IBM / unitxt / 13183672826

81%

push

github

Fixes to vision with WML (#1578)

* minor bug fixes

* adding anls

* Added LMMS_eval template, exact_match_mm metric, RGB image augmentor.

* Added qa.with_context_multiple choice lmms_eval template

* Added relaxed_correctness metric, edited info_vqa card, edit lmms_eval template

* minor script changes

* minor run changes

* Added default template

* Added lmms-lab version of chartqa with default template and relaxed correctness metric and with_type task

* Added lmms-lab card of info_vqa. added default template to info_vqa and docvqa

* Wevsrc new with_domain task, new metric. WIP

* websrc working

* discarding not important changes from main

* workaround for llama_vision with WML

* WML vs local UNITXT reproductions

* Added llama vision benchmark

* Changes for commit

* fix unitxt typo

* ruff

* Update inference engine model and adjust expected targets in tests

Signed-off-by: elronbandel <elronbandel@gmail.com>

* Fix WML Inference Engine tests for images

Signed-off-by: elronbandel <elronbandel@gmail.com>

* Update tests

Signed-off-by: elronbandel <elronbandel@gmail.com>

* Enhance CSV loader with low_memory option and update inference engine tests for set equality

Signed-off-by: elronbandel <elronbandel@gmail.com>

* Add error handling in Loader class to raise UnitxtError on load_iterables failure

Signed-off-by: elronbandel <elronbandel@gmail.com>

---------

Signed-off-by: elronbandel <elronbandel@gmail.com>
Co-authored-by: Amit Alfassy Amit.Alfassy@ibm.com <amit.alfassy@ibm.com>
Co-authored-by: Elron Bandel <elronbandel@gmail.com>

1454 of 1833 branches covered (79.32%)

Branch coverage included in aggregate %.

9192 of 11660 relevant lines covered (78.83%)

0.79 hits per line