-
我理解TVM所做的优化其实也可以由现有的机器学习框架来完成(e.g. Tensorflow, Pytorch..),比如可以通过开放更多和TVM类似API的形式(e.g. 也提供reorder, parallel...)。请问为什么要单独建立一个框架来单独做优化这件事情?TVM有哪些优势是现有框架无法做到的? |
Beta Was this translation helpful? Give feedback.
Answered by
yzh119
Jul 15, 2022
Replies: 2 comments 2 replies
-
PyTorch的算子都是手写(C++/CUDA)或者第三方库(cublas/cutlass/cudnn)完成,并没有算子级别的IR以及相应的变换抽象(reorder/vectorize/...),因此无法开放相关的API。 |
Beta Was this translation helpful? Give feedback.
0 replies
-
TVM的优势:
|
Beta Was this translation helpful? Give feedback.
2 replies
Answer selected by
hbsun2113
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
TVM的优势: