The scaling issues comes from your dataset. MNE expects EEG signal in Volts. Typical EEG signal would then range around 40e-6 Volts. Thus, plotting with raw.plot(scalings=dict(eeg=20e-6)) (default) yields. The best solution is for you to figure out which units the data is recorded in, and then apply a conversion:
raw.apply_function(lambda x: x*1e-6, picks="eeg")
In the example above, the EEG data is scaled from microvolts to volts.
You can also ignore this entirely and figure out a scaling which works for your dataset. The simplest way is to use raw.plot() interactively: press (or hold) the - key on your keyboard until the scale looks correct. See the Help button for all the keyboard shortcuts.
Mathieu