Checks whether flash attention is globally enabled in Keras.

Flash attention is a performance-optimized method for computing attention in large models, such as transformers, allowing for faster and more memory-efficient operations. This function checks the global Keras configuration to determine if flash attention is enabled for compatible layers (e.g., MultiHeadAttention).

Note that enabling flash attention does not guarantee it will always be used. Typically, the inputs must be in float16 or bfloat16 dtype, and input layout requirements may vary depending on the backend.

Usage

config_is_flash_attention_enabled()

Value

FALSE if disabled; otherwise, it indicates that it is enabled.

See also

config_disable_flash_attention() config_enable_flash_attention()