add cpp serializer for flat tensor (ptd) #7841
Merged
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
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/7841
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit c35b15e with merge base 1f1a96f ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D67992901 |
JacobSzwejbka
added
the
release notes: api
dbort
requested changes
Jan 22, 2025
JacobSzwejbka
force-pushed
the
export-D67992901
branch
from
376e7f8 to
ec6e3c8
Compare
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Jan 23, 2025
Summary: Leverage the flatbuffer builder apis to generate a .ptd on device. This will be used by training for checkpointing. No other use cases really exist for generating a .ptd on device right now, so I didnt worry about making this more easily extensible by coming up with a cpp equivalent of a cord or by trying to think ahead of how this might integrate with delegates. Later if we add support for delegates owning the weights under ET training we can revisit this. Differential Revision: D67992901
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D67992901 |
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Jan 23, 2025
Summary: Leverage the flatbuffer builder apis to generate a .ptd on device. This will be used by training for checkpointing. No other use cases really exist for generating a .ptd on device right now, so I didnt worry about making this more easily extensible by coming up with a cpp equivalent of a cord or by trying to think ahead of how this might integrate with delegates. Later if we add support for delegates owning the weights under ET training we can revisit this. Differential Revision: D67992901
JacobSzwejbka
force-pushed
the
export-D67992901
branch
from
ec6e3c8 to
fd69719
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D67992901 |
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Jan 23, 2025
Summary: Leverage the flatbuffer builder apis to generate a .ptd on device. This will be used by training for checkpointing. No other use cases really exist for generating a .ptd on device right now, so I didnt worry about making this more easily extensible by coming up with a cpp equivalent of a cord or by trying to think ahead of how this might integrate with delegates. Later if we add support for delegates owning the weights under ET training we can revisit this. Differential Revision: D67992901
JacobSzwejbka
force-pushed
the
export-D67992901
branch
from
fd69719 to
978d9ce
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D67992901 |
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Jan 23, 2025
Summary: Leverage the flatbuffer builder apis to generate a .ptd on device. This will be used by training for checkpointing. No other use cases really exist for generating a .ptd on device right now, so I didnt worry about making this more easily extensible by coming up with a cpp equivalent of a cord or by trying to think ahead of how this might integrate with delegates. Later if we add support for delegates owning the weights under ET training we can revisit this. Differential Revision: D67992901
JacobSzwejbka
force-pushed
the
export-D67992901
branch
from
978d9ce to
4a8ff3b
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D67992901 |
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Jan 23, 2025
Summary: Leverage the flatbuffer builder apis to generate a .ptd on device. This will be used by training for checkpointing. No other use cases really exist for generating a .ptd on device right now, so I didnt worry about making this more easily extensible by coming up with a cpp equivalent of a cord or by trying to think ahead of how this might integrate with delegates. Later if we add support for delegates owning the weights under ET training we can revisit this. Differential Revision: D67992901
JacobSzwejbka
force-pushed
the
export-D67992901
branch
from
4a8ff3b to
f5916c5
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D67992901 |
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Jan 23, 2025
Summary: Leverage the flatbuffer builder apis to generate a .ptd on device. This will be used by training for checkpointing. No other use cases really exist for generating a .ptd on device right now, so I didnt worry about making this more easily extensible by coming up with a cpp equivalent of a cord or by trying to think ahead of how this might integrate with delegates. Later if we add support for delegates owning the weights under ET training we can revisit this. Differential Revision: D67992901
JacobSzwejbka
force-pushed
the
export-D67992901
branch
from
f5916c5 to
11d2496
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D67992901 |
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Jan 23, 2025
Summary: Leverage the flatbuffer builder apis to generate a .ptd on device. This will be used by training for checkpointing. No other use cases really exist for generating a .ptd on device right now, so I didnt worry about making this more easily extensible by coming up with a cpp equivalent of a cord or by trying to think ahead of how this might integrate with delegates. Later if we add support for delegates owning the weights under ET training we can revisit this. Differential Revision: D67992901
JacobSzwejbka
force-pushed
the
export-D67992901
branch
from
11d2496 to
a2012d9
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D67992901 |
dbort
requested changes
Jan 24, 2025
Contributor
dbort
left a comment
dbort
left a comment
There was a problem hiding this comment.
Thanks for the changes -- the tests look great
JacobSzwejbka
force-pushed
the
export-D67992901
branch
from
a2012d9 to
32223b4
Compare
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Jan 24, 2025
Summary: Leverage the flatbuffer builder apis to generate a .ptd on device. This will be used by training for checkpointing. No other use cases really exist for generating a .ptd on device right now, so I didnt worry about making this more easily extensible by coming up with a cpp equivalent of a cord or by trying to think ahead of how this might integrate with delegates. Later if we add support for delegates owning the weights under ET training we can revisit this. Differential Revision: D67992901
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D67992901 |
dbort
approved these changes
Jan 24, 2025
Contributor
dbort
left a comment
dbort
left a comment
There was a problem hiding this comment.
Thanks for all of the changes -- looks great apart from the new bugs.
Comment on lines
94
to
99
| if (i != 1) { | ||
| // Precalculate the size of the data blob. | ||
| total_segment_size += aligned_size(tensor.nbytes(), tensor_alignment); | ||
| } |
Contributor
There was a problem hiding this comment.
Missing the --i like in the loop below. Please add a test that would have caught these issues.
Contributor
Author
There was a problem hiding this comment.
Ah the existing test does catch this bug. I ran the wrong test (python) before pushing this version
| # points to all the data segments. It will be serialized to flatbuffer. | ||
| flat_tensor = FlatTensor( | ||
| version=0, | ||
| version=0, # Keep in sync with c++ version number in serialize.h |
Contributor
There was a problem hiding this comment.
Consider adding a top-level _SCHEMA_VERSION: int = 0 to raise visibility of this, and to make it map more clearly to the C++ equivalent.
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Jan 24, 2025
Summary: Leverage the flatbuffer builder apis to generate a .ptd on device. This will be used by training for checkpointing. No other use cases really exist for generating a .ptd on device right now, so I didnt worry about making this more easily extensible by coming up with a cpp equivalent of a cord or by trying to think ahead of how this might integrate with delegates. Later if we add support for delegates owning the weights under ET training we can revisit this. Differential Revision: D67992901
JacobSzwejbka
force-pushed
the
export-D67992901
branch
from
32223b4 to
fc6046c
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D67992901 |
Summary: Leverage the flatbuffer builder apis to generate a .ptd on device. This will be used by training for checkpointing. No other use cases really exist for generating a .ptd on device right now, so I didnt worry about making this more easily extensible by coming up with a cpp equivalent of a cord or by trying to think ahead of how this might integrate with delegates. Later if we add support for delegates owning the weights under ET training we can revisit this. Differential Revision: D67992901
JacobSzwejbka
force-pushed
the
export-D67992901
branch
from
fc6046c to
c35b15e
Compare
Contributor
|
This pull request was exported from Phabricator. Differential Revision: D67992901 |
YIWENX14
pushed a commit
that referenced
this pull request
Jan 28, 2025
Differential Revision: D67992901 Pull Request resolved: #7841
zonglinpeng
pushed a commit
to zonglinpeng/executorch
that referenced
this pull request
Jan 30, 2025
Differential Revision: D67992901 Pull Request resolved: pytorch#7841
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
Summary:
Leverage the flatbuffer builder apis to generate a .ptd on device. This will be used by training for checkpointing.
No other use cases really exist for generating a .ptd on device right now, so I didnt worry about making this more easily extensible by coming up with a cpp equivalent of a cord or by trying to think ahead of how this might integrate with delegates.
Later if we add support for delegates owning the weights under ET training we can revisit this.
Differential Revision: D67992901