The standard way is, as is pointed out in a parallel comment, checking when the user logs in (since they have to send their unhashed password to the server as part of logging in).

Another approach is to use things like https://haveibeenpwned.com/ to detect users whose credentials are compromised and alert them.

Another more involved approach is to actually try to find collisions with user passwords (using either common wordlists, generated wordlists, or straight brute force) and then notifying users if they find a hit.

OP wrote "When I logged into" so presumably at this point the website can check the password against a password strength/complexity algorithm or top 1000 lists. Ther user had to change the password immediately. Only if you try to skip the step it will auto-reset.

I don't know what method GH is using, but the standard attack on hashed passwords is the use of rainbow tables[1]

[1] https://en.wikipedia.org/wiki/Rainbow_table

Most modern hash algorithm like bcrypt or scrypt add automatically a salt to prevent the use of rainbow tables.