11# Kernels
22
33<div align =" center " >
4- <img src =" https://github.com/user-attachments/assets/64a652f3-0cd3-4829-b3c1-df13f7933569 " width =" 450 " height =" 450 " alt =" kernel-builder logo " >
4+ <a href =" https://huggingface.co/kernels " >
5+ <img class =" block dark:hidden " src =" https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/kernels/kernels-thumbnail-light.png " alt =" Kernels " />
6+ <img class =" hidden dark:block " src =" https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/kernels/kernels-thumbnail-dark.png " alt =" Kernels " />
7+ </a >
58</div >
69
710The Kernel Hub allows Python libraries and applications to load compute
8- kernels directly from the [ Hub] ( https://hf.co/ ) . To support this kind
9- of dynamic loading, Hub kernels differ from traditional Python kernel
10- packages in that they are made to be:
11+ kernels directly from the [ Hub] ( https://huggingface.co/ ) . Kernels are a first-class
12+ repository type on the Hub, with dedicated pages that surface supported
13+ hardware and versions. To support dynamic loading, Hub kernels differ from
14+ traditional Python kernel packages in that they are made to be:
1115
1216- ** Portable** : a kernel can be loaded from paths outside ` PYTHONPATH ` .
1317- ** Unique** : multiple versions of the same kernel can be loaded in the
@@ -16,8 +20,7 @@ packages in that they are made to be:
1620 the different PyTorch build configurations (various CUDA versions
1721 and C++ ABIs). Furthermore, older C library versions must be supported.
1822
19- You can [ search for kernels] ( https://huggingface.co/models?other=kernels ) on
20- the Hub.
23+ Browse available kernels at [ huggingface.co/kernels] ( https://huggingface.co/kernels ) .
2124
2225If you're looking for a more involved "Why kernels?" answer, refer to
2326[ this page] ( ./why_kernels.md ) .
0 commit comments