Exponential CDF gradient is failing

[AlexAndorra]

Description of your problem

The gradient of the Exponential distribution in PyMC is failing because it contains missing or infinite values.
If you use the new find_constrained_prior function for instance:

pm.find_constrained_prior(pm.Exponential, lower=0, upper=1, init_guess={"lam": 1})

you get a ValueError: array must not contain infs or NaNs

Complete error traceback

ValueError                                Traceback (most recent call last)
<ipython-input-3-1eea11105ccf> in <module>
----> 1 pm.find_constrained_prior(pm.Exponential, lower=0, upper=1, init_guess={"lam": 1})

~/tptm_alex/pymc/pymc/func_utils.py in find_constrained_prior(distribution, lower, upper, init_guess, mass, fixed_params)
    112         jac = "2-point"
    113
--> 114     opt = optimize.least_squares(cdf_error_fn, x0=list(init_guess.values()), jac=jac)
    115     if not opt.success:
    116         raise ValueError("Optimization of parameters failed.")

~/opt/anaconda3/envs/pymc-dev-py39/lib/python3.9/site-packages/scipy/optimize/_lsq/least_squares.py in least_squares(fun, x0, jac, bounds, method, ftol, xtol, gtol, x_scale, loss, f_scale, diff_step, tr_solver, tr_options, jac_sparsity, max_nfev, verbose, args, kwargs)
    926
    927     elif method == 'trf':
--> 928         result = trf(fun_wrapped, jac_wrapped, x0, f0, J0, lb, ub, ftol, xtol,
    929                      gtol, max_nfev, x_scale, loss_function, tr_solver,
    930                      tr_options.copy(), verbose)

~/opt/anaconda3/envs/pymc-dev-py39/lib/python3.9/site-packages/scipy/optimize/_lsq/trf.py in trf(fun, jac, x0, f0, J0, lb, ub, ftol, xtol, gtol, max_nfev, x_scale, loss_function, tr_solver, tr_options, verbose)
    117     # functions are kept the most readable.
    118     if np.all(lb == -np.inf) and np.all(ub == np.inf):
--> 119         return trf_no_bounds(
    120             fun, jac, x0, f0, J0, ftol, xtol, gtol, max_nfev, x_scale,
    121             loss_function, tr_solver, tr_options, verbose)

~/opt/anaconda3/envs/pymc-dev-py39/lib/python3.9/site-packages/scipy/optimize/_lsq/trf.py in trf_no_bounds(fun, jac, x0, f0, J0, ftol, xtol, gtol, max_nfev, x_scale, loss_function, tr_solver, tr_options, verbose)
    465         if tr_solver == 'exact':
    466             J_h = J * d
--> 467             U, s, V = svd(J_h, full_matrices=False)
    468             V = V.T
    469             uf = U.T.dot(f)

~/opt/anaconda3/envs/pymc-dev-py39/lib/python3.9/site-packages/scipy/linalg/decomp_svd.py in svd(a, full_matrices, compute_uv, overwrite_a, check_finite, lapack_driver)
    106
    107     """
--> 108     a1 = _asarray_validated(a, check_finite=check_finite)
    109     if len(a1.shape) != 2:
    110         raise ValueError('expected matrix')

~/opt/anaconda3/envs/pymc-dev-py39/lib/python3.9/site-packages/scipy/_lib/_util.py in _asarray_validated(a, check_finite, sparse_ok, objects_ok, mask_ok, as_inexact)
    291             raise ValueError('masked arrays are not supported')
    292     toarray = np.asarray_chkfinite if check_finite else np.asarray
--> 293     a = toarray(a)
    294     if not objects_ok:
    295         if a.dtype is np.dtype('O'):

~/opt/anaconda3/envs/pymc-dev-py39/lib/python3.9/site-packages/numpy/lib/function_base.py in asarray_chkfinite(a, dtype, order)
    486     a = asarray(a, dtype=dtype, order=order)
    487     if a.dtype.char in typecodes['AllFloat'] and not np.isfinite(a).all():
--> 488         raise ValueError(
    489             "array must not contain infs or NaNs")
    490     return a

ValueError: array must not contain infs or NaNs

Versions and main components

• PyMC/PyMC3 Version: dev version
• Aesara/Theano Version: 2.3.3
• Python Version: 3.9.7
• Operating system: MacOS
• How did you install PyMC/PyMC3: local install

[ricardoV94]

This should be fixed by aesara-devs/aesara#725