tensorplay.nn.attention.restore_flash_attention_impl

tensorplay.nn.attention.restore_flash_attention_impl(_raise_warn: bool = True) → None

Restore the default FA2 implementation