Objective function not passed extra arguments when Hessian is provided

[ForceBru]

Code

from jax.config import config
config.update("jax_enable_x64", True)

import jax
import jax.numpy as jnp
import cyipopt

def basic_function(x: jnp.ndarray, data: jnp.ndarray):
    return (x[0] * data + x[1]).mean()

cyipopt.minimize_ipopt(
    basic_function,
    jnp.array([1.0, 2.0]), (jnp.array([1.,2,5,1,5,2,8,2,0,4,9]), ),
    jac=jax.jacfwd(basic_function),
    hess=jax.hessian(basic_function),
)

Error

# Message from JAX trimmed
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/forcebru/mambaforge/lib/python3.10/site-packages/cyipopt/scipy_interface.py", line 314, in minimize_ipopt
    x, info = nlp.solve(_x0)
  File "cyipopt/cython/ipopt_wrapper.pyx", line 642, in ipopt_wrapper.Problem.solve
  File "cyipopt/cython/ipopt_wrapper.pyx", line 895, in ipopt_wrapper.hessian_cb
  File "/Users/forcebru/mambaforge/lib/python3.10/site-packages/cyipopt/scipy_interface.py", line 164, in hessian
    H = obj_factor * self.obj_hess(x)  # type: ignore
  File "/Users/forcebru/mambaforge/lib/python3.10/site-packages/jax/_src/api.py", line 1286, in jacfun
    y, jac = vmap(pushfwd, out_axes=(None, -1))(_std_basis(dyn_args))
TypeError: basic_function() missing 1 required positional argument: 'data'

I think in this code:

cyipopt/cyipopt/scipy_interface.py

Lines 163 to 164 in 8c68027

	def hessian(self, x, lagrange, obj_factor):
	H = obj_factor * self.obj_hess(x) # type: ignore

...self.obj_hess should be called with self.args and self.kwargs, like self.fun and self.jac:

cyipopt/cyipopt/scipy_interface.py

Line 149 in 8c68027

return self.jac(x, *self.args, **self.kwargs) # .T

• cyipopt 1.2.0

[brocksam]

Thanks for reporting this. I agree, it does look like self.args and self.kwargs are needed where you suggest. Would you be happy to submit a PR making this change?

[ForceBru]

Sure, I'll give it a try this week!

[moorepants]

This now works on master:

$ ipython
Python 3.10.10 | packaged by conda-forge | (main, Mar 24 2023, 20:08:06) [GCC 11.3.0]
Type 'copyright', 'credits' or 'license' for more information
IPython 8.12.0 -- An enhanced Interactive Python. Type '?' for help.

In [1]: %paste
from jax.config import config
config.update("jax_enable_x64", True)

import jax
import jax.numpy as jnp
import cyipopt

def basic_function(x: jnp.ndarray, data: jnp.ndarray):
    return (x[0] * data + x[1]).mean()

cyipopt.minimize_ipopt(
    basic_function,
    jnp.array([1.0, 2.0]), (jnp.array([1.,2,5,1,5,2,8,2,0,4,9]), ),
    jac=jax.jacfwd(basic_function),
    hess=jax.hessian(basic_function),
)

## -- End pasted text --
No GPU/TPU found, falling back to CPU. (Set TF_CPP_MIN_LOG_LEVEL=0 and rerun for more info.)

******************************************************************************
This program contains Ipopt, a library for large-scale nonlinear optimization.
 Ipopt is released as open source code under the Eclipse Public License (EPL).
         For more information visit https://github.com/coin-or/Ipopt
******************************************************************************

Out[1]: 
 message: b'It seems that the iterates diverge.'
 success: False
  status: 4
     fun: -1.1315674525145682e+21
       x: [-2.956e+20 -8.339e+19]
     nit: 34
    info:     status: 4
                   x: [-2.956e+20 -8.339e+19]
                   g: []
             obj_val: -1.1315674525145682e+21
              mult_g: []
            mult_x_L: [ 0.000e+00  0.000e+00]
            mult_x_U: [ 0.000e+00  0.000e+00]
          status_msg: b'It seems that the iterates diverge.'
    nfev: 35
    njev: 36

Solved by #197

[moorepants]

although, no idea about this: status_msg: b'It seems that the iterates diverge.', but the error about the args/kwargs is gone.