minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

[diffusion] fix: keep Qwen-Image 2.1 prefix KV per layer under Cache-DiT (#40472)

Browse Source
This commit is contained in:
WenhaoZhang
2026-09-21 08:48:31 +08:00
committed by GitHub
parent 501b7851e4
commit b912db67ea
4 changed files with 93 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/cookbook/diffusion/Qwen-Image/Qwen-Image-2.1.mdx
+8
View File
@@ -216,3 +216,11 @@ Keep eager execution as the default. Breakable CUDA Graph replay requires
matching resolution and condition-prefix length; unseen shapes run eagerly.
Text buckets alone do not guarantee replay. SageAttention and Cache-DiT can
change numerical results and require quality checks for your workload.
### Cache-DiT
Enable `--enable-cache-dit true` or `SGLANG_CACHE_DIT_ENABLED=true`. 2.1 prefix
KV is per layer: each block slices caches by `_layer_id`. Cache-DiT wraps
`transformer_blocks` and forwards the same extras to every layer; without that
slice, later layers reuse layer 0 and the image collapses to color noise.
See the [Cache-DiT guide](/docs/sglang-diffusion/cache_dit).