minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 13 Packages Projects Releases Wiki Activity

[Speculative Decoding] Add FA4-based Spec Support (#21080)

Browse Source 
Co-authored-by: luhongyu.4869 <luhongyu.4869@bytedance.com>
This commit is contained in:
narutolhy
2026-04-04 02:09:45 -07:00
committed by GitHub
co-authored by luhongyu.4869
parent 34d5765e2f
commit 24763256b9
5 changed files with 94 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/advanced_features/attention_backend.md
+1 -1
View File
@@ -19,7 +19,7 @@ The support matrix is split into two parts: MHA (standard attention) and MLA (mu
|---------------------------------|-----------------------------|------------------|-----------------|-----------------|-----------------|--------------------|----------------|
| **FlashInfer** | ✅ | ✅ | ❌ | ✅ | ✅ | ✅ | ❌ |
| **FA3 (FlashAttention 3)** | ✅ | ✅ | ❌ | ✅ | ✅ | ✅ | ✅ |
| **FA4 (FlashAttention 4)** | 128 | ❌ | ✅ | ❌ | ❌ | ❌ | ✅ |
| **FA4 (FlashAttention 4)** | 128 | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ |
| **Triton** | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
| **Torch Native (SDPA)** | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ |
| **FlexAttention (PyTorch)** | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ |