| 1 | # SPDX-License-Identifier: Apache-2.0 |
| 2 | # SPDX-FileCopyrightText: Copyright contributors to the vLLM project |
| 3 | """Qwen4Exp low-latency GEMM hook for AMD ROCm.""" |
| 4 | |
| 5 | import torch |
| 6 | from torch import nn |
| 7 | |
| 8 | |
| 9 | def enable_qwen4_exp_low_latency_gemm( |
| 10 | module: nn.Module, |
| 11 | dtype: torch.dtype, |
| 12 | ) -> None: |
| 13 | """Keep the standard vLLM linear methods on AMD ROCm.""" |
| 14 | del module, dtype |