register torch._C._functorch.unwrap_if_dead as identity
#661
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
So I'm not opposed, but as far as I can tell, #660 is about the TensorParallelism implementation in NeMo . So I wonder if we would take on this problem a bit further up the stack (e.g. at the level of the custom autograd.Function, Module, ...)? |
Signed-off-by: Masaki Kozuki <mkozuki@nvidia.com>
crcrpar
force-pushed
the
crpa/unwrap_if_dead-for-custom-autograd-function
branch
from
94096d0 to
dc18d00
Compare
crcrpar
commented
Jun 27, 2024
t-vi
approved these changes
Jun 27, 2024
There was a problem hiding this comment.
Thank you @crcrpar . Great analysis that we want to not fall over in autograd.Function.
To my mind we might want to warn in autograd.Function.apply, but we can do that elsewhere (probably wants to be in jit_ext to have lookaside with warning and then trace through the original apply), but let's have this for now.
thunder/torch/__init__.py
Outdated
| @@ -4936,6 +4936,14 @@ def _set_grad_enabled_with_warning(enabled: bool) -> None: | |||
| register_function(torch._C._set_grad_enabled, _set_grad_enabled_with_warning) | |||
|
|
|||
|
|
|||
| def _unwrap_if_dead(tensor): | |||
| warnings.warn("torch._C._functorch.unwrap_if_dead have no effect under thunder.jit") | |||
There was a problem hiding this comment.
probably better as warn_once, but hey.
Signed-off-by: Masaki Kozuki <mkozuki@nvidia.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What does this PR do?
Fixes #660.
The function is called inside of custom
torch.autograd.Function. If one uses onlytorchfunctions in their customFunction, then it should be supported.