[ROCm] Optimize Deepseek R1 on MI300X (#18242)

Co-authored-by: Chen, Todd <zhenchen@amd.com>
2026-02-25 09:01:14 +08:00
parent f1088beb6a
commit b9cf1563de
3 changed files with 7 additions and 2 deletions
--- a/python/sglang/srt/layers/quantization/fp8_utils.py
+++ b/python/sglang/srt/layers/quantization/fp8_utils.py
@@ -61,6 +61,7 @@ if _use_aiter:

    aiter_per1x128_quant = get_hip_quant(aiter.QuantType.per_1x128)

+
 if _is_cuda:
    from sgl_kernel import fp8_blockwise_scaled_mm, fp8_scaled_mm