biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Attention] FA4 integration (#32974)

Browse Source 
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Signed-off-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
Co-authored-by: Matthew Bonanni <mbonanni@redhat.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
This commit is contained in:
Lucas Wilkinson
2026-03-01 18:44:57 -05:00
committed by GitHub
parent 57a96e26c9
commit 8b5014d3dd
15 changed files with 818 additions and 55 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
docs/design/attention_backends.md
View File

@@ -168,6 +168,7 @@ Priority is **1 = highest** (tried first).
| `FLASHINFER` | TRTLLM† | fp16, bf16 | `auto`, `bfloat16`, `fp8`, `fp8_e4m3`, `fp8_e5m2` | 16, 32, 64 | 64, 128, 256 | ✅ | ❌ | ✅ | Decoder | 10.x |
| `FLASH_ATTN` | FA2* | fp16, bf16 | `auto`, `bfloat16` | %16 | Any | ❌ | ❌ | ✅ | All | ≥8.0 |
| `FLASH_ATTN` | FA3* | fp16, bf16 | `auto`, `bfloat16`, `fp8`, `fp8_e4m3`, `fp8_e5m2` | %16 | Any | ✅ | ❌ | ✅ | All | 9.x |
| `FLASH_ATTN` | FA4* | fp16, bf16 | `auto`, `bfloat16` | %16 | Any | ❌ | ❌ | ✅ | All | ≥10.0 |
| `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, 32 | 64, 128, 256 | ❌ | ❌ | ❌ | Decoder | N/A |
@@ -178,7 +179,7 @@ Priority is **1 = highest** (tried first).
> **†** FlashInfer uses TRTLLM attention on Blackwell (SM100), which supports sinks. Disable via `--attention-config.use_trtllm_attention=0`.
>
> **\*** Specify the FlashAttention version via `--attention-config.flash_attn_version=2` or `3`. Default is FA3 on SM90, FA2 otherwise.
> **\*** Specify the FlashAttention version via `--attention-config.flash_attn_version=2`, `3`, or `4`. Default is FA4 on SM100+ (Blackwell), FA3 on SM90 (Hopper), FA2 otherwise.
## MLA (Multi-head Latent Attention) Backends