6676 port generative networks transformer

[marksgraham]

Towards #6676 .

Description

Adds a simple decoder-only transformer architecture.

Types of changes

• Non-breaking change (fix or new feature that would not break existing functionality).
• Breaking change (fix or new feature that would cause existing functionality to change).
• New tests added to cover the changes.
• Integration tests passed locally by running ./runtests.sh -f -u --net --coverage.
• Quick tests passed locally by running ./runtests.sh --quick --unittests --disttests.
• In-line docstrings updated.
• Documentation updated, tested make html command in the docs/ folder.

[KumoLiu]

Thanks for the PR! Overall looks good to me.

Could you please help add more test?
Similarly like this, multiple loops cover the most of the situation. This makes it easier to find bugs later on when refactoring. At the moment, it appears that there is only one hardcoded case.

MONAI/tests/test_selfattention.py

Lines 28 to 35 in d5585c3

	for dropout_rate in np.linspace(0, 1, 4):
	for hidden_size in [360, 480, 600, 768]:
	for num_heads in [4, 6, 8, 12]:
	test_case = [
	{"hidden_size": hidden_size, "num_heads": num_heads, "dropout_rate": dropout_rate},
	(2, 512, hidden_size),
	(2, 512, hidden_size),
	]

[KumoLiu]

/build

[atbenmurray]

Taking a look today