[ET-VK] buffer implementation of rotary positional embeddings #15679
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
added 3 commits
November 6, 2025 16:25
Pull Request resolved: #15619 Title says it all! Previously, to_dim_order_copy was handled by removing the op. However, this is not possible if the op is modifying the dtype of the original tensor, so these instances of the op would be skipped by the partitioner. This diff adds an implementation dtype conversion, which allows to_dim_order_copy to be lowered. ghstack-source-id: 321555048 @exported-using-ghexport Differential Revision: [D86340341](https://our.internmc.facebook.com/intern/diff/D86340341/)
…numel limit Pull Request resolved: #15651 Title says it all! To give a concrete example, Llama3.2-1B-Instruct will have attn weights with size `{1, 32, max_seq_len, max_context_len}`. Usually `max_seq_len == max_context_len`, and if `max_context_len = 2048` Then the attention weight tensors will have sizes `{1, 32, 2048, 2048}` which will contain 134217728 elements. The `maxStorageBufferRange` for Adreno 750 is also 134217728 (2^27), so using context length of 2048 will produce incorrect results on Adreno 750. In practice, it is unlikely that the prompt sequence length will be equal to the context length, so the solution is to adjust down the `max_seq_len` dim of the attention weight tensors to ensure that the GPU buffer numel limit is not hit. ghstack-source-id: 321555042 @exported-using-ghexport Differential Revision: [D86443407](https://our.internmc.facebook.com/intern/diff/D86443407/)
Pull Request resolved: #15620 Title says it all! ghstack-source-id: 321555043 @exported-using-ghexport Differential Revision: [D86340338](https://our.internmc.facebook.com/intern/diff/D86340338/)
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/15679
Note: Links to docs will display an error until the docs builds have been completed. ⏳ No Failures, 133 PendingAs of commit 962c01e with merge base d361573 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
SS-JIA
approved these changes
Nov 7, 2025
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 PR was created by the merge bot to help merge the original PR into the main branch.
ghstack PR number: #15620 by @SS-JIA
^ Please use this as the source of truth for the PR details, comments, and reviews
ghstack PR base: https://github.com/pytorch/executorch/tree/gh/SS-JIA/356/base
ghstack PR head: https://github.com/pytorch/executorch/tree/gh/SS-JIA/356/head
Merge bot PR base: https://github.com/pytorch/executorch/tree/gh/SS-JIA/357/orig
Merge bot PR head: https://github.com/pytorch/executorch/tree/gh/SS-JIA/356/orig
Differential Revision: D86340338
@diff-train-skip-merge