[Llava] Add csv image loading in C++ runner #5380
Open
Conversation
This file contains 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/5380
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ✅ No FailuresAs of commit 6c1b6b2 with merge base 444480b ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
facebook-github-bot
added
the
CLA Signed
digantdesai
force-pushed
the
llava_csv_image
branch
2 times, most recently
from
6b66a90 to
0944ca3
Compare
|
@digantdesai has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
|
CSV seems like an unfortunate choice of image format. Why not PNG (or any other image format -- load the image and then blast it into a Tensor)? |
digantdesai
force-pushed
the
llava_csv_image
branch
from
0944ca3 to
0874123
Compare
Rationale - We don't want to depend on Torch when building for Android. This add two things, (1) for AoT, python image_util optionally generates a .csv from a jpg. In addition to .pt. (2) add a runtime runner flag which hints at the provided image is a csv. And if so, the runner parses the csv and feeds it to the model. This is very naive and obviously fragile. Added some checks in python. Tested few ways, - On M1, with torch, loaded both .pt and .csv generated from the same jpg. And the LLM produces same text. - On Android, without torch, loaded .csv and it also produces similar text.:wq
digantdesai
force-pushed
the
llava_csv_image
branch
from
0874123 to
6c1b6b2
Compare
|
@digantdesai has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator. |
larryliu0820
approved these changes
Sep 18, 2024
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.
Rationale - We don't want to depend on Torch when building for Android.
This add two things,
(1) for AoT, python image_util optionally generates a .csv from a jpg. In addition to .pt.
(2) add a runtime runner flag which hints at the provided image is a csv. And if so, the runner parses the csv and feeds it to the model.
This is very naive and obviously fragile. Added some checks in python.
Tested few ways,