[pull] main from pytorch:main #12
Closed
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
![pull[bot]](https://avatars.githubusercontent.com/in/12910?s=60&v=4){kind=small}
Summary: . Reviewed By: kirklandsign Differential Revision: D54135699 fbshipit-source-id: c9b160140e5a8f86d9ca04d983dcf3c042532d45
Summary: ## Context Move Vulkan graph runtime from PyTorch directory to ExecuTorch directory to improve development logistics: * ExecuTorch delegate changes will no longer require export to PyTorch directory * Makes it much easier to enable OSS build for Vulkan delegate bypass-github-export-checks Reviewed By: shoumikhin Differential Revision: D54133350 fbshipit-source-id: 7c14d6ffbfc9a49bbabae476d2373986fc8acad3
Summary: Pull Request resolved: #2055 Enables fast perf benchmarking for per channel 4 bit dynamic quant of linear layers. Since XNNPACK delegate now supports 4 bit dqlinear, it should get lowered as well. ghstack-source-id: 216323964 exported-using-ghexport bypass-github-export-checks bypass-github-pytorch-ci-checks bypass-github-executorch-ci-checks Reviewed By: digantdesai, shoumikhin, kirklandsign Differential Revision: D54120422 fbshipit-source-id: 63050ef2fe262224114986a976139d07690d4b42
Summary: Pull Request resolved: #2088 Llama2 7b checkpoints are in bfloat16 and the current code fails at sdpa op due to type mismatch bypass-github-export-checks ghstack-source-id: 216296241 exported-using-ghexport Reviewed By: larryliu0820 Differential Revision: D54137104 fbshipit-source-id: ea1fe9b5b696e5139e367f01c5c07fef6501fef8
Summary: pow(n,2) => n*n Reviewed By: manuelcandales Differential Revision: D54142779 fbshipit-source-id: 91185e07e9fe040df090756d78be2cf5e87f6e1d
Summary: Add options for embedding quantization: bitwidth, group_size on CLI Reviewed By: mavlyutovr Differential Revision: D54159472 fbshipit-source-id: 25b0b560667c3875c911b878ee0f28fd042ff713
…loat Summary: Resolve recurring errors where query is c10::Half and key and value float. This should ideally work from first principles, but somehow it does not. We need to fix this but in the meantime this ugly have will enable us to proceed and allow others to debug other aspects of ET lowering. Reviewed By: mavlyutovr Differential Revision: D54167581 fbshipit-source-id: 6cc4e76e3abbf107014b5b9da00e817ee3b2ab03
Summary: Earlier changes to 4bit working diff results in not working 4 bit support. THis diff restores those and avoids using min/max. This would have also intefered with 8bit quant that expects symmeteric min/max unlike 4bit. Reviewed By: digantdesai Differential Revision: D54198222 fbshipit-source-id: 035d34bbd7f87f8eb7fa61ac6b938ecac651cb00
Summary: Currently it is using a magic formula. However, there are cases where we need to set explicitly Currently, it should be a no-op, existing models and CI wouldn't break. Reviewed By: kimishpatel, iseeyuan, Jack-Khuu, shreydesai Differential Revision: D54121131 fbshipit-source-id: 996534456c9e83e10e7fdaa6973c754d37e734d4
Summary: Pull Request resolved: #2101 In [PR #2062](#2062), we introduced the partitioner and removed this failing test. The test fails because we were using the wrong op name. We fix it to that from [PR #1737](#1737). ghstack-source-id: 216477874 exported-using-ghexport Reviewed By: SS-JIA Differential Revision: D54206402 fbshipit-source-id: 0c9ae2af9a380e8aa0e28b107a33ccd36a89033e
Summary: Pull Request resolved: #2102 Derived classes of `OpNode` are currently used only for prepack or execute, never both. This means they need not have both API. Inspired by [Stephen's comment](https://www.internalfb.com/diff/D53982441?dst_version_fbid=370105355800543&transaction_fbid=940226924354801), we will build on `ExecuteNode` to be initialized with member functions for `create_params_block()`, `get_shader()`, etc. `PrepackNode` doesn't need these members. Hence, it makes sense to separate the classes. Feel free to suggest better names. I don't really like mine. ghstack-source-id: 216481872 exported-using-ghexport Reviewed By: SS-JIA Differential Revision: D54042646 fbshipit-source-id: a8965d69c94cdb3fe9837e9b83b6db8a877949f0
kirklandsign
pushed a commit
that referenced
this pull request
Oct 22, 2024
…ytorch#6407) * Main backup (#12) * Add nnlib as submodule * Adding nnlib submodule * Integrated nnlib API unde backends/cadence/hifi * Fix review comments on PR#3 * Add nnlib as submodule * Adding nnlib submodule * Integrated nnlib API unde backends/cadence/hifi * Fix review comments on PR#3 * Incorporated feedback from Meta team. * lint errors fixed * Adding Sub operator optimized version * Add optimization for add, mul operators * Adding Div operator * Modified div mod to cover truncate and floor modes --------- Co-authored-by: cad-audio <[email protected]> Co-authored-by: cad-audio <[email protected]> * Adding sigmoid optimizations * Adding tanh optimizations * Fixing review comments in 5483 * Adding cflags to prevent compilation halts * Adding cflags to prevent compilation halts * Changing name space of optimized ops; Remove unused ops from file * fixed lint issues. * Namespace updates for cadence ops, adding 6 optimized ops --------- Co-authored-by: cad-audio <[email protected]> Co-authored-by: cad-audio <[email protected]>
kirklandsign
pushed a commit
that referenced
this pull request
Nov 21, 2024
* Main backup (#12) * Add nnlib as submodule * Adding nnlib submodule * Integrated nnlib API unde backends/cadence/hifi * Fix review comments on PR#3 * Add nnlib as submodule * Adding nnlib submodule * Integrated nnlib API unde backends/cadence/hifi * Fix review comments on PR#3 * Incorporated feedback from Meta team. * lint errors fixed * Adding Sub operator optimized version * Add optimization for add, mul operators * Adding Div operator * Modified div mod to cover truncate and floor modes --------- Co-authored-by: cad-audio <[email protected]> Co-authored-by: cad-audio <[email protected]> * Adding sigmoid optimizations * Adding tanh optimizations * Fixing review comments in 5483 * Adding cflags to prevent compilation halts * Adding cflags to prevent compilation halts * Changing name space of optimized ops; Remove unused ops from file * fixed lint issues. * Namespace updates for cadence ops, adding 6 optimized ops * Updating add op with namespace change * Updating add op with namespace change, fix build issue --------- Co-authored-by: cad-audio <[email protected]> Co-authored-by: cad-audio <[email protected]>
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.
See Commits and Changes for more details.
Created by
pull[bot]
Can you help keep this open source service alive? 💖 Please sponsor : )