frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-20 12:54:11 +08:00
Code Issues Packages Projects Releases Wiki Activity

Updated Autograd Basics (markdown)

Manuel
2024-07-03 19:25:40 +02:00
parent 395d1d7c4d
commit 33070aece6
1 changed files with 3 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
Autograd-Basics.md

@ -114,3 +114,6 @@ python test/<test_file.py> -k attn
- tests requiring forward AD to be implemented _when you turn your op from CompositeImplicit to CompositeExplicit_, since the forward AD will no longer be derived for you.
- tests requiring decompositions will fail with messages like `aten.attn.default - couldn't find symbolic meta function/decomposition` _when you turn your op from CompositeImplicit to CompositeExplicit_, since those won't be derived for you anymore either.
- You can skip `attn` in the fake_autocast_device tests for CPU _when you turn your op from CompositeImplicit to CompositeExplicit_.
## Next
Unit 3: Autograd - [Autograd Lab](https://github.com/pytorch/pytorch/wiki/Autograd-Onboarding-Lab)