Use casadi map to more quickly evaluate processed variables

[valentinsulzer]

https://web.casadi.org/docs/#map

[jaskiratsingh2000]

Hi @tinosulzer ! Does the casadi map has to be used in the place of "base_variable_casadis" because, I feel that that's what getting processed.

[valentinsulzer]

instead of this for loop

PyBaMM/pybamm/solvers/processed_variable.py

Lines 123 to 131 in 6ab6c15

	for ts, ys, inputs, base_var_casadi in zip(
	self.all_ts, self.all_ys, self.all_inputs_casadi, self.base_variables_casadi
	):
	for inner_idx, t in enumerate(ts):
	t = ts[inner_idx]
	y = ys[:, inner_idx]
	entries[idx] = float(base_var_casadi(t, y, inputs))
	idx += 1

[jaskiratsingh2000]

Hi @tinosulzer Do you thinnk that something this should work for us?

for ts, ys, inputs, base_var_casadi in zip(
        self.all_ts, self.all_ys, self.all_inputs_casadi, self.base_variables_casadi
):
    num_times = len(ts)
    num_y_vals = ys.shape[0]
    num_entries = num_times * num_y_vals

    # Create a function to evaluate the base_var_casadi over a list of inputs
    def evaluate_base_var_casadi(t, y, inputs):
        return float(base_var_casadi(t, y, inputs))

    # Create casadi functions for the inputs and the base_var_casadi
    input_func = casadi.Function('inputs', [self.input_time] + self.input_vars, ['input'], ['cdense'])
    base_var_func = casadi.Function('base_var_casadi', [self.input_time, self.input_y] + input_func(self.input_time), ['output'], ['cdense'])

    # Evaluate the inputs
    input_vals = [np.array([t] + list(y)) for t, y in zip(ts, ys.T)]
    inputs_casadi = input_func(casadi.horzcat(*input_vals)).full().T

    # Evaluate the processed variables using casadi.map
    entries = casadi.map(evaluate_base_var_casadi, ts, ys.T, inputs_casadi).full().ravel()

    # Set the entries for this output variable
    self.entries[idx:idx + num_entries] = entries

    idx += num_entries

[jaskiratsingh2000]

@tinosulzer Any updates on above what I mentioned?

[valentinsulzer]

I'm not sure exactly the right way to do it since it needs to be optimized to minimize the calculation time, which needs to be tested locally. This is not a good issue to start with. Please check issues labelled "easy"

[jaskiratsingh2000]

Hi @tinosulzer Thanks for letting me know. I'm gently trying to understand the deep of PyBaMM that's why I thought to work on such issues along side issues with labelled "easy"

[rtimms]

closing, see #4464