Adding Up Every Step of the Journey

A model is trained for 45 epochs, and every epoch performs 50 parameter updates.

The learning rate follows a step decay. It starts at $\alpha_0 = 0.1$ and is halved every 10 epochs, staying perfectly constant inside each block:

One clean way to compare two schedules is the total learning budget — the learning rate summed over every single update in the whole run:

$B = \sum_{t=1}^{T} \alpha_t$

where $T$ is the total number of updates. This is the discrete cousin of the area under the schedule curve, $\int \alpha(t)\,dt$, and it measures how far the optimizer is permitted to travel in total across training.

What is the total learning budget $B$ for this run?

Round your answer to two decimal places.