[Kernel] Add B12X dense linear backends (#52016)

Signed-off-by: Luke Alonso <lalonso@gmail.com>
This commit is contained in:
Luke Alonso
2026-08-13 19:48:33 -07:00
committed by GitHub
parent fe4c5dcd4c
commit 3c79b1a8bf
17 changed files with 2753 additions and 3 deletions
+24
View File
@@ -0,0 +1,24 @@
# B12X Linear Backend
[B12X](https://pypi.org/project/b12x/) provides optional CUDA kernels for
NVIDIA SM120 and SM121 GPUs. Install the dependency with:
```bash
uv pip install "vllm[b12x]"
```
B12X participates in automatic kernel selection after established optimized
backends and before emulation. Select it explicitly with:
```bash
vllm serve <model> --linear-backend b12x
```
## Supported Configurations
| Backend | Supported configurations |
| ------- | ------------------------ |
| Linear | Per-tensor FP8, 128x128 block FP8, MXFP8, NVFP4, and MXFP4 |
Dense W4A16 layers are not handled by B12X and continue to use another
compatible backend such as Marlin.