[Bufferization] Enable OneShot #10
Open
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
* Add conda env yaml * Add mlp example * Add dumps to mlp example (enabled by default)
Signed-off-by: Ilya Enkovich <[email protected]> Co-authored-by: Laurent Montigny <[email protected]>
…nels. (#4) * Add lowering pass and backend to experiment with matmul kernels. Signed-off-by: Ilya Enkovich <[email protected]> * Use MKL for matmul kernels. Signed-off-by: Ilya Enkovich <[email protected]> --------- Signed-off-by: Ilya Enkovich <[email protected]>
Devjiu
force-pushed
the
dmitriim/enable_oneshot_buffer
branch
2 times, most recently
from
68c5230 to
ee73567
Compare
Signed-off-by: Ilya Enkovich <[email protected]>
This commit introduces FuseLinalgOpsPass that merges Transpose in any matmul argument for Linalg dialect. The conversion from torch to linalg has also been changed and `aten::mm` is converted to `linalg::matmul` instead of `linalg::generic`. Signed-off-by: Dmitrii Makarenko <[email protected]>
This commit enables oneshot bufferization and associated passes. Amount of mallocs reduced, butr correctness should be verified on bigger amount of nets/layers. Signed-off-by: Dmitrii Makarenko <[email protected]>
Devjiu
force-pushed
the
dmitriim/enable_oneshot_buffer
branch
from
ee73567 to
be5ba89
Compare
| @@ -34,6 +34,7 @@ def _build_lowering_pipeline(opts: TestOptions): | |||
| "func.func(linalg-fuse-elementwise-ops)", | |||
| "convert-shape-to-std", | |||
| # Bufferize. | |||
| "one-shot-bufferize", | |||
| "func.func(scf-bufferize)", | |||
| "func.func(tm-tensor-bufferize)", | |||
There was a problem hiding this comment.
Shouldn't one-shot replace all of these?
There was a problem hiding this comment.
not all of this, it just do some dep analysis on early stage. AFAIU next bufferization passes converts dialect constructions to buffers, so we still need them
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.
This commit enables oneshot bufferization and associated passes.
Amount of mallocs reduced, but correctness should be verified on bigger
amount of nets/layers.
in shape: torch.Size([100, 128, 128])
w shape: (8192, 16384)
b shape: (8192,)
vanilla pytorch
toch_mlir
toch_mlir oneshot