Merge branch 'master' into sliding-window

modules/generative_head/composite_head_A.py

modules/generative_head/composite_head_B.py

modules/generative_head/composite_head_C.py

modules/generative_head/composite_head_D.py

modules/generative_head/double_substitution_head.py

modules/generative_head/substitution_head.py

modules/token_embedding/double_substitution_embedding.py

modules/token_embedding/substitution_embedding.py