Pathfinder failure fitting Mauna Loa GP example

[fonnesbeck]

When attempting to fit the Mauna Loa example in pymc-examples:

with pm.Model() as model:
    # yearly periodic component x long term trend
    η_per = pm.HalfCauchy("η_per", beta=2, initval=1.0)
    ℓ_pdecay = pm.Gamma("ℓ_pdecay", alpha=10, beta=0.075)
    period = pm.Normal("period", mu=1, sigma=0.05)
    ℓ_psmooth = pm.Gamma("ℓ_psmooth ", alpha=4, beta=3)
    cov_seasonal = (
        η_per**2 * pm.gp.cov.Periodic(1, period, ℓ_psmooth) * pm.gp.cov.Matern52(1, ℓ_pdecay)
    )
    gp_seasonal = pm.gp.Marginal(cov_func=cov_seasonal)

    # small/medium term irregularities
    η_med = pm.HalfCauchy("η_med", beta=0.5, initval=0.1)
    ℓ_med = pm.Gamma("ℓ_med", alpha=2, beta=0.75)
    α = pm.Gamma("α", alpha=5, beta=2)
    cov_medium = η_med**2 * pm.gp.cov.RatQuad(1, ℓ_med, α)
    gp_medium = pm.gp.Marginal(cov_func=cov_medium)

    # long term trend
    η_trend = pm.HalfCauchy("η_trend", beta=2, initval=2.0)
    ℓ_trend = pm.Gamma("ℓ_trend", alpha=4, beta=0.1)
    cov_trend = η_trend**2 * pm.gp.cov.ExpQuad(1, ℓ_trend)
    gp_trend = pm.gp.Marginal(cov_func=cov_trend)

    # noise model
    η_noise = pm.HalfNormal("η_noise", sigma=0.5, initval=0.05)
    ℓ_noise = pm.Gamma("ℓ_noise", alpha=2, beta=4)
    σ = pm.HalfNormal("σ", sigma=0.25, initval=0.05)
    cov_noise = η_noise**2 * pm.gp.cov.Matern32(1, ℓ_noise) + pm.gp.cov.WhiteNoise(σ)

    # The Gaussian process is a sum of these three components
    gp = gp_seasonal + gp_medium + gp_trend

    # Since the normal noise model and the GP are conjugates, we use `Marginal` with the `.marginal_likelihood` method
    y_ = gp.marginal_likelihood("y", X=t, y=y, noise=cov_noise)

    approx = pmx.fit(method='pathfinder')

It appears to generate rewrite errors:

ERROR (pytensor.graph.rewriting.basic): SequentialGraphRewriter apply <pytensor.tensor.rewriting.elemwise.InplaceElemwiseOptimizer object at 0x7f01855aabd0>
ERROR (pytensor.graph.rewriting.basic): Traceback:
ERROR (pytensor.graph.rewriting.basic): Traceback (most recent call last):
  File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/rewriting/basic.py", line 292, in apply
    sub_prof = rewriter.apply(fgraph)
               ^^^^^^^^^^^^^^^^^^^^^^
  File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/tensor/rewriting/elemwise.py", line 192, in apply
    and not fgraph.has_destroyers([node.inputs[i]])
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 393, in has_destroyers
    droot, _, root_destroyer = self.refresh_droot_impact()
                               ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 431, in refresh_droot_impact
    self.droot, self.impact, self.root_destroyer = _build_droot_impact(self)
                                                   ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 200, in _build_droot_impact
    raise InconsistencyError(f"Multiple destroyers of {input_root}")
pytensor.graph.utils.InconsistencyError: Multiple destroyers of Blockwise{SolveTriangular{unit_diagonal=False, lower=False, check_finite=True, b_ndim=2, overwrite_b=True}, (m,m),(m,n)->(m,n)}.0

ERROR (pytensor.graph.rewriting.basic): SequentialGraphRewriter apply <pytensor.scan.rewriting.ScanInplaceOptimizer object at 0x7f01852667e0>
ERROR (pytensor.graph.rewriting.basic): Traceback:
ERROR (pytensor.graph.rewriting.basic): Traceback (most recent call last):
  File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/rewriting/basic.py", line 292, in apply
    sub_prof = rewriter.apply(fgraph)
               ^^^^^^^^^^^^^^^^^^^^^^
  File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/scan/rewriting.py", line 1035, in apply
    for scan_idx, original_node in enumerate(reversed(fgraph.toposort())):
                                                      ^^^^^^^^^^^^^^^^^
  File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/fg.py", line 760, in toposort
    return io_toposort(self.inputs, self.outputs, self.orderings())
                                                  ^^^^^^^^^^^^^^^^
  File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/fg.py", line 784, in orderings
    orderings = feature.orderings(self)
                ^^^^^^^^^^^^^^^^^^^^^^^
  File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 683, in orderings
    droot, impact, __ignore = self.refresh_droot_impact()
                              ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 431, in refresh_droot_impact
    self.droot, self.impact, self.root_destroyer = _build_droot_impact(self)
                                                   ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 200, in _build_droot_impact
    raise InconsistencyError(f"Multiple destroyers of {input_root}")
pytensor.graph.utils.InconsistencyError: Multiple destroyers of Blockwise{SolveTriangular{unit_diagonal=False, lower=False, check_finite=True, b_ndim=2, overwrite_b=True}, (m,m),(m,n)->(m,n)}.0

pymc_extras: 0.2.4
pymc : 5.22.0
pytensor : 2.30.3

[zaxtax]

Do you still have this problem when building with what's on main?

…

On Tue, 15 Apr 2025, 15:29 Chris Fonnesbeck, ***@***.***> wrote: When attempting to fit the Mauna Loa example in pymc-examples: with pm.Model() as model: # yearly periodic component x long term trend η_per = pm.HalfCauchy("η_per", beta=2, initval=1.0) ℓ_pdecay = pm.Gamma("ℓ_pdecay", alpha=10, beta=0.075) period = pm.Normal("period", mu=1, sigma=0.05) ℓ_psmooth = pm.Gamma("ℓ_psmooth ", alpha=4, beta=3) cov_seasonal = ( η_per**2 * pm.gp.cov.Periodic(1, period, ℓ_psmooth) * pm.gp.cov.Matern52(1, ℓ_pdecay) ) gp_seasonal = pm.gp.Marginal(cov_func=cov_seasonal) # small/medium term irregularities η_med = pm.HalfCauchy("η_med", beta=0.5, initval=0.1) ℓ_med = pm.Gamma("ℓ_med", alpha=2, beta=0.75) α = pm.Gamma("α", alpha=5, beta=2) cov_medium = η_med**2 * pm.gp.cov.RatQuad(1, ℓ_med, α) gp_medium = pm.gp.Marginal(cov_func=cov_medium) # long term trend η_trend = pm.HalfCauchy("η_trend", beta=2, initval=2.0) ℓ_trend = pm.Gamma("ℓ_trend", alpha=4, beta=0.1) cov_trend = η_trend**2 * pm.gp.cov.ExpQuad(1, ℓ_trend) gp_trend = pm.gp.Marginal(cov_func=cov_trend) # noise model η_noise = pm.HalfNormal("η_noise", sigma=0.5, initval=0.05) ℓ_noise = pm.Gamma("ℓ_noise", alpha=2, beta=4) σ = pm.HalfNormal("σ", sigma=0.25, initval=0.05) cov_noise = η_noise**2 * pm.gp.cov.Matern32(1, ℓ_noise) + pm.gp.cov.WhiteNoise(σ) # The Gaussian process is a sum of these three components gp = gp_seasonal + gp_medium + gp_trend # Since the normal noise model and the GP are conjugates, we use `Marginal` with the `.marginal_likelihood` method y_ = gp.marginal_likelihood("y", X=t, y=y, noise=cov_noise) approx = pmx.fit(method='pathfinder') It appears to generate rewrite errors: ERROR (pytensor.graph.rewriting.basic): SequentialGraphRewriter apply <pytensor.tensor.rewriting.elemwise.InplaceElemwiseOptimizer object at 0x7f01855aabd0> ERROR (pytensor.graph.rewriting.basic): Traceback: ERROR (pytensor.graph.rewriting.basic): Traceback (most recent call last): File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/rewriting/basic.py", line 292, in apply sub_prof = rewriter.apply(fgraph) ^^^^^^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/tensor/rewriting/elemwise.py", line 192, in apply and not fgraph.has_destroyers([node.inputs[i]]) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 393, in has_destroyers droot, _, root_destroyer = self.refresh_droot_impact() ^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 431, in refresh_droot_impact self.droot, self.impact, self.root_destroyer = _build_droot_impact(self) ^^^^^^^^^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 200, in _build_droot_impact raise InconsistencyError(f"Multiple destroyers of {input_root}") pytensor.graph.utils.InconsistencyError: Multiple destroyers of Blockwise{SolveTriangular{unit_diagonal=False, lower=False, check_finite=True, b_ndim=2, overwrite_b=True}, (m,m),(m,n)->(m,n)}.0 ERROR (pytensor.graph.rewriting.basic): SequentialGraphRewriter apply <pytensor.scan.rewriting.ScanInplaceOptimizer object at 0x7f01852667e0> ERROR (pytensor.graph.rewriting.basic): Traceback: ERROR (pytensor.graph.rewriting.basic): Traceback (most recent call last): File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/rewriting/basic.py", line 292, in apply sub_prof = rewriter.apply(fgraph) ^^^^^^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/scan/rewriting.py", line 1035, in apply for scan_idx, original_node in enumerate(reversed(fgraph.toposort())): ^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/fg.py", line 760, in toposort return io_toposort(self.inputs, self.outputs, self.orderings()) ^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/fg.py", line 784, in orderings orderings = feature.orderings(self) ^^^^^^^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 683, in orderings droot, impact, __ignore = self.refresh_droot_impact() ^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 431, in refresh_droot_impact self.droot, self.impact, self.root_destroyer = _build_droot_impact(self) ^^^^^^^^^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 200, in _build_droot_impact raise InconsistencyError(f"Multiple destroyers of {input_root}") pytensor.graph.utils.InconsistencyError: Multiple destroyers of Blockwise{SolveTriangular{unit_diagonal=False, lower=False, check_finite=True, b_ndim=2, overwrite_b=True}, (m,m),(m,n)->(m,n)}.0 pymc_extras: 0.2.4 pymc : 5.22.0 pytensor : 2.30.3 — Reply to this email directly, view it on GitHub <#458>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAACCUL7RPCG446ILK52WSD2ZUCT7AVCNFSM6AAAAAB3FUPMEGVHI2DSMVQWIX3LMV43ASLTON2WKOZSHE4TMNBTHE2DMMA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***> *fonnesbeck* created an issue (pymc-devs/pymc-extras#458) <#458> When attempting to fit the Mauna Loa example in pymc-examples: with pm.Model() as model: # yearly periodic component x long term trend η_per = pm.HalfCauchy("η_per", beta=2, initval=1.0) ℓ_pdecay = pm.Gamma("ℓ_pdecay", alpha=10, beta=0.075) period = pm.Normal("period", mu=1, sigma=0.05) ℓ_psmooth = pm.Gamma("ℓ_psmooth ", alpha=4, beta=3) cov_seasonal = ( η_per**2 * pm.gp.cov.Periodic(1, period, ℓ_psmooth) * pm.gp.cov.Matern52(1, ℓ_pdecay) ) gp_seasonal = pm.gp.Marginal(cov_func=cov_seasonal) # small/medium term irregularities η_med = pm.HalfCauchy("η_med", beta=0.5, initval=0.1) ℓ_med = pm.Gamma("ℓ_med", alpha=2, beta=0.75) α = pm.Gamma("α", alpha=5, beta=2) cov_medium = η_med**2 * pm.gp.cov.RatQuad(1, ℓ_med, α) gp_medium = pm.gp.Marginal(cov_func=cov_medium) # long term trend η_trend = pm.HalfCauchy("η_trend", beta=2, initval=2.0) ℓ_trend = pm.Gamma("ℓ_trend", alpha=4, beta=0.1) cov_trend = η_trend**2 * pm.gp.cov.ExpQuad(1, ℓ_trend) gp_trend = pm.gp.Marginal(cov_func=cov_trend) # noise model η_noise = pm.HalfNormal("η_noise", sigma=0.5, initval=0.05) ℓ_noise = pm.Gamma("ℓ_noise", alpha=2, beta=4) σ = pm.HalfNormal("σ", sigma=0.25, initval=0.05) cov_noise = η_noise**2 * pm.gp.cov.Matern32(1, ℓ_noise) + pm.gp.cov.WhiteNoise(σ) # The Gaussian process is a sum of these three components gp = gp_seasonal + gp_medium + gp_trend # Since the normal noise model and the GP are conjugates, we use `Marginal` with the `.marginal_likelihood` method y_ = gp.marginal_likelihood("y", X=t, y=y, noise=cov_noise) approx = pmx.fit(method='pathfinder') It appears to generate rewrite errors: ERROR (pytensor.graph.rewriting.basic): SequentialGraphRewriter apply <pytensor.tensor.rewriting.elemwise.InplaceElemwiseOptimizer object at 0x7f01855aabd0> ERROR (pytensor.graph.rewriting.basic): Traceback: ERROR (pytensor.graph.rewriting.basic): Traceback (most recent call last): File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/rewriting/basic.py", line 292, in apply sub_prof = rewriter.apply(fgraph) ^^^^^^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/tensor/rewriting/elemwise.py", line 192, in apply and not fgraph.has_destroyers([node.inputs[i]]) ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 393, in has_destroyers droot, _, root_destroyer = self.refresh_droot_impact() ^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 431, in refresh_droot_impact self.droot, self.impact, self.root_destroyer = _build_droot_impact(self) ^^^^^^^^^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 200, in _build_droot_impact raise InconsistencyError(f"Multiple destroyers of {input_root}") pytensor.graph.utils.InconsistencyError: Multiple destroyers of Blockwise{SolveTriangular{unit_diagonal=False, lower=False, check_finite=True, b_ndim=2, overwrite_b=True}, (m,m),(m,n)->(m,n)}.0 ERROR (pytensor.graph.rewriting.basic): SequentialGraphRewriter apply <pytensor.scan.rewriting.ScanInplaceOptimizer object at 0x7f01852667e0> ERROR (pytensor.graph.rewriting.basic): Traceback: ERROR (pytensor.graph.rewriting.basic): Traceback (most recent call last): File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/rewriting/basic.py", line 292, in apply sub_prof = rewriter.apply(fgraph) ^^^^^^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/scan/rewriting.py", line 1035, in apply for scan_idx, original_node in enumerate(reversed(fgraph.toposort())): ^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/fg.py", line 760, in toposort return io_toposort(self.inputs, self.outputs, self.orderings()) ^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/fg.py", line 784, in orderings orderings = feature.orderings(self) ^^^^^^^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 683, in orderings droot, impact, __ignore = self.refresh_droot_impact() ^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 431, in refresh_droot_impact self.droot, self.impact, self.root_destroyer = _build_droot_impact(self) ^^^^^^^^^^^^^^^^^^^^^^^^^ File "/var/home/fonnesbeck/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/destroyhandler.py", line 200, in _build_droot_impact raise InconsistencyError(f"Multiple destroyers of {input_root}") pytensor.graph.utils.InconsistencyError: Multiple destroyers of Blockwise{SolveTriangular{unit_diagonal=False, lower=False, check_finite=True, b_ndim=2, overwrite_b=True}, (m,m),(m,n)->(m,n)}.0 pymc_extras: 0.2.4 pymc : 5.22.0 pytensor : 2.30.3 — Reply to this email directly, view it on GitHub <#458>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAACCUL7RPCG446ILK52WSD2ZUCT7AVCNFSM6AAAAAB3FUPMEGVHI2DSMVQWIX3LMV43ASLTON2WKOZSHE4TMNBTHE2DMMA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[aphc14]

Do you still have this problem when building with what's on main?

Getting a different problem now:

/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/jaxopt/_src/osqp.py:299: SyntaxWarning: invalid escape sequence '\m'
  """Operator Splitting Solver for Quadratic Programs.

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/compile/function/types.py:1037, in Function.__call__(self, output_subset, *args, **kwargs)
   1036 try:
-> 1037     outputs = vm() if output_subset is None else vm(output_subset=output_subset)
   1038 except Exception:

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/op.py:531, in Op.make_py_thunk.<locals>.rval(p, i, o, n, cm)
    523 @is_thunk_type
    524 def rval(
    525     p=p,
   (...)    529     cm=node_compute_map,
    530 ):
--> 531     r = p(n, [x[0] for x in i], o)
    532     for entry in cm:

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/tensor/slinalg.py:72, in Cholesky.perform(self, node, inputs, outputs)
     71 if self.overwrite_a and x.flags["C_CONTIGUOUS"]:
---> 72     out[0] = scipy_linalg.cholesky(
     73         x.T,
     74         lower=not self.lower,
     75         check_finite=self.check_finite,
     76         overwrite_a=True,
     77     ).T
     78 else:

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/scipy/linalg/_decomp_cholesky.py:101, in cholesky(a, lower, overwrite_a, check_finite)
     47 """
     48 Compute the Cholesky decomposition of a matrix.
     49 
   (...)     99 
    100 """
--> 101 c, lower = _cholesky(a, lower=lower, overwrite_a=overwrite_a, clean=True,
    102                      check_finite=check_finite)
    103 return c

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/scipy/linalg/_decomp_cholesky.py:18, in _cholesky(a, lower, overwrite_a, clean, check_finite)
     16 """Common code for cholesky() and cho_factor()."""
---> 18 a1 = asarray_chkfinite(a) if check_finite else asarray(a)
     19 a1 = atleast_2d(a1)

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/numpy/lib/function_base.py:630, in asarray_chkfinite(a, dtype, order)
    629 if a.dtype.char in typecodes['AllFloat'] and not np.isfinite(a).all():
--> 630     raise ValueError(
    631         "array must not contain infs or NaNs")
    632 return a

ValueError: array must not contain infs or NaNs

During handling of the above exception, another exception occurred:

ValueError                                Traceback (most recent call last)
File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/compile/function/types.py:1037, in Function.__call__(self, output_subset, *args, **kwargs)
   1036 try:
-> 1037     outputs = vm() if output_subset is None else vm(output_subset=output_subset)
   1038 except Exception:

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/op.py:531, in Op.make_py_thunk.<locals>.rval(p, i, o, n, cm)
    523 @is_thunk_type
    524 def rval(
    525     p=p,
   (...)    529     cm=node_compute_map,
    530 ):
--> 531     r = p(n, [x[0] for x in i], o)
    532     for entry in cm:

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py:712, in LogLike.perform(self, node, inputs, outputs)
    711 phi = inputs[0]
--> 712 logP = np.apply_along_axis(self.logp_func, axis=-1, arr=phi)
    713 # replace nan with -inf since np.argmax will return the first index at nan

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/numpy/lib/shape_base.py:402, in apply_along_axis(func1d, axis, arr, *args, **kwargs)
    401 for ind in inds:
--> 402     buff[ind] = asanyarray(func1d(inarr_view[ind], *args, **kwargs))
    404 if not isinstance(res, matrix):
    405     # wrap the array, to preserve subclasses

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py:923, in make_single_pathfinder_fn.<locals>.logp_func(x)
    922 def logp_func(x):
--> 923     logp, _ = logp_dlogp_func(x)
    924     return logp

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/compile/function/types.py:1047, in Function.__call__(self, output_subset, *args, **kwargs)
   1046         thunk = self.vm.thunks[self.vm.position_of_error]
-> 1047     raise_with_op(
   1048         self.maker.fgraph,
   1049         node=self.vm.nodes[self.vm.position_of_error],
   1050         thunk=thunk,
   1051         storage_map=getattr(self.vm, "storage_map", None),
   1052     )
   1053 else:
   1054     # old-style linkers raise their own exceptions

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/link/utils.py:526, in raise_with_op(fgraph, node, thunk, exc_info, storage_map)
    524     # Some exception need extra parameter in inputs. So forget the
    525     # extra long error message in that case.
--> 526 raise exc_value.with_traceback(exc_trace)

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/compile/function/types.py:1037, in Function.__call__(self, output_subset, *args, **kwargs)
   1036 try:
-> 1037     outputs = vm() if output_subset is None else vm(output_subset=output_subset)
   1038 except Exception:

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/op.py:531, in Op.make_py_thunk.<locals>.rval(p, i, o, n, cm)
    523 @is_thunk_type
    524 def rval(
    525     p=p,
   (...)    529     cm=node_compute_map,
    530 ):
--> 531     r = p(n, [x[0] for x in i], o)
    532     for entry in cm:

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/tensor/slinalg.py:72, in Cholesky.perform(self, node, inputs, outputs)
     71 if self.overwrite_a and x.flags["C_CONTIGUOUS"]:
---> 72     out[0] = scipy_linalg.cholesky(
     73         x.T,
     74         lower=not self.lower,
     75         check_finite=self.check_finite,
     76         overwrite_a=True,
     77     ).T
     78 else:

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/scipy/linalg/_decomp_cholesky.py:101, in cholesky(a, lower, overwrite_a, check_finite)
     47 """
     48 Compute the Cholesky decomposition of a matrix.
     49 
   (...)     99 
    100 """
--> 101 c, lower = _cholesky(a, lower=lower, overwrite_a=overwrite_a, clean=True,
    102                      check_finite=check_finite)
    103 return c

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/scipy/linalg/_decomp_cholesky.py:18, in _cholesky(a, lower, overwrite_a, clean, check_finite)
     16 """Common code for cholesky() and cho_factor()."""
---> 18 a1 = asarray_chkfinite(a) if check_finite else asarray(a)
     19 a1 = atleast_2d(a1)

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/numpy/lib/function_base.py:630, in asarray_chkfinite(a, dtype, order)
    629 if a.dtype.char in typecodes['AllFloat'] and not np.isfinite(a).all():
--> 630     raise ValueError(
    631         "array must not contain infs or NaNs")
    632 return a

ValueError: array must not contain infs or NaNs
Apply node that caused the error: Cholesky{lower=True, check_finite=True, on_error='nan', overwrite_a=True}(Composite{...}.4)
Toposort index: 813
Inputs types: [TensorType(float64, shape=(545, 545))]
Inputs shapes: [(545, 545)]
Inputs strides: [(4360, 8)]
Inputs values: ['not shown']
Outputs clients: [[ExtractDiag{offset=0, axis1=0, axis2=1, view=False}(Cholesky{lower=True, check_finite=True, on_error='nan', overwrite_a=True}.0), ExpandDims{axis=0}(Cholesky{lower=True, check_finite=True, on_error='nan', overwrite_a=True}.0), Isnan(Cholesky{lower=True, check_finite=True, on_error='nan', overwrite_a=True}.0), Transpose{axes=[1, 0]}(Cholesky{lower=True, check_finite=True, on_error='nan', overwrite_a=True}.0)]]

Backtrace when the node is created (use PyTensor flag traceback__limit=N to make it longer):
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py", line 144, in get_logp_dlogp_of_ravel_inputs
    [model.logp(jacobian=jacobian), model.dlogp(jacobian=jacobian)],
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc/model/core.py", line 696, in logp
    rv_logps = transformed_conditional_logp(
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc/logprob/basic.py", line 595, in transformed_conditional_logp
    temp_logp_terms = conditional_logp(
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc/logprob/basic.py", line 529, in conditional_logp
    node_logprobs = _logprob(
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/functools.py", line 912, in wrapper
    return dispatch(args[0].__class__)(*args, **kw)
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc/distributions/distribution.py", line 140, in logp
    return class_logp(value, *dist_params)
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc/distributions/multivariate.py", line 295, in logp
    quaddist, logdet, posdef = quaddist_chol(value, mu, cov)
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc/distributions/multivariate.py", line 178, in quaddist_chol
    chol_cov = nan_lower_cholesky(cov)

HINT: Use the PyTensor flag `exception_verbosity=high` for a debug print-out and storage map footprint of this Apply node.

During handling of the above exception, another exception occurred:

ValueError                                Traceback (most recent call last)
File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py:1538, in multipath_pathfinder(model, num_paths, num_draws, num_draws_per_path, maxcor, maxiter, ftol, gtol, maxls, num_elbo_draws, jitter, epsilon, importance_sampling, progressbar, concurrent, random_seed, pathfinder_kwargs, compile_kwargs)
   1537 task = progress.add_task(desc.format(path_idx=0), completed=0, total=num_paths)
-> 1538 for path_idx, result in enumerate(generator, start=1):
   1539     try:

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py:1112, in make_generator(concurrent, fn, seeds, max_workers)
   1111 else:
-> 1112     yield from _execute_serially(fn, seeds)

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py:1097, in _execute_serially(fn, seeds)
   1096 rng = np.random.default_rng(seed)
-> 1097 yield fn(rng)

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py:959, in make_single_pathfinder_fn.<locals>.single_pathfinder_fn(random_seed)
    958 reseed_rngs(rngs, bfgs_seeds)
--> 959 psi, logP_psi, logQ_psi, elbo_argmax = pathfinder_body_fn(x, g)
    961 if np.all(~np.isfinite(logQ_psi)):

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/compile/function/types.py:1047, in Function.__call__(self, output_subset, *args, **kwargs)
   1046         thunk = self.vm.thunks[self.vm.position_of_error]
-> 1047     raise_with_op(
   1048         self.maker.fgraph,
   1049         node=self.vm.nodes[self.vm.position_of_error],
   1050         thunk=thunk,
   1051         storage_map=getattr(self.vm, "storage_map", None),
   1052     )
   1053 else:
   1054     # old-style linkers raise their own exceptions

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/link/utils.py:526, in raise_with_op(fgraph, node, thunk, exc_info, storage_map)
    524     # Some exception need extra parameter in inputs. So forget the
    525     # extra long error message in that case.
--> 526 raise exc_value.with_traceback(exc_trace)

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/compile/function/types.py:1037, in Function.__call__(self, output_subset, *args, **kwargs)
   1036 try:
-> 1037     outputs = vm() if output_subset is None else vm(output_subset=output_subset)
   1038 except Exception:

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/op.py:531, in Op.make_py_thunk.<locals>.rval(p, i, o, n, cm)
    523 @is_thunk_type
    524 def rval(
    525     p=p,
   (...)    529     cm=node_compute_map,
    530 ):
--> 531     r = p(n, [x[0] for x in i], o)
    532     for entry in cm:

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py:712, in LogLike.perform(self, node, inputs, outputs)
    711 phi = inputs[0]
--> 712 logP = np.apply_along_axis(self.logp_func, axis=-1, arr=phi)
    713 # replace nan with -inf since np.argmax will return the first index at nan

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/numpy/lib/shape_base.py:402, in apply_along_axis(func1d, axis, arr, *args, **kwargs)
    401 for ind in inds:
--> 402     buff[ind] = asanyarray(func1d(inarr_view[ind], *args, **kwargs))
    404 if not isinstance(res, matrix):
    405     # wrap the array, to preserve subclasses

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py:923, in make_single_pathfinder_fn.<locals>.logp_func(x)
    922 def logp_func(x):
--> 923     logp, _ = logp_dlogp_func(x)
    924     return logp

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/compile/function/types.py:1047, in Function.__call__(self, output_subset, *args, **kwargs)
   1046         thunk = self.vm.thunks[self.vm.position_of_error]
-> 1047     raise_with_op(
   1048         self.maker.fgraph,
   1049         node=self.vm.nodes[self.vm.position_of_error],
   1050         thunk=thunk,
   1051         storage_map=getattr(self.vm, "storage_map", None),
   1052     )
   1053 else:
   1054     # old-style linkers raise their own exceptions

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/link/utils.py:526, in raise_with_op(fgraph, node, thunk, exc_info, storage_map)
    524     # Some exception need extra parameter in inputs. So forget the
    525     # extra long error message in that case.
--> 526 raise exc_value.with_traceback(exc_trace)

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/compile/function/types.py:1037, in Function.__call__(self, output_subset, *args, **kwargs)
   1036 try:
-> 1037     outputs = vm() if output_subset is None else vm(output_subset=output_subset)
   1038 except Exception:

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/op.py:531, in Op.make_py_thunk.<locals>.rval(p, i, o, n, cm)
    523 @is_thunk_type
    524 def rval(
    525     p=p,
   (...)    529     cm=node_compute_map,
    530 ):
--> 531     r = p(n, [x[0] for x in i], o)
    532     for entry in cm:

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/tensor/slinalg.py:72, in Cholesky.perform(self, node, inputs, outputs)
     71 if self.overwrite_a and x.flags["C_CONTIGUOUS"]:
---> 72     out[0] = scipy_linalg.cholesky(
     73         x.T,
     74         lower=not self.lower,
     75         check_finite=self.check_finite,
     76         overwrite_a=True,
     77     ).T
     78 else:

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/scipy/linalg/_decomp_cholesky.py:101, in cholesky(a, lower, overwrite_a, check_finite)
     47 """
     48 Compute the Cholesky decomposition of a matrix.
     49 
   (...)     99 
    100 """
--> 101 c, lower = _cholesky(a, lower=lower, overwrite_a=overwrite_a, clean=True,
    102                      check_finite=check_finite)
    103 return c

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/scipy/linalg/_decomp_cholesky.py:18, in _cholesky(a, lower, overwrite_a, clean, check_finite)
     16 """Common code for cholesky() and cho_factor()."""
---> 18 a1 = asarray_chkfinite(a) if check_finite else asarray(a)
     19 a1 = atleast_2d(a1)

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/numpy/lib/function_base.py:630, in asarray_chkfinite(a, dtype, order)
    629 if a.dtype.char in typecodes['AllFloat'] and not np.isfinite(a).all():
--> 630     raise ValueError(
    631         "array must not contain infs or NaNs")
    632 return a

ValueError: array must not contain infs or NaNs
Apply node that caused the error: Cholesky{lower=True, check_finite=True, on_error='nan', overwrite_a=True}(Composite{...}.4)
Toposort index: 813
Inputs types: [TensorType(float64, shape=(545, 545))]
Inputs shapes: [(545, 545)]
Inputs strides: [(4360, 8)]
Inputs values: ['not shown']
Outputs clients: [[ExtractDiag{offset=0, axis1=0, axis2=1, view=False}(Cholesky{lower=True, check_finite=True, on_error='nan', overwrite_a=True}.0), ExpandDims{axis=0}(Cholesky{lower=True, check_finite=True, on_error='nan', overwrite_a=True}.0), Isnan(Cholesky{lower=True, check_finite=True, on_error='nan', overwrite_a=True}.0), Transpose{axes=[1, 0]}(Cholesky{lower=True, check_finite=True, on_error='nan', overwrite_a=True}.0)]]

Backtrace when the node is created (use PyTensor flag traceback__limit=N to make it longer):
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py", line 144, in get_logp_dlogp_of_ravel_inputs
    [model.logp(jacobian=jacobian), model.dlogp(jacobian=jacobian)],
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc/model/core.py", line 696, in logp
    rv_logps = transformed_conditional_logp(
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc/logprob/basic.py", line 595, in transformed_conditional_logp
    temp_logp_terms = conditional_logp(
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc/logprob/basic.py", line 529, in conditional_logp
    node_logprobs = _logprob(
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/functools.py", line 912, in wrapper
    return dispatch(args[0].__class__)(*args, **kw)
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc/distributions/distribution.py", line 140, in logp
    return class_logp(value, *dist_params)
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc/distributions/multivariate.py", line 295, in logp
    quaddist, logdet, posdef = quaddist_chol(value, mu, cov)
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc/distributions/multivariate.py", line 178, in quaddist_chol
    chol_cov = nan_lower_cholesky(cov)

HINT: Use the PyTensor flag `exception_verbosity=high` for a debug print-out and storage map footprint of this Apply node.
Apply node that caused the error: LogLike{logp_func=<function make_single_pathfinder_fn.<locals>.logp_func at 0x7f16567b4fe0>}(if{inplace}.0)
Toposort index: 164
Inputs types: [TensorType(float64, shape=(None, 10, None))]
Inputs shapes: [(30, 10, 12)]
Inputs strides: [(960, 8, 80)]
Inputs values: ['not shown']
Outputs clients: [[Sub(LogLike{logp_func=<function make_single_pathfinder_fn.<locals>.logp_func at 0x7f16567b4fe0>}.0, AdvancedSetSubtensor.0)]]

Backtrace when the node is created (use PyTensor flag traceback__limit=N to make it longer):
  File "/tmp/ipykernel_880351/3623004794.py", line 42, in <module>
    approx = pmx.fit(method="pathfinder")
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/fit.py", line 35, in fit
    return fit_pathfinder(**kwargs)
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py", line 1717, in fit_pathfinder
    mp_result = multipath_pathfinder(
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py", line 1507, in multipath_pathfinder
    single_pathfinder_fn = make_single_pathfinder_fn(
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py", line 940, in make_single_pathfinder_fn
    pathfinder_body_fn = make_pathfinder_body(
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py", line 838, in make_pathfinder_body
    logP_phi = loglike(phi)
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/graph/op.py", line 293, in __call__
    node = self.make_node(*inputs, **kwargs)
  File "/home/aphc/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py", line 707, in make_node
    outputs = pt.tensor(dtype="float64", shape=(None, None))

HINT: Use the PyTensor flag `exception_verbosity=high` for a debug print-out and storage map footprint of this Apply node.

During handling of the above exception, another exception occurred:

ValueError                                Traceback (most recent call last)
Cell In[14], line 42
     37 y_ = gp.marginal_likelihood("y", X=t, y=y, noise=cov_noise)
     39 # this line calls an optimizer to find the MAP
     40 # mp = pm.find_MAP(include_transformed=True)
---> 42 approx = pmx.fit(method="pathfinder")

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/fit.py:35, in fit(method, **kwargs)
     32 if method == "pathfinder":
     33     from pymc_extras.inference.pathfinder import fit_pathfinder
---> 35     return fit_pathfinder(**kwargs)
     37 if method == "laplace":
     38     from pymc_extras.inference.laplace import fit_laplace

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py:1717, in fit_pathfinder(model, num_paths, num_draws, num_draws_per_path, maxcor, maxiter, ftol, gtol, maxls, num_elbo_draws, jitter, epsilon, importance_sampling, progressbar, concurrent, random_seed, postprocessing_backend, inference_backend, pathfinder_kwargs, compile_kwargs)
   1714     maxcor = max(maxcor, 5)
   1716 if inference_backend == "pymc":
-> 1717     mp_result = multipath_pathfinder(
   1718         model,
   1719         num_paths=num_paths,
   1720         num_draws=num_draws,
   1721         num_draws_per_path=num_draws_per_path,
   1722         maxcor=maxcor,
   1723         maxiter=maxiter,
   1724         ftol=ftol,
   1725         gtol=gtol,
   1726         maxls=maxls,
   1727         num_elbo_draws=num_elbo_draws,
   1728         jitter=jitter,
   1729         epsilon=epsilon,
   1730         importance_sampling=importance_sampling,
   1731         progressbar=progressbar,
   1732         concurrent=concurrent,
   1733         random_seed=random_seed,
   1734         pathfinder_kwargs=pathfinder_kwargs,
   1735         compile_kwargs=compile_kwargs,
   1736     )
   1737     pathfinder_samples = mp_result.samples
   1738 elif inference_backend == "blackjax":

File ~/projects/pymc-devs/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pymc_extras/inference/pathfinder/pathfinder.py:1597, in multipath_pathfinder(model, num_paths, num_draws, num_draws_per_path, maxcor, maxiter, ftol, gtol, maxls, num_elbo_draws, jitter, epsilon, importance_sampling, progressbar, concurrent, random_seed, pathfinder_kwargs, compile_kwargs)
   1593             raise ValueError(
   1594                 "All paths failed. Consider decreasing the jitter or reparameterizing the model."
   1595             )
   1596     else:
-> 1597         raise ValueError(
   1598             "BUG: Failed to iterate!"
   1599             "Please report this issue at: "
   1600             "https://github.com/pymc-devs/pymc-extras/issues "
   1601             "with your code to reproduce the issue and the following details:\n"
   1602             f"pathfinder_config: \n{pathfinder_config}\n"
   1603             f"compile_kwargs: {compile_kwargs}\n"
   1604             f"pathfinder_kwargs: {pathfinder_kwargs}\n"
   1605             f"num_paths: {num_paths}\n"
   1606             f"num_draws: {num_draws}\n"
   1607         )
   1609 return mpr

ValueError: BUG: Failed to iterate!Please report this issue at: https://github.com/pymc-devs/pymc-extras/issues with your code to reproduce the issue and the following details:
pathfinder_config: 
PathfinderConfig(num_draws=1000, maxcor=8, maxiter=1000, ftol=1e-05, gtol=1e-08, maxls=1000, jitter=2.0, epsilon=1e-08, num_elbo_draws=10)
compile_kwargs: {}
pathfinder_kwargs: {}
num_paths: 4
num_draws: 1000

[aphc14]

Okay this should be fixed with the submitted PR above :)