, rng=random_state).owner.outputs
-    y = pt.random.normal(0, sigma, size=size, rng=next_rng)
+    next_rng, x = pt.random.normal(nu, sigma, size=size, rng=random_state, return_next_rng=True)
+    next_rng, y = pt.random.normal(0, sigma, size=size, rng=next_rng, return_next_rng=True)
     return pt.sqrt(x**2 + y**2)