[MM] Simplify _apply_hf_processor_main (#53275)
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
@@ -385,17 +385,19 @@ def test_processing_correctness(
|
||||
num_batches: int,
|
||||
simplify_rate: float,
|
||||
):
|
||||
if model_id == "google/gemma-3n-E2B-it":
|
||||
pytest.skip("Fix later")
|
||||
if model_id == "OpenGVLab/InternVL2-2B":
|
||||
pytest.skip("Fix later")
|
||||
if model_id == "openvla/openvla-7b":
|
||||
pytest.skip(
|
||||
"OpenVLA uses a custom vLLM processor because its HF remote "
|
||||
"processor is incompatible with current Transformers."
|
||||
)
|
||||
if model_id == "jinaai/jina-reranker-m0":
|
||||
pytest.skip("Fix later")
|
||||
pytest.skip(
|
||||
"JinaVL reverses the multi-modal item order inside "
|
||||
"_apply_hf_processor_main to match its document-before-query "
|
||||
"score template, which breaks the positional correspondence "
|
||||
"between multi-modal kwargs and hashes assumed by the "
|
||||
"processing cache. Needs a fix that keeps both consistent."
|
||||
)
|
||||
if model_id == "mistralai/Voxtral-Mini-4B-Realtime-2602":
|
||||
pytest.skip(
|
||||
"Voxtral Realtime doesn't make use of any place-holder "
|
||||
@@ -403,22 +405,12 @@ def test_processing_correctness(
|
||||
"correctness test as is. Let's revisit adapting this "
|
||||
"test once more realtime models exist."
|
||||
)
|
||||
if model_id == "CohereLabs/cohere-transcribe-03-2026":
|
||||
pytest.skip("Fix later")
|
||||
if model_id.startswith("OpenMOSS-Team/MOSS-Audio-"):
|
||||
pytest.skip(
|
||||
"MOSS-Audio uses a custom processor that dynamically expands "
|
||||
"audio placeholders from processed audio lengths. Its vLLM "
|
||||
"processor paths are covered by test_moss_audio.py."
|
||||
)
|
||||
# TODO: Remove when transformers 5.15.0 is released, which contains
|
||||
# https://github.com/huggingface/transformers/pull/47483.
|
||||
if model_id == "microsoft/VibeVoice-ASR-HF":
|
||||
pytest.skip(
|
||||
"VibeVoice ASR requires audio as a positional argument and hence "
|
||||
"cannot pass the processing correctness test as is. Its generation "
|
||||
"is covered by test_transformers_audio.py."
|
||||
)
|
||||
if model_id == "lmms-lab-encoder/LLaVA-OneVision-2-8B-Instruct":
|
||||
pytest.skip(
|
||||
"LLaVA-OneVision-2 video processing routes frames through custom "
|
||||
|
||||
Reference in New Issue
Block a user