Perplexity from NLL

Compute the perplexity of a language model from its mean negative log-likelihood per token (natural log): $\exp (\text{mean NLL})\backslash exp(\backslash text\{mean\; NLL\})$.

Implement solve(mean_nll).

Returns. Return a float, at least 1 for non-negative NLL.

Examples

Example 1

Input:

solve(1.0)

Output:

2.718282

Example 2

Input:

solve(0.0)

Output:

1.0