Is kompute support FP8 / bfloat16 / float16 type constructor?
Is kompute support FP8 / bfloat16 / float16 type constructor?