From 527bcd14d46595b2bde9e48f3d2f10879623b891 Mon Sep 17 00:00:00 2001 From: jennyyyyzhen <47012288+jennyyyyzhen@users.noreply.github.com> Date: Sat, 31 Jan 2026 01:03:57 -0800 Subject: [PATCH] [ROCM] Enable aiter attn backend for qwen3-next model (#32492) Signed-off-by: jennyyyyzhen --- docs/design/attention_backends.md | 2 +- vllm/v1/attention/backends/rocm_aiter_fa.py | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) diff --git a/docs/design/attention_backends.md b/docs/design/attention_backends.md index ef96d002d7..6e84dde924 100644 --- a/docs/design/attention_backends.md +++ b/docs/design/attention_backends.md @@ -168,7 +168,7 @@ Priority is **1 = highest** (tried first). | `FLASH_ATTN` | FA3* | fp16, bf16 | `auto`, `bfloat16`, `fp8`, `fp8_e4m3`, `fp8_e5m2` | %16 | Any | ✅ | ❌ | All | 9.x | | `FLASH_ATTN_DIFFKV` | | fp16, bf16 | `auto` | Any | Any | ❌ | ❌ | Decoder | Any | | `FLEX_ATTENTION` | | fp16, bf16, fp32 | `auto`, `bfloat16` | Any | Any | ❌ | ✅ | Decoder, Encoder Only | Any | -| `ROCM_AITER_FA` | | fp16, bf16 | `auto` | %16 | 64, 128, 256 | ❌ | ❌ | Decoder | N/A | +| `ROCM_AITER_FA` | | fp16, bf16 | `auto` | 16, 32 | 64, 128, 256 | ❌ | ❌ | Decoder | N/A | | `ROCM_AITER_UNIFIED_ATTN` | | fp16, bf16 | `auto` | Any | Any | ❌ | ❌ | Decoder | N/A | | `ROCM_ATTN` | | fp16, bf16, fp32 | `auto` | 16, 32, 544 | 32, 64, 96, 128, 160, 192, 224, 256 | ❌ | ❌ | Decoder | N/A | | `TREE_ATTN` | | fp16, bf16 | `auto` | %16 | 32, 64, 96, 128, 160, 192, 224, 256 | ❌ | ❌ | Decoder | Any | diff --git a/vllm/v1/attention/backends/rocm_aiter_fa.py b/vllm/v1/attention/backends/rocm_aiter_fa.py index e122882c7b..8ed06fc8d5 100644 --- a/vllm/v1/attention/backends/rocm_aiter_fa.py +++ b/vllm/v1/attention/backends/rocm_aiter_fa.py @@ -683,7 +683,7 @@ class AiterFlashAttentionBackend(AttentionBackend): @staticmethod def get_supported_kernel_block_sizes() -> list[int | MultipleOf]: - return [MultipleOf(16)] + return [16, 32] @classmethod def get_supported_head_sizes(cls) -> list[int]: