Draft: Yifeit/assume pure #308
Draft
Conversation
This file contains hidden or 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
tengyifei
force-pushed
the
yifeit/assume-pure
branch
2 times, most recently
from
cd77464 to
d577a40
Compare
yaoshiang
reviewed
Jun 14, 2025
| def replace_nn_linear_with_einsum(module: torch.nn.Module, config: DictConfig): | ||
| """Recursively replace `nn.Linear` layers with `EinsumLinear` in the module. | ||
|
|
||
| Without this patch, an `nn.Linear` module in PyTorch/XLA will lower to reshapes |
There was a problem hiding this comment.
I recall that xla's einsum worked fine when used without custom-op - can you point me to where the einsum becomes a custom op and remind me why?
do we need to use this torch_xla.distributed.spmd.xla_sharding.apply_xla_patch_to_nn_linear - could we use assume_pure on the linear model directly in this code? That question may be missing some basic knowledge of assume-pure, my apologies if so.
tengyifei
added a commit
to pytorch/xla
that referenced
this pull request
Jun 16, 2025
- Support nested tuples in `assume_pure(mark_sharding)` - Add a `PureModule` from AI-Hypercomputer/torchprime#308 - Support `PureModule(EinsumLinear)` which uses `torch.ops.xla.einsum_linear_forward`
tengyifei
added a commit
to pytorch/xla
that referenced
this pull request
Jun 16, 2025
- Support nested tuples in `assume_pure(mark_sharding)` - Add a `PureModule` from AI-Hypercomputer/torchprime#308 - Support `PureModule(EinsumLinear)` which uses `torch.ops.xla.einsum_linear_forward`
tengyifei
added a commit
to pytorch/xla
that referenced
this pull request
Jun 16, 2025
- Support nested tuples in `assume_pure(mark_sharding)` - Add a `PureModule` from AI-Hypercomputer/torchprime#308 - Support `PureModule(EinsumLinear)` which uses `torch.ops.xla.einsum_linear_forward`
tengyifei
added a commit
to pytorch/xla
that referenced
this pull request
Jun 16, 2025
- Support nested tuples in `assume_pure(mark_sharding)` - Add a `PureModule` from AI-Hypercomputer/torchprime#308 - Support `PureModule(EinsumLinear)` which uses `torch.ops.xla.einsum_linear_forward`
TESTED: python3 torchprime/torch_xla_models/train.py model.pure_modules=[LlamaMLP,Linear]
tengyifei
force-pushed
the
yifeit/assume-pure
branch
from
574f142 to
c8ba416
Compare
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.
No description provided.