Implement load_into for Mmap Data Loader by keyprocedure · Pull Request #11654 · pytorch/executorch

keyprocedure · 2025-06-13T22:12:32Z

Summary

Adds load_into() to support memory mapped loading directly into the caller's buffer.
Enables copying a specific byte range from the file without creating an internal buffer.
Resuses input validation from load().

Test plan

Added unit tests to MmapDataLoaderTest for load_into() to validate copying for both aligned and offset data.
All Mmap Data Loader tests pass via:

./build-ninja/extension/data_loader/test/extension_data_loader_test --gtest_filter='MmapDataLoaderTest.*'

Full set of Data Loader tests pass via:

./build-ninja/extension/data_loader/test/extension_data_loader_test

pytorch-bot · 2025-06-13T22:12:36Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/11654

📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❌ 1 New Failure

As of commit 5589d40 with merge base 7bd15b9 ():

NEW FAILURE - The following job has failed:

pull / test-moshi-linux / linux-job (gh)
test_exported_decoder_xnnpack

This comment was automatically generated by Dr. CI and updates every 15 minutes.

keyprocedure · 2025-06-13T22:16:42Z

@pytorchbot label "release notes: none"

extension/data_loader/mmap_data_loader.cpp

JacobSzwejbka · 2025-06-17T05:17:42Z

extension/data_loader/mmap_data_loader.h


  ET_NODISCARD executorch::runtime::Result<size_t> size() const override;

+  ET_NODISCARD executorch::runtime::Error validate_input(


nit: Can you make this private

Done, moved to private section

JacobSzwejbka · 2025-06-18T17:17:58Z

just letting ci run again and then will merge

keyprocedure · 2025-06-18T20:07:37Z

CI failure looks unrelated: MimiModel is missing _context_for_encoder_decoder

JacobSzwejbka · 2025-06-23T16:44:29Z

Thanks for contributing!

keyprocedure · 2025-06-23T19:35:16Z

I'm happy to help! Thanks for the review.

This reverts commit 18e4240.

### Summary Revert #11654 and #11858 due to Meta-internal build breakage. I will re-land these (preserving attribution).

### Summary - Adds `load_into()` to support memory mapped loading directly into the caller's buffer. - Enables copying a specific byte range from the file without creating an internal buffer. - Resuses input validation from `load()`. Fixes pytorch#11561 ### Test plan - Added unit tests to `MmapDataLoaderTest` for `load_into()` to validate copying for both aligned and offset data. - All Mmap Data Loader tests pass via: ``` ./build-ninja/extension/data_loader/test/extension_data_loader_test --gtest_filter='MmapDataLoaderTest.*' ``` - Full set of Data Loader tests pass via: ``` ./build-ninja/extension/data_loader/test/extension_data_loader_test ```

GregoryComer · 2025-06-24T22:25:02Z

FYI I temporarily reverted this due to a meta-internal build issue. Sorry about that. I'm going to re-merge it shortly, preserving commit authorship. We'll also fix the CI so that it builds with -Wshadow.

keyprocedure · 2025-06-24T22:54:19Z

Got it. Let me know if you'd like me to change anything.

### Summary - Adds `load_into()` to support memory mapped loading directly into the caller's buffer. - Enables copying a specific byte range from the file without creating an internal buffer. - Resuses input validation from `load()`. Fixes #11561 ### Test plan - Added unit tests to `MmapDataLoaderTest` for `load_into()` to validate copying for both aligned and offset data. - All Mmap Data Loader tests pass via: ``` ./build-ninja/extension/data_loader/test/extension_data_loader_test --gtest_filter='MmapDataLoaderTest.*' ``` - Full set of Data Loader tests pass via: ``` ./build-ninja/extension/data_loader/test/extension_data_loader_test ```

### Summary - Adds `load_into()` to support memory mapped loading directly into the caller's buffer. - Enables copying a specific byte range from the file without creating an internal buffer. - Resuses input validation from `load()`. Fixes pytorch#11561 ### Test plan - Added unit tests to `MmapDataLoaderTest` for `load_into()` to validate copying for both aligned and offset data. - All Mmap Data Loader tests pass via: ``` ./build-ninja/extension/data_loader/test/extension_data_loader_test --gtest_filter='MmapDataLoaderTest.*' ``` - Full set of Data Loader tests pass via: ``` ./build-ninja/extension/data_loader/test/extension_data_loader_test ```

### Summary Revert pytorch#11654 and pytorch#11858 due to Meta-internal build breakage. I will re-land these (preserving attribution).

### Summary - Adds `load_into()` to support memory mapped loading directly into the caller's buffer. - Enables copying a specific byte range from the file without creating an internal buffer. - Resuses input validation from `load()`. Fixes pytorch#11561 ### Test plan - Added unit tests to `MmapDataLoaderTest` for `load_into()` to validate copying for both aligned and offset data. - All Mmap Data Loader tests pass via: ``` ./build-ninja/extension/data_loader/test/extension_data_loader_test --gtest_filter='MmapDataLoaderTest.*' ``` - Full set of Data Loader tests pass via: ``` ./build-ninja/extension/data_loader/test/extension_data_loader_test ```

keyprocedure added 2 commits June 13, 2025 14:43

Implement load_into(); Add unit tests

9cfd878

Apply lintrunner formatting

0b3e9ac

keyprocedure requested review from JacobSzwejbka and lucylq as code owners June 13, 2025 22:12

facebook-github-bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Jun 13, 2025

pytorch-bot bot added the release notes: none Do not include this in the release notes label Jun 13, 2025

keyprocedure changed the title ~~Implement load_into to Mmap Data Loader~~ Implement load_into for Mmap Data Loader Jun 13, 2025

mergennachin mentioned this pull request Jun 16, 2025

Implement load_into for SharedPtrDataLoader and add test (#11562) #11707

Open

JacobSzwejbka requested changes Jun 16, 2025

View reviewed changes

extension/data_loader/mmap_data_loader.cpp Outdated Show resolved Hide resolved

extension/data_loader/mmap_data_loader.cpp Show resolved Hide resolved

Remove mlock from load_into()

3779d45

JacobSzwejbka requested changes Jun 17, 2025

View reviewed changes

keyprocedure added 3 commits June 17, 2025 08:23

Move validate_input to private section

17b3599

Update comments

a972fa8

Merge branch 'main' into add-mmap-load-into

5589d40

JacobSzwejbka approved these changes Jun 18, 2025

View reviewed changes

JacobSzwejbka merged commit 18e4240 into pytorch:main Jun 23, 2025
95 of 96 checks passed

GregoryComer added a commit to GregoryComer/executorch that referenced this pull request Jun 24, 2025

Revert "Implement load_into for Mmap Data Loader (pytorch#11654)"

a285663

This reverts commit 18e4240.

GregoryComer mentioned this pull request Jun 24, 2025

Revert mmap loader changes #11928

Merged

GregoryComer added a commit that referenced this pull request Jun 24, 2025

Revert mmap loader changes (#11928)

d5fe5fa

### Summary Revert #11654 and #11858 due to Meta-internal build breakage. I will re-land these (preserving attribution).

GregoryComer mentioned this pull request Jun 24, 2025

Reland "Implement load_into for Mmap Data Loader" #11935

Merged

hinriksnaer pushed a commit to hinriksnaer/executorch that referenced this pull request Jun 26, 2025

Revert mmap loader changes (pytorch#11928)

6b314e5

### Summary Revert pytorch#11654 and pytorch#11858 due to Meta-internal build breakage. I will re-land these (preserving attribution).

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Implement load_into for Mmap Data Loader#11654

Implement load_into for Mmap Data Loader#11654
JacobSzwejbka merged 6 commits intopytorch:mainfrom
keyprocedure:add-mmap-load-into

keyprocedure commented Jun 13, 2025 •

edited

Loading

Uh oh!

pytorch-bot bot commented Jun 13, 2025 •

edited

Loading

Uh oh!

keyprocedure commented Jun 13, 2025

Uh oh!

Uh oh!

Uh oh!

JacobSzwejbka Jun 17, 2025

Uh oh!

keyprocedure Jun 17, 2025

Uh oh!

JacobSzwejbka commented Jun 18, 2025

Uh oh!

keyprocedure commented Jun 18, 2025

Uh oh!

Uh oh!

JacobSzwejbka commented Jun 23, 2025

Uh oh!

keyprocedure commented Jun 23, 2025

Uh oh!

GregoryComer commented Jun 24, 2025

Uh oh!

keyprocedure commented Jun 24, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants


		ET_NODISCARD executorch::runtime::Result<size_t> size() const override;

		ET_NODISCARD executorch::runtime::Error validate_input(

Conversation

keyprocedure commented Jun 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test plan

Uh oh!

pytorch-bot bot commented Jun 13, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/11654

❌ 1 New Failure

Uh oh!

keyprocedure commented Jun 13, 2025

Uh oh!

Uh oh!

Uh oh!

JacobSzwejbka Jun 17, 2025

Choose a reason for hiding this comment

Uh oh!

keyprocedure Jun 17, 2025

Choose a reason for hiding this comment

Uh oh!

JacobSzwejbka commented Jun 18, 2025

Uh oh!

keyprocedure commented Jun 18, 2025

Uh oh!

Uh oh!

JacobSzwejbka commented Jun 23, 2025

Uh oh!

keyprocedure commented Jun 23, 2025

Uh oh!

GregoryComer commented Jun 24, 2025

Uh oh!

keyprocedure commented Jun 24, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

keyprocedure commented Jun 13, 2025 •

edited

Loading

pytorch-bot bot commented Jun 13, 2025 •

edited

Loading