flashinfer.gemm.mm_fp16_fp8_workspace_bytes

flashinfer.gemm.mm_fp16_fp8_workspace_bytes(m: int, n: int, k: int, device: device | None = None, tactic: str | None = None) → int

Workspace bytes mm_fp16_fp8() needs for an [M, K] x [N, K] problem.

Same arguments, answer and errors as mm_bf16_fp8_workspace_bytes(): the two functions run the same kernels and the workspace does not depend on the activation dtype.