Arm backend: Add bfloat16 support to VGF backend.#20299
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/20299
Note: Links to docs will display an error until the docs builds have been completed. ❗ 2 Active SEVsThere are 2 currently active SEVs. If your PR is affected, please view them below:
❌ 2 New Failures, 1 Unrelated Failure, 2 Unclassified FailuresAs of commit f015b75 with merge base da9158b ( NEW FAILURES - The following jobs have failed:
UNCLASSIFIED FAILURES - DrCI could not classify the following jobs because the workflow did not run on the merge base. The failures may be pre-existing on trunk or introduced by this PR:
BROKEN TRUNK - The following job failed but were present on the merge base:👉 Rebase onto the `viable/strict` branch to avoid these failures
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
volkkov
left a comment
There was a problem hiding this comment.
Good improvement overall. One edge case: empty string vs null aren't treated the same downstream — might need a normalization step.
Thanks for the comment volkkov, could you please be a bit more specific? I.e. which string |
0650cd0 to
cf8b909
Compare
- Add bf16 extension to default VgfCompileSpec - Handle bf16 in VGFSetup.sh - Needs bumping of Vulkan SDK to 1.4.350.0 to include VK_FORMAT_R16_SFLOAT_FPENCODING_BFLOAT16_ARM Initially tested with a single operator test of matmul. Signed-off-by: Erik Lundell <erik.lundell@arm.com> Change-Id: I74b0c15b5a4f9194c437e8e69d2349e9c282878b
digantdesai
left a comment
There was a problem hiding this comment.
LGTM, I will let @SS-JIA stamp it due to the submodule update.
|
Hi @SS-JIA what do you think about this? |
Initially tested with a single operator test of matmul.
cc @digantdesai @freddan80 @per @zingo @oscarandersson8218 @mansnils @Sebastian-Larsson @robell @rascani