emit programs with mutable buffers #2233
Closed
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/2233
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 97279b7 with merge base e7197a1 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
|
This pull request was exported from Phabricator. Differential Revision: D53713544 |
tarun292
approved these changes
Mar 4, 2024
JacobSzwejbka
force-pushed
the
export-D53713544
branch
from
1fed8e9 to
b583417
Compare
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Mar 4, 2024
Summary: Meaningful changes to the emitter logic here. Before we would ignore the tensor spec passed in and try to decide if the placeholder was a constant and if it was we would create a new spec from the actual value for that constant. That drops meta data on the input spec which is not great. Now instead of that we just look up the storage of the concrete tensor and hook it up to the spec. Also added some logic to seperate out behavior for mutable buffers specifically. While working on this I also discovered a bug that memory planning is planning space for parameters and constant buffers if its told to allocate space for inputs which is really bad lol. Oh one big assumption this diff makes is that the buffer does not have a meaningful initial state. I should probably throw out a warning during emission about this in the short term. Long term we will handle them properly. bypass-github-export-checks Reviewed By: tarun292 Differential Revision: D53713544
|
This pull request was exported from Phabricator. Differential Revision: D53713544 |
JacobSzwejbka
force-pushed
the
export-D53713544
branch
from
b583417 to
395c95e
Compare
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Mar 5, 2024
Summary: Meaningful changes to the emitter logic here. Before we would ignore the tensor spec passed in and try to decide if the placeholder was a constant and if it was we would create a new spec from the actual value for that constant. That drops meta data on the input spec which is not great. Now instead of that we just look up the storage of the concrete tensor and hook it up to the spec. Also added some logic to seperate out behavior for mutable buffers specifically. While working on this I also discovered a bug that memory planning is planning space for parameters and constant buffers if its told to allocate space for inputs which is really bad lol. Oh one big assumption this diff makes is that the buffer does not have a meaningful initial state. I should probably throw out a warning during emission about this in the short term. Long term we will handle them properly. bypass-github-export-checks Reviewed By: tarun292 Differential Revision: D53713544
|
This pull request was exported from Phabricator. Differential Revision: D53713544 |
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Mar 5, 2024
Summary: Meaningful changes to the emitter logic here. Before we would ignore the tensor spec passed in and try to decide if the placeholder was a constant and if it was we would create a new spec from the actual value for that constant. That drops meta data on the input spec which is not great. Now instead of that we just look up the storage of the concrete tensor and hook it up to the spec. Also added some logic to seperate out behavior for mutable buffers specifically. While working on this I also discovered a bug that memory planning is planning space for parameters and constant buffers if its told to allocate space for inputs which is really bad lol. Oh one big assumption this diff makes is that the buffer does not have a meaningful initial state. I should probably throw out a warning during emission about this in the short term. Long term we will handle them properly. bypass-github-export-checks Reviewed By: tarun292 Differential Revision: D53713544
JacobSzwejbka
force-pushed
the
export-D53713544
branch
from
395c95e to
b77fd57
Compare
|
This pull request was exported from Phabricator. Differential Revision: D53713544 |
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Mar 5, 2024
Summary: Meaningful changes to the emitter logic here. Before we would ignore the tensor spec passed in and try to decide if the placeholder was a constant and if it was we would create a new spec from the actual value for that constant. That drops meta data on the input spec which is not great. Now instead of that we just look up the storage of the concrete tensor and hook it up to the spec. Also added some logic to seperate out behavior for mutable buffers specifically. While working on this I also discovered a bug that memory planning is planning space for parameters and constant buffers if its told to allocate space for inputs which is really bad lol. Oh one big assumption this diff makes is that the buffer does not have a meaningful initial state. I should probably throw out a warning during emission about this in the short term. Long term we will handle them properly. bypass-github-export-checks Reviewed By: tarun292 Differential Revision: D53713544
JacobSzwejbka
force-pushed
the
export-D53713544
branch
from
b77fd57 to
ad047a7
Compare
|
This pull request was exported from Phabricator. Differential Revision: D53713544 |
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Mar 5, 2024
Summary: Meaningful changes to the emitter logic here. Before we would ignore the tensor spec passed in and try to decide if the placeholder was a constant and if it was we would create a new spec from the actual value for that constant. That drops meta data on the input spec which is not great. Now instead of that we just look up the storage of the concrete tensor and hook it up to the spec. Also added some logic to seperate out behavior for mutable buffers specifically. While working on this I also discovered a bug that memory planning is planning space for parameters and constant buffers if its told to allocate space for inputs which is really bad lol. Oh one big assumption this diff makes is that the buffer does not have a meaningful initial state. I should probably throw out a warning during emission about this in the short term. Long term we will handle them properly. bypass-github-export-checks Reviewed By: tarun292 Differential Revision: D53713544
JacobSzwejbka
force-pushed
the
export-D53713544
branch
from
ad047a7 to
d22066b
Compare
|
This pull request was exported from Phabricator. Differential Revision: D53713544 |
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Mar 5, 2024
Summary: Meaningful changes to the emitter logic here. Before we would ignore the tensor spec passed in and try to decide if the placeholder was a constant and if it was we would create a new spec from the actual value for that constant. That drops meta data on the input spec which is not great. Now instead of that we just look up the storage of the concrete tensor and hook it up to the spec. Also added some logic to seperate out behavior for mutable buffers specifically. While working on this I also discovered a bug that memory planning is planning space for parameters and constant buffers if its told to allocate space for inputs which is really bad lol. Oh one big assumption this diff makes is that the buffer does not have a meaningful initial state. I should probably throw out a warning during emission about this in the short term. Long term we will handle them properly. bypass-github-export-checks Reviewed By: tarun292 Differential Revision: D53713544
JacobSzwejbka
force-pushed
the
export-D53713544
branch
from
d22066b to
3ea027e
Compare
|
This pull request was exported from Phabricator. Differential Revision: D53713544 |
JacobSzwejbka
force-pushed
the
export-D53713544
branch
from
3ea027e to
8bd7b83
Compare
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Mar 5, 2024
Summary: Meaningful changes to the emitter logic here. Before we would ignore the tensor spec passed in and try to decide if the placeholder was a constant and if it was we would create a new spec from the actual value for that constant. That drops meta data on the input spec which is not great. Now instead of that we just look up the storage of the concrete tensor and hook it up to the spec. Also added some logic to seperate out behavior for mutable buffers specifically. While working on this I also discovered a bug that memory planning is planning space for parameters and constant buffers if its told to allocate space for inputs which is really bad lol. Oh one big assumption this diff makes is that the buffer does not have a meaningful initial state. I should probably throw out a warning during emission about this in the short term. Long term we will handle them properly. bypass-github-export-checks Reviewed By: tarun292 Differential Revision: D53713544
|
This pull request was exported from Phabricator. Differential Revision: D53713544 |
JacobSzwejbka
added a commit
to JacobSzwejbka/executorch-1
that referenced
this pull request
Mar 5, 2024
Summary: Meaningful changes to the emitter logic here. Before we would ignore the tensor spec passed in and try to decide if the placeholder was a constant and if it was we would create a new spec from the actual value for that constant. That drops meta data on the input spec which is not great. Now instead of that we just look up the storage of the concrete tensor and hook it up to the spec. Also added some logic to seperate out behavior for mutable buffers specifically. While working on this I also discovered a bug that memory planning is planning space for parameters and constant buffers if its told to allocate space for inputs which is really bad lol. Oh one big assumption this diff makes is that the buffer does not have a meaningful initial state. I should probably throw out a warning during emission about this in the short term. Long term we will handle them properly. bypass-github-export-checks Reviewed By: tarun292 Differential Revision: D53713544
JacobSzwejbka
force-pushed
the
export-D53713544
branch
from
8bd7b83 to
74a1da8
Compare
|
This pull request was exported from Phabricator. Differential Revision: D53713544 |
Summary: Meaningful changes to the emitter logic here. Before we would ignore the tensor spec passed in and try to decide if the placeholder was a constant and if it was we would create a new spec from the actual value for that constant. That drops meta data on the input spec which is not great. Now instead of that we just look up the storage of the concrete tensor and hook it up to the spec. Also added some logic to seperate out behavior for mutable buffers specifically. While working on this I also discovered a bug that memory planning is planning space for parameters and constant buffers if its told to allocate space for inputs which is really bad lol. Oh one big assumption this diff makes is that the buffer does not have a meaningful initial state. I should probably throw out a warning during emission about this in the short term. Long term we will handle them properly. bypass-github-export-checks Reviewed By: tarun292 Differential Revision: D53713544
JacobSzwejbka
force-pushed
the
export-D53713544
branch
from
74a1da8 to
97279b7
Compare
|
This pull request was exported from Phabricator. Differential Revision: D53713544 |
|
This pull request has been merged in aef3a7c. |
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.
Summary:
Meaningful changes to the emitter logic here.
Before we would ignore the tensor spec passed in and try to decide if the placeholder was a constant and if it was we would create a new spec from the actual value for that constant. That drops meta data on the input spec which is not great. Now instead of that we just look up the storage of the concrete tensor and hook it up to the spec.
Also added some logic to seperate out behavior for mutable buffers specifically.
While working on this I also discovered a bug that memory planning is planning space for parameters and constant buffers if its told to allocate space for inputs which is really bad lol.
Oh one big assumption this diff makes is that the buffer does not have a meaningful initial state. I should probably throw out a warning during emission about this in the short term. Long term we will handle them properly.
Reviewed By: tarun292
Differential Revision: D53713544