[llm] Update metadata max_seq_len based on the max range of dynamic shapes #11611
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/11611
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit db863ef with merge base a6d8440 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
|
@larryliu0820 has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
larryliu0820
added
the
release notes: none
larryliu0820
force-pushed
the
larryliu0820-patch-9
branch
from
e48419f to
b45e4c1
Compare
|
@larryliu0820 has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
facebook-github-bot
pushed a commit
that referenced
this pull request
Jun 12, 2025
Summary: See if any CI is broken by this. Differential Revision: D76530379 Pulled By: larryliu0820
facebook-github-bot
force-pushed
the
larryliu0820-patch-9
branch
from
b45e4c1 to
a897dae
Compare
|
This pull request was exported from Phabricator. Differential Revision: D76530379 |
facebook-github-bot
pushed a commit
that referenced
this pull request
Jun 13, 2025
Summary: With some testing it seems like we can't export with token dimension max being `max_seq_len` in dynamic shape, if we only export with `tokens`. However if we export with both `tokens` and `input_pos`, we can set token dimension max value to be `max_seq_len`. This diff fix 2 things: * Change dynamic shape based on different inputs * Change pte's metadata `get_max_seq_len` and `get_max_context_len` based on the value of token dimension max value in dynamic shape. Differential Revision: D76530379 Pulled By: larryliu0820
facebook-github-bot
force-pushed
the
larryliu0820-patch-9
branch
from
a897dae to
d2a5153
Compare
|
This pull request was exported from Phabricator. Differential Revision: D76530379 |
facebook-github-bot
pushed a commit
that referenced
this pull request
Jun 17, 2025
Summary: With some testing it seems like we can't export with token dimension max being `max_seq_len` in dynamic shape, if we only export with `tokens`. However if we export with both `tokens` and `input_pos`, we can set token dimension max value to be `max_seq_len`. This diff fix 2 things: * Change dynamic shape based on different inputs * Change pte's metadata `get_max_seq_len` and `get_max_context_len` based on the value of token dimension max value in dynamic shape. Test Plan: Imported from GitHub, without a `Test Plan:` line. To repro the issue with max value being `max_seq_len` for `tokens` only argument, change the `_get_dynamic_shape()` and run: ``` buck run fbcode//executorch/examples/models/llama/fb:cria05 -- cria_0_5b -E 8,0 -d fp32 cria.pte ``` Differential Revision: D76530379 Pulled By: larryliu0820
facebook-github-bot
force-pushed
the
larryliu0820-patch-9
branch
from
d2a5153 to
695adc7
Compare
|
This pull request was exported from Phabricator. Differential Revision: D76530379 |
larryliu0820
changed the title
|
@larryliu0820 has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
kimishpatel
approved these changes
Jun 17, 2025
facebook-github-bot
pushed a commit
that referenced
this pull request
Jun 17, 2025
…11611) Summary: With some testing it seems like we can't export with token dimension max being `max_seq_len` in dynamic shape, if we only export with `tokens`. However if we export with both `tokens` and `input_pos`, we can set token dimension max value to be `max_seq_len`. This diff fix 2 things: * Change dynamic shape based on different inputs * Change pte's metadata `get_max_seq_len` and `get_max_context_len` based on the value of token dimension max value in dynamic shape. Reviewed By: kimishpatel Differential Revision: D76530379 Pulled By: larryliu0820
facebook-github-bot
force-pushed
the
larryliu0820-patch-9
branch
from
0b301b7 to
d7d00f6
Compare
|
This pull request was exported from Phabricator. Differential Revision: D76530379 |
…11611) Summary: With some testing it seems like we can't export with token dimension max being `max_seq_len` in dynamic shape, if we only export with `tokens`. However if we export with both `tokens` and `input_pos`, we can set token dimension max value to be `max_seq_len`. This diff fix 2 things: * Change dynamic shape based on different inputs * Change pte's metadata `get_max_seq_len` and `get_max_context_len` based on the value of token dimension max value in dynamic shape. Reviewed By: kimishpatel Differential Revision: D76530379 Pulled By: larryliu0820
facebook-github-bot
force-pushed
the
larryliu0820-patch-9
branch
from
d7d00f6 to
db863ef
Compare
|
This pull request was exported from Phabricator. Differential Revision: D76530379 |
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.
With some testing it seems like we can't export with token dimension max being
max_seq_lenin dynamic shape, if we only export withtokens.However if we export with both
tokensandinput_pos, we can set token dimension max value to bemax_seq_len.This diff fix 2 things:
get_max_seq_lenandget_max_context_lenbased on the value of token dimension max value in dynamic shape.