search

apache / tvm

Open deep learning compiler stack for cpu, gpu and specialized accelerators
https://tvm.apache.org/
Apache License 2.0
11.41k stars 3.4k forks source link

[Relax] Alloc BYOC workspace with R.builtin.alloc_tensor #17110

Closed vinx13 closed 1 week ago

vinx13 commented 2 weeks ago

This makes the allocation go through memory planning and make it compatible with cuda graph.

cc @yongwww @masahi

yongwww commented 1 week ago

@vinx13 please take a look at lint