Disable Flash Attention with `USE_FLASH_ATTENTION` #580

Job	Run time
matrix	3s
build-and-push-image (xpu-ipex, xpu-ipex-, always, true, PLATFORM=xpu, true, Dockerfile-intel)	8m 28s
build-and-push-image (turing, turing-, always, true, 75, DEFAULT_USE_FLASH_ATTENTION=False, true,...	25m 8s
build-and-push-image (ampere, always, true, 80, true, Dockerfile-cuda)	26m 12s
build-and-push-image (hpu, hpu-, always, true, PLATFORM=hpu, true, Dockerfile-intel)	8m 16s
build-and-push-image (a10, 86-, always, true, 86, true, Dockerfile-cuda)	24m 29s
build-and-push-image (RTX 4000, 89-, always, true, 89, true, Dockerfile-cuda)	24m 33s
build-and-push-image (All, cuda-, always, false, false, Dockerfile-cuda-all)	47m 41s
build-and-push-image (Hopper, hopper-, always, true, 90, true, Dockerfile-cuda)	27m 25s
build-and-push-image (cpu-ipex, cpu-ipex-, always, true, PLATFORM=cpu, true, Dockerfile-intel)	8m 26s
build-and-push-image (cpu, cpu-, always, true, true, Dockerfile)	13m 5s
	3h 33m 46s

Provide feedback