Initialize n-best list of hypotheses.

Args: param1: param2: support: target_image: target_image_l: vali

Number of hypotheses in the list.

Extract a non-constant sub-tree from an equation.

Clean solutions of second order differential equations.

Evaluate equation `pred_equation` given as a string.

model_predict is a callable that takes an X of shape (n_datapoints, n_variables) and returns scalar predictions

Evaluate sklearn model at model_path.

Evaluate sklearn model at model_path.

Same API as fit functions in sklearn: X [Number_of_points, Number_of_features], Y [Number_of_points]

Decode a sentence given initial start. `x`: - LongTensor(bs, slen) <EOS> W1 W2 W3 <EOS> <PAD> <EOS> W1 W2 W3

Detect whether some expressions contain complex numbers.

Detect whether some expressions contain a NaN / Infinity symbol.

Used for the transformer

Return True if an equation is not the validation set and is numerically meaningfull (i.e. values all differs from nan, -inf, +inf, all zeros)

Assert both symbolically and numerically that the equation is not in the validation set Args: idx: index to the Eq in the

For experiments before 2021-02-03, there was no key 'benchmark_name'. They were all using the AI Feynman dataset

Re-index coefficients (i.e. if a1 is there and not a0, replace a1 by a0, and recursively).

Remove the multiplicative factor of an expression, and return it.

We updated our benchmark name from "ours" to "ours-nc".

Simplify a differential equation by removing non-zero factors.

Return the smallest sub-tree in an expression that contains all given symbols.

Replace x1 -> x, x2 -> y, x3 -> z in order to work with equivalent formats. Also, add 0 times all variables in order to make all variables ap

returns training dataloader

returns validation dataloader