audio tester class #45391
eustlb
merged
42 commits into
main
from
tarekziade-audio-test
audio tester
3562c7f1
tweak check repo for audio tester
0817bdbd
audio -> ALM
356c922e
zucchini-nlp
commented
on 2026-04-13
zucchini-nlp
commented
on 2026-04-13
ALMTester: no audio/text defaults; better input prep
9663a8e5
Merge branch 'main' into tarekziade-audio-test
73c45481
udpate test_sdpa_can_dispatch_composite_models to hanlde ALMs
a599b1de
propagate to other model classes
a7d54dc5
cleaner
a302c3ec
updates
8fcba58d
audio_mask_key + updates
66acc9ed
typo
63ca77e0
simplify granite speech
7588135e
nits
41fed1c8
some more cleaning
e5971c7f
add test_mismatching_num_audio_tokens
59703ddd
add get_placeholder_mask
6a67f32b
specific to musicflamingo
b59f9583
granite speech fix
bb986b66
let's factorise alm/vlm testers
670c68c2
make fix-repo
c9534432
eustlb
commented
on 2026-04-22
unskip test_sdpa_can_dispatch_on_flash on qwen2_audio
87404099
should not be skipped
dde65f61
make fix-repo
19b37c5a
test_mismatching_num_audio_tokens should be skipped for voxtral_realtime
b47621a9
eustlb
approved these changes
on 2026-04-23
nit
b9d30be1
_special_token_ids as property and skipped in prepare_config_and_inpu…
8d2e4b76
MoE params in common class
cbd526f2
add _TEXT_MODEL_TESTER_DEFAULTS to avoid divergence
12dfcd04
nit
95b1f202
clearer inits
c2aa666e
_prepare_modality_inputs return dict
5e36c9f8
Merge branch 'main' into tarekziade-audio-test
ca5ff0b0
format
184227cb
split line for readability
d77fbb95
ran python utils/check_modular_conversion.py --fix_and_overwrite
902dbba3
testing auto cancel
dcdead1d
testing auto cancel - part 2
628343dc
Merge branch 'main' into tarekziade-audio-test
4c357680
Merge branch 'main' into tarekziade-audio-test
3f5f4d52
remove comment
c1a47720
udpate granite speech plus tests
93223153
fix test
95da7983
Login to write a write a comment.
Login via GitHub