You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As described by #6953, this appears to be an actual bug that prevents vector-valued inputs for the sde_fn from working. As a result, the second example in the example notebook will not work under PyMC v5.
Cell In[53], line 2
1 def osc_sde(xy, tau, a):
----> 2 x, y = xy[:, 0], xy[:, 1]
3 dx = tau * (x - x**3.0 / 3.0 + y)
4 dy = (1.0 / tau) * (a - x)
File ~/repos/pymc-examples/.pixi/envs/default/lib/python3.12/site-packages/pytensor/tensor/variable.py:502, in _tensor_py_operators.__getitem__(self, args)
500 # Check if the number of dimensions isn't too large.
501 if self.ndim < index_dim_count:
--> 502 raise IndexError("too many indices for array")
504 # Convert an Ellipsis if provided into an appropriate number of
505 # slice(None).
506 if len(ellipses) > 1:
IndexError: too many indices for array
PyMC version information:
PyMC 5.16.2
Context for the issue:
No response
The text was updated successfully, but these errors were encountered:
Yes the distribution is defined as a scalar timeseries in the core case. The error you hit immediately is because init_dist has shape=size, and not say shape=(*size, 2) if the core case was multivariate vector of length 2.
A multivariate version would be a feature request right now, even if it worked before in V3.
Describe the issue:
As described by #6953, this appears to be an actual bug that prevents vector-valued inputs for the
sde_fn
from working. As a result, the second example in the example notebook will not work under PyMC v5.In the scan function, the value of
xy
will always be a scalar because of how the arguments are split.That is,
prev_y
will always be scalar.Reproduceable code example:
Error message:
PyMC version information:
PyMC 5.16.2
Context for the issue:
No response
The text was updated successfully, but these errors were encountered: